Category Archives: Ai

What if someone told you that a psychedelicdrug could help alleviate symptoms of depression, anxiety, and other mental health conditions?

A growing number of studies suggest that the use of ketamine, which is approved for use as an anesthetic, might also be helpfulin areas where antidepressants have fallen short.This expandingneed, and a changing perception of the value ofpsychedelics as medications,might explain why companies such as Nue Life, a next-generation mental wellness startup, are emerging.

Synthesized in 1962, ketamine is a newer drug compared with others in the class of psychedelics with potential for therapeutic effects. There'secstasy, a.k.a. MDMA, which was first developed in 1912. Then camedimethyltryptamine (DMT) in 1931 andlysergic acid diethylamide(LSD) in 1938.

Ketamine therapy is gainingtraction at a time when the world's mental health crisis is worsening. Between August 2020 and February2021, thenumber of Americanadults with recent symptoms ofan anxiety or depressive disorder grewfrom 36.4 percent to 41.5 percent, according to the Centers for Disease Control and Prevention.TheWorld Health Organization estimatesthatdepression affects around 280 million people across the globe.

As Nue Life CEO Juan Pablo Cappello sees it, themarket need is clear because there are only subpar options available to treat mental health conditions. He explains that antidepressants relieve andmask symptoms -- not to mention that several of these medications have a laundry list of side effects.

Many options available today are indeed dated. The Food and Drug Administration approved Eli Lilly's antidepressant, Prozac, in 1987. One of the most commonly prescribed antidepressants is Zoloft, which wonapproval to be sold in the U.S. in 1990.

And it's no secret that talk therapy canbe expensive, and many patients are not covered by health insurance. But Cappello wonders that instead of masking the symptoms with medication, is there a way to try to address the root cause of suffering and disease?

"Especially during Covid and early lockdown, we saw a huge need for helping women, veterans, and other folks experiencing symptoms of depression, and we wanted to find solutions for those who were not having success with traditional medications and treatments," he says.

Cappello's Chilean grandmother first piqued his interest in plant medicine and indigenous traditions. That interest would not be a fleeting one.Cappello, who is a lawyer by trade, spent the past 20 years supporting psychedelic research. And, after digging into research on psychedelics and wearable tech, Cappello and his team "envisioned combining the two to create apersonalized treatment plan that can drive profound change."

And so Nue Life was born.

The company is funded by the likes of Jack Abraham, of Atomic, and Jon Oringer, of Shutterstock. Right nowNue Lifeoffers at-home ketamine therapy, but plans to expand its offering as research and legislation evolve. The company is not associated with any research institutions or hospitals.

Here's how it works: Prospective patients contact the company throughthe app or by phone fora telemedicine consultation. After undergoing what the company says is a rigorous screening process, candidates for the therapy areprescribed oralketamineand Nue Life sendsa single dose to the patient's house. The patient is required to have a physical sitter who will monitor the ketamine experience with them, which lasts about two hours on average.

"Ketamine is a very safe and effective drug, but it is critical to us that someone has a personal sitter,"Cappello says. Sitters must go through an online training course, though nurse practitioners are also available on demand should a sitter have any questions or concerns. KetaminegotFDA approval as an anesthetic in 1970 buthas yet to receive approval to treat depression and other conditions. But, more recently, the nasal spray Spravato (aclose relative of ketamine) snagged FDA approval in March 2019 for use with treatment-resistant depression.

But ketamine is not without notoriety. Some know it as a horse tranquilizer, others recognize the drug as "Special K," because of its recreational popularity among the rave and club scene. At high doses, the drug can induce seizures or send users down so-called "K-holes," where some describe out-of-body experiences. It's also a knowndate rape drug.But in low doses, it can relax the mind and alleviate pain. Research suggests that ketamine can increase the production of the neurotransmitter glutamate, which helpsneurons regrow lost connections. To date, Nue Life has arranged 22,500 ketamine experiences with no serious negative outcomes.

The company offers two programs, according to its webpage. A one-month program includes six ketamine experiences, which runs $1,250. The four-month program offers18 ketamine treatments for$2,750.

Dolingout ketamineto patients is only the first part of the treatment; the second is to then use data and A.I. to recommend next steps."The real mission is to leverage those experiences to help [patients] do the deep work necessary to heal the underlying root cause of trauma and disease,"Cappello says.

Nue Life measures patient outcomes through data collection,and more data means better patient recommendations. That's where the A.I. comes in. For instance, the A.I. incorporated into the Nue Life platform might recommend that a patient check in with their psychotherapist and offer to schedule a therapy session through an onlinetool. It also uses digital phenotyping, a process that pulls data from smartphones and other devices to study patterns in behavior. The company obtainsprior consent before connecting its app toa patients' wearable devices.

Digital phenotyping can help NueLifevalidate subjective responses of its patients before and after ketaminetherapyby looking at the digital markers of depression, anxiety, and PTSD, the company says. If a patient says they're doing well, but in reality they're not sleeping, exercising, or interacting with their phone in a healthy manner, then the company is able to probe its patients a little bit deeper. "Tech can be a tool that allows us to deliver more effective and personalized experiences to our patients,"Cappello says.

A big-picture question, and an obstacle for Nue Life as well asothercompanies in the space, is how long will it take for society, law enforcement, and regulators to overcome the stigma that's associated with drugs like ketamine. But as more companies, such asthe German startup Atai Life Sciences, getthe green light to conduct clinical trials examining ketamine and depression (Atai is using a non-psychedelic form of ketamine), these treatments will likely grow in popularity. And the segmentis developing, with at least one forecaster pegging the addressable target market for psychedelics in the realm of $10.75 billion by 2027.

"The need and opportunity have never been greater for a new approach to mental health, focusing on deep healing," Cappello says.

View post:

Can Ketamine and AI Improve Wellness? | Inc.com - Inc.

This week San Francisco based company Komodo Health which has a Healthcare Map that gathers de-identified patient data from hundreds of sources announced its partnership with nonprofit Chan Zuckerberg Initiative. Through the partnership, the 50 organizations in Chan Zuckerberg Initiatives Rare as One Network will gain access to Komodos Healthcare Map technology, which they anticipate will allow the networks to catch diagnoses earlier, improve research speed, and close gaps in current care.

Historically patients with rare diseases prove difficult to track and analyze. However, the company believes that its Healthcare Map is expected to close this gap in need in conjunction with AI and analytics to unearth a more complete view of these patient populations.

Specifically, Komodos database includes the de-identified healthcare data of over 330 million patients. When joined with analytics and AI, this Healthcare map can provide insights on populations based on a variety of factors, from demographics to disease stage to geography, ultimately identifying rare disease patients and potential corresponding providers. The hope is that as a result, this will support earlier diagnosis so patients can receive treatment.

Our work with CZIs Rare as One Network gives us an incredible opportunity to empower over 50 patient-led organizations with data-driven software and intelligence that can enhance their work to accelerate research, unlocking treatments and cures for rare disease, said Web Sun, president and co-founder of Komodo Health in the press release. Komodo will now be providing the critical insights needed for these advocacy groups to dig deeper into patient behaviors and patterns of care and put breakthroughs into the hands of patients faster.

The partnership will allow CZIs Rare as One Networks 50 organizations to draw on Komodos technology for a variety of purposes. The technology will support patient advocacy leaders in the organizations to match patients with appropriate researchers and healthcare providers, analyze demographic disparities, and note patient cohorts. As a result, they hope to use the technology to more completely see the diagnostic journey for rare disease patients and to adjust care to shorten those timelines to successful treatments.

We are excited to support the Rare As One Network organizations as they utilize Komodos software to address unanswered questions in their disease areas, said Heidi Bjornson-Pennell, CZI Rare As One Program Manager in a news release. We believe in the power of data and technology to unearth the critical insights needed to address the unmet needs of these patient communities.

Photo: JuSun, Getty Images

Visit link:

AI and analytic tools platform partners with rare disease advocacy groups - MedCity News

The 21st century belongs to Artificial Intelligence (AI) and how it can be used to make our lives easier. We have seen a lot of movies based on the subject of artificial intelligence in recent times. However, most of these movies are intense and show AI in a bad light. But Netflix doesnt seem to be doing that and is giving this sub-genre a chance to rejuvenate.

Titled AI Love You, the Thai movie tells the story of an AI building powered by human feelings. While getting updated, the software faces some problem and the AI falls in love with a real girl. The AI escapes the building and goes into the body of a real man so that he could be with the woman it loves. Will it able to win her affection? Well find this when AI Love You releases on Netflix.

RELATED ARTICLES

Netflix 'Chosen': 5 things you need to know about Danish sci-fi series

Netflix 'Bigbug': 5 things you need to know about the sci-fi comedy

The upcoming movie has an interesting premise and will be a great addition to the platforms catalog that consists of massive projects like Black Mirror, Altered Carbon, and Love, Death, and Robots. The makers of AI Love You will be hoping for a similar kind of reaction as well.

Heres everything we know about the upcoming movie.

The upcoming Thai sci-fi movie will be premiering exclusively on Netflix on Tuesday, February 15. According to the Netflix Press Site, the romantic feature film will be available for streaming at 3:01 am EST. Viewers can download the Netflix app through Googles Play Store and Apples App Store to watch the movie.

The official synopsis reads, A modern love story set in the near future where an AI building is powered by human feelings. Due to a software glitch, it falls in love with a real girl, escapes the building into the body of a real man, and tries to win her affections.

Pimchanok Leuwisetpaiboon will be seen playing the female lead in the movie whose name is Lana. Meanwhile, Mario Maurer plays Bobby in the movie.

Pimchanok Luevisadpaibul

One of the prominent names in the Thai entertainment industry, PimchanokLuevisadpaibul began her acting career in 2009 with the movie Power Kids. In 2010, she made her TV debut with Wai Puan Guan Lah Fun. Since then, she has been considered one of the brightest talents in the industry and garnered a lot of praise for her acting skills.

Crazy Little Thing Call Love, Suddenly Its Magic, Pandin Mahatsajun, Slam Dance, and The Sand Princess are some of the high-end ventures she has appeared on.

Other actors to feature in the movie include Sahajak Boonthanakit (Mr. Wilson), Michael S. New (Alan), and David Asavanond (The Hawk).

The movie is being helmed by the dynamic duo of David Asavanond and Stephan Zlotescu. Meanwhile, it is penned by Philip Gelatt.

Mark Boot, Pttawan Choetkiattikun, Yangyang Li, Chunxiang Long, and Adam Zachary Smith are the producers of the movie.

You can watch the trailer here.

Read more from the original source:

'AI Love You' on Netflix: 5 things to know about the Thai sci-fi romance - MEAWW

BalkansCat/iStock Editorial via Getty Images

Welcome to the qualifying round of the 2022 US Investment Olympics.

The goal of the games is simple: beat the S&P 500, either by generating higher returns or playing dirty and going for higher risk-adjusted returns.

Let the games begin.

Like the 2022 Winter Olympics in Beijing, the US Investment Olympics are not easy to qualify for. Mutual funds are automatically barred from participation: Their fees are just too high for them to have a realistic shot against the S&P 500. Hedge funds have even higher fees and theoretically should be hedged, so they can't compete with the stock market either. In fact, the only securities capable of matching the index are exchange-traded funds (ETFs).

So far, there are eight ETF contestants representing three themes:

Although less expensive than the average mutual or hedge fund, the ETFs have fees of 64 basis points (bps) and are not cheap compared to low-cost index trackers. But then again, top-notch performance isn't free.

Despite their contemporary themes, our ETFs have yet to resonate much with the investment community. Their cumulative assets under management (AUM) are only $700 million, even though some have track records going back to 2012. But then again, who doesn't love cheering for the underdog?

Smart Money, Crowd Intelligence, and AI ETFs AUM, in US Millions

FactorResearch

So how did our eight ETFs fare against the S&P 500? We created equal-weighted indices for the three groups, with Smart Money's track record going back to 2012, AI's to 2016, and Crowd Intelligence's to 2019.

Since all invest in US stocks, they all performed in line with the S&P 500. Some have beaten the benchmark on occasion but not consistently. The judges are not especially impressed.

Outperforming the S&P 500: Smart Money, Crowd Intelligence, and AI ETFs

FactorResearch

Of course, the Olympics, like finance, is all about data and details. Eyeballing an investment's chart is not a particularly scientific approach to performance evaluation. The judges want to know what sort of alpha our competitors have generated since their inception. Smart Money yielded a negative alpha of -3.0% per annum since 2012, Crowd Intelligence -7.2% per year since 2019, and AI -0.9% since 2017.

A cynic might say the smart money isn't that smart, the crowd not that wise, and AI not that intelligent.

Alpha Generation: Smart Money, Crowd Intelligence, and AI ETFs

FactorResearch

But before eliminating all these contestants from medal contention, our judges examine their risk-management characteristics. Our ETFs may not have the longest track records, but they all experienced the last severe stock market shock: the COVID-19 crisis. So how did they do?

Smart Money and Crowd Intelligence fell further than the S&P 500 in March 2020, while AI did marginally better. Perhaps humans are overrated and AI is better at risk management?

Less Downside? Maximum Drawdowns during 2020 COVID-19 Crisis

FactorResearch

Although lower drawdowns may help investors stick to an investment strategy, on a stand-alone basis, they are not especially helpful metrics. After all, cash would outperform in a down market too, but it is unlikely to beat the benchmark over time. So the judges turn to risk-adjusted returns and the Sharpe ratio.

AI beat Smart Money and Crowd Intelligence, but none of our contenders generated higher Sharpe ratios than the S&P 500. That means none of them qualify to advance.

Better Risk-Adjusted Returns? Sharpe Ratios, 2019-2021

FactorResearch

Although these ETFs had distinct flavors, they exhibited similar behavior: In fact, they all outperformed the S&P 500 in 2020. The question is why.

A factor exposure analysis reveals that they have almost identical exposures: negative exposure to value and positive exposure to the size and momentum factors. Our competitors were all overweight outperforming small-cap growth stocks.

Smart money investors like hedge funds may not appreciate that the crowd is picking up the same risk exposure as they are. And they all might be surprised that the AI ETFs are too.

The right factor exposure can help outperform the S&P 500 over time, but it does not resemble alpha. In fact, it is the investment world equivalent of doping. Especially when hidden within thematic products.

Though it wouldn't have mattered in this round, it would have been cause for disqualification.

Thus far, the S&P 500 is beating the field.

Disclaimer: Please note that the content of this site should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute.

Original Post

Editor's Note: The summary bullets for this article were chosen by Seeking Alpha editors.

See more here:

The U.S. Investment Olympics: Smart Money, Crowd Intelligence, And AI - Seeking Alpha

AZEEM AZHAR: Welcome to the Exponential View podcast with me, Azeem Azhar. As an entrepreneur, investor and analyst, Ive been an insider in the technology industry for more than 20 years. During that time, Ive watched exponentially developing technologies change our economies and much more besides, and there is much, much more change to come. Now, I wrote about these dynamics in my book, The Exponential Age, and I dig into them every week in my newsletter, Exponential View, and speak to the people making it happen right here on this podcast. So Ive been really interested in what has been going on in the intersection between breakthroughs in biology and medical research and also in fields of data science and machine learning for quite a while. Researchers and entrepreneurs are making quite impressive progress. One area thats particularly close to home is tackling the questions around human nutrition. What makes for good food for each of us, and todays guest is answering that question and hes doing so by bringing together new research in medicine and biology and nutrition as well as novel techniques in data science, in machine learning, wearable technologies and AI. Jonathan Wolf is a CEO and co-founder of ZOE, a personalized nutrition company that uses some of these state-of-the-art techniques to offer targeted health advice to its customers. One of the most impressive things about ZOE is the speed and scale of its studies. Jonathans background is in high growth consumer tech companies like Yahoo and Criteo, and Matt knows her growth has been put to good use at ZOE, which looks to work on a much quicker time scale than scientific research has been used to. Jonathan Wolf, welcome to Exponential View.

JONATHAN WOLF: Azeem, its a pleasure to be here.

AZEEM AZHAR: I have been fascinated by ZOE for a few years. I think we met to talk about it in 2017 or 2018. The thing that drew me to it was your focus on the problem of personalized nutrition by bringing out insights in something that many of us may know about, the microbiome that lives in our gut, and connecting that unique variability of the microbiome, our own genetics, our own lifestyles with the food that is best for us. So its a really tremendous promise. But for someone who isnt familiar with those ideas, can you just tell us, what are you aiming to do with ZOE?

JONATHAN WOLF: You did a fantastic summary really of what were trying to do, which is, we start with this idea of personalization and the realization that really everything that weve got through healthcare and public health advice for the last 150 years has been sort of this generic one size fits all advice, but it turns out theres vastly more variety and variability between individuals than I think science had historically realized, in part because it just didnt have the ways to measure that. And so what that means is if you can go from generic advice to something thats actually personalized to you, it can have a really radical impact on your health. But the heart of ZOE is, what about if we were actually able to measure you and understand you as an individual and your personal responses, and then give you this guidance thats personalized to you and the support to understand how to actually bring that into your life.

AZEEM AZHAR: One of the key drivers for that is this idea of the microbiome. One thing that I find fascinating is that the microbiome seems to determine quite a lot of our personal responses to our environment. What is the microbiome and why is it important?

JONATHAN WOLF: Let me talk a bit, I guess, about where ZOE came from because I think that sort of helps to answer that question. I have two co-founders and one of my co-founders is Professor Tim Spector and he set up something called the UK Twin Study almost 30 years ago. And so hes been following 12,000 twins. They are the most intensively studied population in the planet. So what was really interesting is he started to look at all of these identical twins and started to discover that one identical twin might get breast cancer and the other doesnt, one identical twin might get obese and the other one didnt. In fact, you just saw over and over again these profoundly different outcomes with individuals who not only had exactly the same genes, they also had exactly the same upbringing for their first 18 years.

JONATHAN WOLF: So really both genes and upbringing were the same. And I think if you looked at the state of the science 20 years ago, people basically said thats impossible. And so Tim and his team started to investigate this. And after quite a few years of research, basically came down to the view there were two things that seemed to really differentiate identical twins. One of them is their gut microbiome. So these are the microbes, the bacteria mainly in your gut that youve already talked about, and the other is your food, your nutrition. And so suddenly we had this explanation for why even identical twins are different. Its not that genes dont matter at all, but it turns out that your microbes have this very important role in shaping you and the food that you eat. This is what takes you to the point, which is really where Tim was when I met him where he started doing these studies where he fed identical twins exactly the same food in a standardized setting. What he found was that you would see profoundly different responses in the hours after eating those meals, even for identical twins. One twin might see this big spike in blood sugar, the other one hardly change. In another twin, you might see this big spike in their blood fat, the other hardly change. And again, they were genetically the same but they have these very different microbes.

AZEEM AZHAR: And just to clarify the point about the response in blood fat or blood sugar is that those can be markers for other types of health risks. So spiking blood sugar after a meal can increase your insulin resistance and take you down a path of sort of diabetes type conditions. And so if you know that thats happening with the food, thats a food that you might want to have in more moderation.

JONATHAN WOLF: Its partly that, absolutely, and its partly a sign of the differences within your metabolic pathways. Actually, the first big study that ZOE did, we took 1,000 people and we did both an in clinic day and then a few weeks at home. What we saw was this vast variation in almost every response. So we saw a tenfold variation amongst a population that was viewed as healthy by doctors in blood sugar, in blood fat, but also in inflammatory markers, really almost anything that you choose to look at. And as you say, if you understand that, then you can start to say, okay, how can I guide an individual in terms of what they should eat shaped by those responses and also shaped by their individual microbes where we can start to understand the links between particular foods and particular microbes and start to put somebody on a different path because the sort of food that we are generally eating today in the 21st century is a very long way away from the food that we evolved to eat and hour after hour, day after day, cause inflammation that starts to lead to sort of negative outcomes, whether those are cardiovascular disease, whether its leading towards diabetes, as youre mentioning, and there are also some really interesting potential links between this and things like your immune system.

AZEEM AZHAR: Yeah. I want to dig into both the science and the connection between ZOE as a company and its desire to do the science in some detail, but lets go back and sort of perhaps in a nutshell encapsulate what are the specifics of what ZOE measures and what it tells you?

JONATHAN WOLF: Today the product is live in the US and well be launching in the UK shortly. What we do is we take this largest nutrition science study in the world and then we allow you to do a very simple at-home test and then we can compare your data using machine learning with the data from all of those studies. What that means is that it starts with a box that arrives at home and you unwrap and an app that takes you through a process that very easily explains how you can send us a microbiome sample, so that is a sample of your poop. How we can measure your blood sugar. So theres a blood sugar sensor called a continuous glucose monitor they can put on your arm with a standardized meal that allows us to understand exactly how you respond to a meal with sugar and fat in it. And we also do a blood test to understand whats happening to your fat, so your lipids. The fat is just as important as sugar in terms of whats going on. We take all of that data and we also get you to track what youre eating for a few days and give us some more context about your health. All of that information, we can then compare with thousands of people who took part in these weeks of clinical study in hospital and everybody else whos been participating in ZOE since. We go away and we analyze that, and that allows us to come back to you and give you this sort of insights report which helps you to understand, okay, how am I unique? How does my blood sugar compare with other people? How does my microbiome compare? But whats really important is, what does it actually mean? So it allows us to feed back to you for every meal that you eat a score that we calculate, which tells you how good that meal is for you and a program that then helps you to understand, okay, how can I shift the way that Im currently eating towards the food that actually really best for me. Everything is really about understanding what is it that you can add into your diet that is actually going to support you and is right for you and that therefore shifts your outcome over the coming weeks, months, and ultimately years.

AZEEM AZHAR: I love the sort of process of measurement. Im a knowledge is power, not an ignorance is bliss kind of person, and that sort of Ive experimented with a number of methods of measuring my biomarkers. I ran some small scale experiments because I was the only person who was being experimented on, but I did try controlled studies and I discovered some really slightly remarkable and off the charts thing. So one thing was that I ate quite a lot of lentils and I found that on days that I would fast, having my lentils at the end of the day would result in a really significant spike in my glucose levels, which is problematic over the long term. Its something you want to avoid. But I also found really curiously that if I had Gulab Jamun, which is a type of Pakistani donut with a rose sugar source, I wouldnt see that much of a postprandial spike in my blood glucose levels to until 30 minutes later. Maybe thats just my own curious makeup, but there was this sense that actually in something as simple as the everyday task of eating my dinner and treating myself to perhaps a sweet afterwards, I was actually flying in fog. My assumptions about what was going to be good for me and what was going to be bad for me were quite inverted.

JONATHAN WOLF: I think thats exactly right. I think maybe if one steps back a bit, we have faced 50 years of confusing and often conflicting nutritional advice. In fact, often you can figure out sort of when somebody was in their 20s from their core nutritional beliefs because thats when you sort of pick up your strongest views about eggs are good or bad or fat is good or bad, or you name it. The challenge is that much of that advice has been changed or its confusing. We live in this world where clearly all of that advice has not worked because we are at our most unhealthy in terms of metabolic disease and obesity that we have ever been.

AZEEM AZHAR: Yeah. But its more complicated than that as well. I mean, not only has the advice changed and some of its not been good, its that what your studies and others have shown is that what might give you a good blood lipid response may not be as good for your cousin or for me or for someone else, that there is enough individual variation for generic advice to be very generic.

JONATHAN WOLF: And part of the reason for why this advice has generally been so poor is that we see this huge variation in response. Imagine that youre measuring everyone in the population and youre trying to say fat is bad, well, actually if what you see is that there are some people in the population who are really struggling to process high levels of fat after a meal and there are other people actually who process that really easily and youre just constantly looking at the average. But I think its important to say we havent found anybody for whom eating broccoli and kale is bad for them. We havent found anybody who drinking Coca-Cola is good for them. So I think one has to recognize theres a very large degree of personalization, but it doesnt suddenly say, Oh, you know what? Theres no advice that applies to anybody. Everybodys completely different. And so your child will be like, Well, Im personalized so I can eat sweets all day. Thats not true, but theres this vast complexity and many of the foods that are most popular continue to be highly controversial and also highly variable. I give you one example. One of the foods where we give the biggest variation advice is a banana. We have some people where the scores for banana are very high and we have other people where the score for banana is very low. But the key is, how do I understand how to put together a meal that works for me? And also the timing of when you eat things. Well, actually your responses change dramatically during the day. So if you measure your responses after breakfast and you look at exactly the same food at lunchtime or dinner, you will see often twofold differences in those meals, and that is itself influenced by what else you eat.

AZEEM AZHAR: When I learned about the microbiome, it was probably from TV ads about 15 years ago when people were selling those little yogurt drinks that claimed to have active bacteria. You started to learn the idea of healthy bacteria. And then its really been perhaps in the last 10 or 12 years that weve started to see popular science books talk about the microbiome, and the first consumer testing services started in the US probably a year or two before ZOE. And even some of the references in the academic papers. You referred to the banana. I remember reading a paper from 2015 about people having a worse glycemic response, a sugar response to a banana than a biscuit. And its only 2015. I mean, thats only seven years ago. So it feels like some of the science has been evolving. And of course the computational capability or the data analytics over the top have also been evolving. So were on quite fresh terrain in a sense, arent we?

JONATHAN WOLF: Correct. I think whats amazing is that 20 years ago, basically there was nothing about the microbiome in the scientific literature. And so the way that people in the field talk about this, its almost as if we suddenly discovered we had a liver and the microbiome is basically that. It turns out this is an organ with trillions of bacteria, so as many bacteria as there are cells in our body, and where we now understand that a very big part of our immune system is actually about managing that bacteria inside our gut. And those bacteria play this crucial role because if were eating food with fiber in it, we dont digest that food directly but the microbes break it down. Its almost like a little chemical factory which is very specialized and it would take particular sorts of food, itll break it down and it will create a lot of byproducts. And those byproducts cross the cell walls in our guts and actually go into our body. They go into our bloodstream and they get distributed to every cell in our body and they influence our metabolism. Another way to think about this is this is a complex ecosystem inside each of us. Theres not one or two bacteria. There are probably 100 different strains of bacteria that you can find at high resolution and probably many more that are so tiny that we cant pick them up even with the incredibly complex shotgun sequencing that we do today. And so why do we have those bacteria? And the answer is, its sort of like a toolbox to deal with all sorts of things that we cant deal with ourselves. These bacteria can change in 30 minutes. So that means if we change our diet between today and tomorrow, nothing about us changes. But actually lets say we are starting to eat something which has lots of dairy in and we werent eating anything with any dairy previously. Actually your whole microbiome can effectively reshape itself, focus on the microbes that can deal with that, break it down and give you a lot of those beneficial byproducts.

AZEEM AZHAR: But you said within 30 minutes.

JONATHAN WOLF: The average lifetime of a bacterias ability to replicate is about 30 minutes. And so what that means is that within 24 hours, you could see a complete reshaping. Relative proportions of bacteria can change dramatically based upon the food that we eat. And so it gives you this sort of flexibility. If you think about it, our DNA is incredibly fixed. You and I share 99.9% of the same DNA. Our microbiome is going to be at most 25% the same. Thats amazing, our microbiome is so different. Another thing to recognize is that all of our microbiomes are now very restricted in comparisons to sort of microbiomes you see from people who live more sort of hunter gatherer lifestyles in the way that our ancestors wouldve done. And so that means that our microbiomes are much more limited.

AZEEM AZHAR: So theres some sense in which theres almost more genetic material kicking around and genetic variability within our microbiomes than within our own DNA?

JONATHAN WOLF: Correct. About a hundred fold more than we have in our own DNA.

AZEEM AZHAR: And so the traditional questions of nature versus nurture, lets sequence my DNA and tell me what my outcomes could be, misses sort of 100X of the other genetic material that might be shaping or influencing my expressions?

JONATHAN WOLF: Correct. I think theres two things that are sort of exciting about that. One is what the ZOE test kit does is we do a full sequencing of your microbiome to be able to identify all of the different microbes. And that includes the fact that your microbes are unique. The other thing to realize though is your microbiome is not fixed. I think one of the things about this whole idea about sort of genes is destiny is its very depressing. Whatever happens to me, it was all fixed by my parents, theres nothing I can do about it. If Im going to get sick, if Im going to put on weight, theres nothing I can do.

JONATHAN WOLF: But actually once you start to understand how important the microbiome is, I think its really exciting because it says your microbiome is changeable. Its malleable. So if I start to change what I eat, then actually I can start to shape my microbiome and that is actually going to affect my metabolism. And so you can start to think about things like how do you eat the food thats right for you, which is going to sort of work with your body, support the good bugs. One of the great Nature Medicine papers that we released last year identified for the first time a set of good and bad bugs. How can you find the food that supports those good bugs, actually start to shift your microbiome and support your health in the long run.

AZEEM AZHAR: Look, the science is quite new and yet were getting these products out to consumers within a couple of decades of the sort of science being understood, which seems quite fast to me. But theres also this issue around the nature of these types of studies. Theres so much variability in the microbiome and also within our genetics and our lifestyle that in order to start to do studies, you need lots and lots of people. I mean, it cant be done with 100 people, and you need to do them over time. Youve got to get people to eat food every day and observe them and monitor them. Im kind of curious about the process by which weve been able to, in a way, accelerate sort of raw signs into something thats being delivered today. Just help me understand, how have we done this as quickly as we have?

JONATHAN WOLF: I think youre right that its very unusual. I think whats unique here and is sort of bringing this sort of consumer internet and sort of technology perspective about the data, which is to say rather than doing a one off study, which is traditionally how science is done, then you go through a long period of commercializing it. How do we think about constantly collecting more data from individuals who are free to choose to participate but want to to sort of improve the science for everybody. And therefore we can push this forward much, much faster than otherwise, and actually really start to make a difference in peoples health.

AZEEM AZHAR: Ongoing data collection turns into a data network effect. So it gives you data that no one else has, that it gives your machine learning and data scientist teams better data from which they can test predictions and they can sort of find new pathways that perhaps werent visible when you had a smaller set of data. So thats pretty exciting. Its exciting, I guess, from the product quality standpoint and its exciting from the strategic moats that you build around your business. But Im quite curious about those landmark papers that you published in Nature Medicine. Its very rare for startups to do that. Sometimes scientists publishes some papers, then they go off and raise money and they commercialize what they had published. Or sometimes the company is very successful and then like DeepMind, manages to get a remit where it can just churn out groundbreaking paper after groundbreaking paper. But you are in a sort of reverse Goldilock Zone with ZOE, which was, you didnt have that groundbreaking predict study, that large scale study, nor had you already had the success that DeepMind had. And yet you were able to sort of pull that off at that beginning. I mean, how did you persuade your investors to essentially invest in science research?

JONATHAN WOLF: Youre quite right. It is very unusual. I think one of our frustrations sometimes is that in this space around sort of health and wellness in which were operating, there are an awful lot of companies that make a lot of claims where the underpinning science for this is very weak. There are various examples, arent there, of particularly west coast Silicon Valley companies making claims that have turned out to lead to criminal charges. So if youre going to do something where youre going to be giving health advice, that you need to do this properly. One of the things we realized was if were going to do this, we need to get many of the top scientists around the world to collaborate with us because this is a very, what they call in science multidisciplinary project.

JONATHAN WOLF: So weve talked about microbiome quite a bit, but we spend an enormous amount of time focusing in on metabolism that includes blood sugar and diabetes. We spend enormous amount of time thinking about inflammation and fats. We need to understand nutrition and nutritional science. We need to understand very large scale human studies, and we need to understand machine learning, all of these different things. And so when we created ZOE from the very beginning, what we said is we are going to create this product, people will have to pay for the product because it costs money in order to do these tests and deliver the results and all the rest of it. But actually we do want to publish as much as possible in order to make this information available to the rest of the world and in order to make sure that actually we can work with these amazing scientists who are really going to be able to help us to come up with the answer.

AZEEM AZHAR: The way you set this up, the scientists are getting incredibly specialists. They dont necessarily see whats happening across in the silo. They perhaps are in funding tracks that take a long time that mean that studies are going to be really slow. Youve brought together an interdisciplinary team of scientists with a product mindset and then an execution mindset that comes out of sort of consumer internet. Is there a message that you have of the model that ZOEs taken that perhaps universities or governments could learn from?

JONATHAN WOLF: The challenge to understand is that most science studies, particularly around nutrition, are very small. So whenever youve seen something on the front page about a particular food being a superfood or liable to kill you or give your cancer, most of those studies have involved 20 or 30 people. That means that the accuracy of that data is very low because theres just not enough information, particularly given this huge personal variation thats going on. Now, the reason why thats happening is not because those scientists are bad scientists. Its because the amount of funding that you can get to support a nutrition science study is very small. I wont go into all the boring details, but the net result is that you end up funding maybe 30 tiny studies rather than one large scale study that can follow this over enough time to really get useful data. And so what is exciting, I think, for many of the scientists working with us is that theyve been able to participate in what is the largest nutrition science study in the world that therefore gives you that depth of data that allows them to answer many, many questions, often questions we hadnt even thought of at the point that we started the study because its got that scale of data.

AZEEM AZHAR: ZOE has given us another example of what this kind of scientific collaboration could look like because you launched a really large scale COVID study as well.

JONATHAN WOLF: Thats right. This was obviously not something that we ever expected to do at the point that we started ZOE. Honestly, I dont think Id ever spent a single moment thinking about infectious diseases. Like so many of us in the world, didnt know anything about it, and now we all feel like were sort of experts one way or the other having lived through it. But in March, 2020, as COVID was hitting, I was sitting here in UK and we went into lockdown within ZOE a few weeks before the government said that we should, we felt like we had to do something. At that point, hospitals in Northern Italy were collapsing and it looked as though in the UK the NHS would buckle. People would die because they wouldnt have ventilators. And so had this fantastic conversation with Tim and George and Tim had this idea. He said, Could we use everything weve built for ZOE nutrition, this amazing app that we use to control everything, our back end, or the rest of these things, to be able to follow COVID symptoms and link that to his twins, his sort of 12,000 twins and try and understand some more about what causes you to get severe COVID or not. I said, Look, Tim, I think its a brilliant idea, but actually if were going to do that, lets not just do it for the twins, lets do it for the entire population. Lets try and see if we can get millions of people to participate in this and then we would have enough data to understand whats the level of COVID around the country. All of this was invisible. There was no testing around the world if you think back to where we were back then. But actually we could use an app, symptom reporting and everything that weve built about sort of this large scale Citizen Science to understand what was going on. We launched it, we had a million people use it within the first 24 hours. Ultimately weve had over four and a half million people use it. And even today, more than a million people are using it every week. And on the back of that, we were able to do just extraordinary science in a way that sort of has never been done before, which is built by this idea of very, very large scale citizen participation using their phones, and this is really just an example of what underpins ZOE.

AZEEM AZHAR: What sort of discoveries did you make as a result of the COVID study?

JONATHAN WOLF: The first paper in Nature Medicine that identified anosmia, which is the loss of smell and taste that were all now used to, was our paper identifying that this was the biggest symptom that you could use to differentiate COVID, I would say at that time. This is no longer true with Omicron, but that is one example. We were able to track hotspots, so where were people getting COVID in that first year when there was no testing and identify that, and then you would see this flow through a few weeks later. And so that data was being passed to the NHS and the government for them to help plan what they were responding for. And then more recently, we have had a whole series of papers in the Lance and elsewhere looking at the effectiveness of vaccines. We were, for example, one of the first people to show you the decay of the efficacy of those first vaccines and therefore the importance of boosters. Most of the Long COVID data that came out has come from us looking at what does Long COVID have.

AZEEM AZHAR: But thats an incredibly rich set of results, which I think speaks to this approach. But lets, if we can, get back to ZOE itself.

JONATHAN WOLF: Of course.

AZEEM AZHAR: Weve talked a lot about the science and we touched briefly on the consumer experience. As I understand it with sort of the ZOE experience, once one can sign up, which you can in the US and hopefully soon in the UK, you get a pack and that pack includes a continuous glucose monitor, a mechanism to take a stool sample. Ive tried this in the past myself. Its not the most fun part of ones say, but it has to be done so you can get your microbiome sequenced. And then there was some testing for blood lipids as well, I think, going on within the pack which gets sort of sent back. But what happens in those first couple of weeks then for me as a consumer, a customer of ZOEs? What does the rest of the experience look like?

JONATHAN WOLF: People are doing this because they want to have some objective around their health. And so the testing period that you talk about is just really the first part of this process. I think what you find in that first period that really excite people is starting to think about your food and starting to see these reactions. And when you get the results back a few weeks later, you start to see for the first time, wow. Many people say, I had no idea that actually I have really bad blood fat responses or that actually my blood sugar responses are quite good. Now I suddenly understand. Been trying to do this ketogenic diet and it always felt like a disaster. Actually it turns out that actually my blood sugar control is better than my blood fat control. So sort of that first stage is the inside. But the next stage, which I think is the most important for people, is starting the program. The program combines little tidbits of sort of teaching to help you to understand step by step, day by day, and a sort of guided program for you to understand how do you take the diet you currently have and how do you start to shift it? So how do you start by identifying the foods that are really good for you?

AZEEM AZHAR: What do these foods look like and how different is one for one person from another?

JONATHAN WOLF: Theyre mainly plants as you would expect. Theyre mainly plants because plants have these two key characteristics, they have fiber and they have polyphenols. These are sort of the critical inputs into your microbes. They vary enormously between individuals because the microbiome are so different. Artichokes and red peppers and apples may be particularly my top boosters, yours could be something completely different. And so part of it is identifying these individual foods. But then the second stage, which is really critical, is understanding how do you combine a meal.

JONATHAN WOLF: Any meal that doesnt have any carbohydrates is not going to impact your blood sugar. So if you were to just eat butter all day, every day, your blood sugar will not move at all. And so for anyone sort of thinking about this, it sort of emphasizes the complexity of whats required to get a good diet because youre actually managing many, many different factors that need to go in. So coming back to ZOE, you have this program that takes you through step by step supported by coaches to understand, firstly, about individual foods, then understanding how do you combine foods, and actually thats really where the magic comes. And theres an app where you can just play with a meal, for example, or where we just give you recipes and recommendations. We just identify whats right for you.

AZEEM AZHAR: How hard is it to distill that science into something that users can understand?

JONATHAN WOLF: I mean, it took us about a year of hard work with the data science team to try and come up with these real time scores. So it was a lot of work because what we give you is just a single score for any food or any meal.

AZEEM AZHAR: Im curious about the challenge of simplification. When we start to boil these things down to single scores, a lot of nuance gets lost. So where have you found that balance between looking at this complex degree of variability, all of this different data expressed across many different dimensions in sort of the mathematical space and coming out and saying, well, therefore this is a 17 or this is a 71.

JONATHAN WOLF: One of the great benefits that we have is because this is personalized to you, were not trying to boil down to one number as the answer for everybody. Thats really hard. Actually we have this enormous amount of information about you already Azeem. And so that allows us to say, well, for you, actually I can make this much simpler. Because I know all of this stuff, the objective here, which is really critical is Actually how do I just make this simple enough that you can figure out how to apply it? So what we are able to do is to boil that down. That doesnt mean we hide the rest of this. Actually if you go onto the app, if you look anything up, it will give you this total score. But then it explains to you, what does this mean for you in terms of your blood sugar, in terms of your blood fat, in terms of your gut health. And so you can drill down into a lot of the data that sits underneath and understand it. But for the vast majority of our customers, what they want to know is, can you actually help me to feel better to know that I am actually on a better path? And so for us, what were trying to do is make simple for you how to understand whats the food for you so that you can get those outcomes. And if you want to dig into lots of detail, you absolutely can.

AZEEM AZHAR: When we look at technologies like this that can be applied for questions of health and nutrition, Im always curious about how long My Netflix subscription has been going on for years and years and years. But when I used a continuous glucose monitor, I used it for about 10 or 12 weeks. And at the end of the 12 weeks, I kind of had a pretty robust measure in my head of where my glucose was going to end up given activity and exercise and food and sleep. When people work with something like ZOE or with the sort of similar products that will come out in the years to come, do you imagine well just do this on an ongoing basis like a Netflix product or would this be something that we do for six months and then come back to in a few years?

JONATHAN WOLF: I hope that we are a lifetime partner rather than sort of the one off fling because I think we know that if you want to make permanent changes to your health, then actually thats an ongoing commitment. I think one of the reasons that I think thats what weve seen already is that theres a lot of interest from our current members to say, How can I retest? I want to understand whether I am managing to shift my microbiome. Am I getting it to a better place? And also as Im aging, how does that change? We havent really touched on this but one of the most interesting things that weve seen through this research is massive changes that women go through menopause. And so for many women, the way that they were eating in their 20s and 30s just stops working at some point during their 40s and then into their 50s, and its not because theyre doing know anything wrong, its because actually thereve been these really profound changes in their metabolism. That means that whats right for you now is not the same as when you were younger. And we see those changes also with men, but its in a much slower and more gradual pace. And so what that means is this is not a sort of one off. I think for many people listening, this will seem familiar. It was definitely true for me. I was 21, I could eat anything. Didnt seem to have any impact. I can tell you that that is not true anymore. And so I think that means that we see people who want to have that permanent partner. And of course also food is very complex. So you say I know what I want to do, but actually then youre in a new environment. You go to a restaurant and youre trying to figure out whats actually right for me. Thats very complicated. How do we partner with you? How do we support you with that? Or how do we help you to understand, hey, youve had all these, how do we deliver you new recipes that continue to be right for you because youre sort of sick of what youve been eating?

AZEEM AZHAR: I want to just come back to this idea of these types of technologies. What I find fascinating about ZOE is that there is a consumerization of health. I mean that in a good way, that I think a lot of us have benefited from being able to track things more effectively. So I wear a Whoop Strap, an Apple watch, and I do these DEXA scans that tell me about my body fat and so on. And Ive found it a really thrilling and productive experience to be able to have access to better data and better insight about my own body. What do you think about how these technologies ought to be deployed into our societies more widely? Are these sorts of things that you would like to see health systems embed, or do you think that this is something that should be a sort of consumer driven choice?

JONATHAN WOLF: I think the question is, whats the evidence? Where are the randomized control trials showing the impact on health? And I think that varies a lot between these different devices. If you think about the simplest device, its just a mobile phone. And so the first thing I think youd really like to see the healthcare system do is fully take that on board. We are still incredibly early in terms of how that links through to our local physician, GP, whatever it is. On these other wearables, so Im going to sound more Luddite than you want me to be Azeem, but I think that for many of these, it is not yet clear what the value of these streams of data are. And so as weve looked at it, what are we collecting? Stuff that we know is really high value. So that blood sugar response, thats really high value. Understanding your microbiome is really high value. Understanding your diet and your background health, incredibly high value. And as we think about what were doing with turning the ZOE COVID study into this wider health study, we believe that being able to get people to continue to share sort of their core symptom data over time, which gives you this insight into your immune system, is going to be incredibly high value. I do believe that some of these other wearables definitely will give us added insight, but in many cases, theres very little today to be able to show that there is incremental impact. I think that if you want to move from stuff thats just like fun internet into health, you got to prove this stuff actually has an effect.

AZEEM AZHAR: Youve had so much experience now working with large numbers of people sticking to a program and not sticking to a program. We think about what takes someone from wherever they are today to a better health outcome. The first is sort of better diagnosis of what the underlying drivers of that current condition is. Then it is sort of a better prescription, a high quality prescription of what they need to do in order to effect the change. And then the third is the question of compliance, which is, am I sticking to the program? Am I walking 45 minutes every day and doing weights and eating well regularly is the bit thats hard, because its always easier to make your to-do list than it is to do it. So, have you learned anything about what the triggers for better, more sort of psychologically safe compliance are for the ordinary woman or man in the street?

JONATHAN WOLF: Yes. And I think the first thing I would say is that is by far the hardest thing. If you ask me like what were my mistakes, five years ago I naively thought the biggest challenge was how do we come up with a personalized guidance thats actually better than the average guidance. And now I realize actually that this question of compliance, how do you get people to actually follow this for long enough to start to feel the results, is in fact the hardest problem. What the advice is itself is relevant because particularly in the nutrition space, most advice is not advice that you can follow in the long run. And so the first thing is how do you actually give good advice that is sustainable? I think the fantastic news is that thats where ZOEs actually giving this advice about food, which if youre eating the right food, there is no quantity. Once youre eating the right food, its going to put you in the right path, the reasons we probably dont have time to go into right now. But there is still this question about, okay, how do I follow it? I think theres a number of things weve learned, but I would say sort of support and accountability seem to be the most critical. One of the key things in the ZOE program is that there is a coach that is available 24/7 who is actually going to support you, who you check in with. Its very similar to the model that is proven with exercise where the model that really works is having some sort of physical trainer who you are going to meet once a week. Because youve got that in the calendar, youre going to go and do it and youre going to do the exercise. And because theyre saying, Come on, do it again, you do it again, like it has this profound difference between something which is just self starting.

AZEEM AZHAR: ZOE obviously is having a lot of success in the area of gut health, something that was hardly acknowledged 20 years ago. Youre using a range of emerging technologies distilled in a nice way for consumers, but you also started to do new science and pivoting your technology to other areas, of course, the UK COVID study being a great example. What do you really think the potential of ZOE could be now? Will it be to focus in that gut health area or do you think that this creates a platform for new types of science?

JONATHAN WOLF: Today, our focus is personalized nutrition. We think that there is the opportunity there to change the health of millions of people around the globe. Its incredibly exciting and incredibly impactful. The gut microbiome and gut health is a core component of that. In the longer run, I absolutely do believe that this is a platform that can impact health more broadly and that as we understand better the interplay between our microbiome and all of this other data, we will start to understand how does that link into people who get cancer, how this might affect dementia, all sorts of other diseases, and indeed start to understand how you might be able to have early interventions that could really change those outcomes. And whether that is probiotics or diet or even more traditional pharmaceutical interventions with this insight, I think that whats clear is if you can start to have very large numbers of people who are sharing their health data in order to try and improve these outcomes, then we could be in a different way of doing this science and hopefully one which is much faster and which starts to understand personalization rather than the sort of generic one.

AZEEM AZHAR: When you did your own study, what was the most pleasant surprise you got, a recommendation for a food that was going to work really well for you?

JONATHAN WOLF: I had basically given up all dairy in my diet before I started, having been told like that its bad for me and bad for people. And interestingly, I now have yogurt as part of my breakfast every morning. I now have a lot of cheese through my week, conscious of some of the environmental impact, so some balance of that. But actually thats really surprising and I have grown to really like it. But interestingly, particularly both yogurt and cheese are fermented products. So that makes a huge difference to the nutritional value of these foods. And actually it turns out that I have very bad blood sugar control, but actually my blood fat control is much better. So actually a big part of my diet has been sort of reducing highly processed carbohydrates, which were a big part of my diet. And so cheese is one of the great joys, yogurt, and the fact that actually if I pick the right fruit, theres a lot of fruits that score really highly for me because of the way that theyre supporting my microbiome. What I would say is if you ask my children, theyd say that basically they all know its drummed into them that they have to eat certain foods because it supports their microbiome because this is setting themselves up for life. And so I think its had a profound impact also on how I think about what my children are eating. And before that, I didnt really put very much thought into it and I think I was giving them terrible food now that I understand this better.

AZEEM AZHAR: Jonathan Wolf, thank you so much for taking the time today.

JONATHAN WOLF: Its been a real pleasure, Azeem. Thank you for listening to me.

AZEEM AZHAR: Well, thanks for listening to this podcast. To become a premium subscriber of my weekly newsletter, go to http://www.exponentialview.co/listener. Youll find a 20% off discount there. And stay in touch. Follow me on Twitter, Im @Azeem, A-Z-E-E-M. This podcast was produced by Mischa Frankl-Duval, Fred Casella and Marija Gavrilov. Bojan Sabioncello is the sound editor.

See the original post here:

The Future of Healthcare: Personalization and AI (with ZOE's Jonathan Wolf) - Harvard Business Review

The skyrocketing scale of AI has been hard to miss in recent years. The most advanced algorithms now have hundreds of billions of connections, and it takes millions of dollars and a supercomputer to train them. But as eye-catching as big AI is, progress isnt all about scalework on the opposite end of the spectrum is just as crucial to the future of the field.

Some researchers are trying to make building AI faster, more efficient, and more accessible, and one area ripe for improvement is the learning process itself. Because AI models and the data sets they feed on have grown exponentially, advanced models can take days or weeks to train, even on supercomputers.

Might there be a better way? Perhaps.

A new paper published on the preprint server arXiv describes how a type of algorithm called a hypernetwork could make the training process much more efficient. The hypernetwork in the study learned the internal connections (or parameters) of a million example algorithms so it could pre-configure the parameters of new, untrained algorithms.

The AI, called GHN-2, can predict and set the parameters of an untrained neural network in a fraction of a second. And in most cases, the algorithms using GHN-2s parameters performed as well as algorithms that had cycled through thousands of rounds of training.

Theres room for improvement, and algorithms developed using the method still need additional training to achieve state-of-the-art results. But the approach could positively impact the field if it reduces the energy, computing power, and cash needed to build AI.

Although machine learning is partially automatedthat is, no one tells a machine learning algorithm exactly how to accomplish its taskactually building the algorithms is far more hands on. It takes a good deal of skill and experience to tweak and tune a neural networks internal settings so that it can learn a task at a high enough level to be useful.

Its almost like being the coach rather than the player, Demis Hassabis, co-founder of DeepMind, told Wired in 2016. Youre coaxing these things, rather than directly telling them what to do.

To reduce the lift, researchers have been developing tools to automate key steps in this process, like, for example, finding the ideal architecture for a new algorithm. A neural networks architecture is the high level stuff, like the number of layers of artificial neurons and how those layers link together. Finding the best architecture takes a good bit of trial and error, and automating it can save engineers time.

So, in 2018, a team of researchers from Google Brain and the University of Toronto built an algorithm called a graph hypernetwork to do the job. Of course they couldnt actually train a bunch of candidate architectures and pit them against each other to see which would come out on top. The set of possibilities is huge, and training them one by one would quickly get out of hand. Instead, they used the hypernetwork to predictthe parameters of candidate architectures, run them through a task, and then rank them to see which performed best.

The new research builds on this idea. But instead of using a hypernetwork to rank architectures, the team focused on parameter prediction. By building a hypernetwork thats expert at predicting the values of parameters, they thought, perhaps they could then apply it to any new algorithm. And instead of starting with a random set of valueswhich is how training usually beginsthey could give algorithms a big head start in training.

To build a useful AI parameter-picker, you need a good, deep training data set. So the team made onea selection of a million possible algorithmic architecturesto train GHN-2. Because the data set is so large and diverse, the team found GHN-2 can generalize well to architectures its never seen. They can, for example, account for all the typical state-of-the-art architectures that people use, Thomas Kipf, a research scientist at Google Researchs Brain Team in Amsterdam, recently told Quanta. That is one big contribution.

After training, the team put GHN-2 through its paces and compared algorithms using its predictions to traditionally trained algorithms.

The results were impressive.

Traditionally, algorithms use a process called stochastic gradient descent (SGD) to gradually tune a neural networks connections. Each time the algorithm performs a task, the actual output is compared to the desirable output (is this an image of a cat or a dog?), and the networks parameters are adjusted. Over thousands or millions of iterations, training nudges an algorithm toward an optimal state where errors are minimized.

Algorithms using GHN-2s predictionsthat is, with no training whatsoevermatched the accuracy of algorithms that were trained with SGD over thousands of iterations. Crucially, however, it took GHN-2 less than a second to predict a models parameters, whereas the traditionally trained algorithms took some 10,000 times longer to reach the same level.

To be clear, the performance the team achieved isnt yet state-of-the-art. Most machine learning algorithms are trained much more intensively to higher standards. But even if an algorithm like GHN-2 doesnt get its predictions just righta likely outcomestarting with a set of parameters that is, say, 60 percent of the way there is far superior to starting with a set of random parameters. Algorithms would need fewer learning cycles to reach their optimal state.

The results are definitely super impressive, DeepMinds Peter Velikovi told Quanta. They basically cut down the energy costs significantly.

As billion-parameter models give way to trillion-parameter models, its refreshing to see researchers crafting elegant solutions to complement brute force. Efficiency, it seems, may well be prized as much as scale in the years ahead.

Image Credit: Leni Johnston / Unsplash

Looking for ways to stay ahead of the pace of change? Rethink whats possible. Join a highly curated, exclusive cohort of 80 executives for Singularitys flagship Executive Program (EP), a five-day, fully immersive leadership transformation program that disrupts existing ways of thinking. Discover a new mindset, toolset and network of fellow futurists committed to finding solutions to the fast pace of change in the world. Click here to learn more and apply today!

Read this article:

This AI Learned the Design of a Million Algorithms to Help Build New AIs Faster - Singularity Hub

Artificial intelligence (AI) adoption is on the rise. According to a recent McKinsey survey, 55 per cent of companies use artificial intelligence in at least one function, and 27 per cent attribute at least 5 per cent of earnings before interest and taxes to AI, much of that in the form of cost savings.

As AI will dramatically transform nearly every industry it touches, its no surprise that vendors and enterprises are looking for opportunities to deploy AI everywhere they can. But not every project can benefit from AI and attempting to apply AI inappropriately can not only cost time and money but also sour employees, customers, and corporate leaders on future AI projects.

The key factors for determining whether a project is suitable for AI are business value, availability of training data, and cultural readiness for change. Heres a look at how to ensure those criteria are in line for your proposed AI project before your foray into artificial intelligence becomes a sunk cost.

Data scientists in particular gravitate toward an AI-first approach, says Zack Fragoso, data science and AI manager at pizza chain Dominos, which has more than 18,000 locations in over 90 countries around the world. But you cant apply AI everywhere.

Despite being a very traditional line of business, Dominos has been embracing change especially during the pandemic. Customers now have 13 digital ways to order pizzas, and the company generated more than 70 per cent of sales through digital ordering channels in 2020. That has opened up a lot of opportunity for making good on the promise of AI.

The key for Dominos in applying AI, Fragoso says, has been taking a simple approach. At the end of the day, the simple solution runs faster, performs better, and we can explain it to our business partners, he says. Explainability is a big part of it the more people understand the tools and methods we use, the easier it is to gain adoption.

The approach itself is simple: If theres a business problem that needs solving, Dominos looks at the simplest, most traditional solution, and then, if we go up from there, there needs to be a value-add in the performance of the model, Fragoso says.

For example, predicting how long it takes to cook a pizza and put it in a box is simple. We pull that right from our operations research. You can plug in the oven times, he says. But there are some problems that can only be solved with AI, he adds, such as those requiring image recognition or natural language processing.

For example, last year, Dominos ran a loyalty program that rewarded customers for eating pizza any pizza, from any pizza maker. We built a pizza classifier using millions of pictures of different kinds of pizza and put it into an app, Fragoso says.

That project offered two types of business value. First, it enhanced the customer experience, he says. Second, it created a collection of pizza images that the company then used to detect pizza quality and temperature. It was a really great full-circle AI project, he says.

A more practical AI project Dominos undertook was a predictor aimed at improving the accuracy of its pizza tracker, as customers want to know when exactly to come to the store to pick up their food, or when to expect their delivery to arrive, Fragoso says. Adding machine learning to the traditional if-then coding of Dominos pizza tracker resulted in a 100 per cent increase in accuracy, he says.

In building the model, Dominos stuck with its simplest-first principle. The first iteration was a simple regression model, he says. That got us close. Then a decision tree model, where we could look at more facets. Then we actually moved to a neural net because we could capture some of the same variables as in the decision tree but the neural net produces the answer faster. We want our customer experience on the website to be really fast.

There is a place for machine learning, says Sanjay Srivastava, chief digital officer at Genpact, particularly when a company is looking to build processes that are continually improving based on experience. But sometimes all thats needed is a simple correlation, which can be obtained from basic statistical modeling.

Ten-year-old practices around random forests and other statistical tool kits can get you the answer much faster and much cheaper than building a whole MLOps team around it, Srivastava says. You have to know when to fall back to existing techniques that are much simpler and much more effective.

One common area in which AI is often pitched a solution but is usually overkill is in chatbots, he says: In some scenarios, it makes sense. But in 90 per cent of scenarios you know the questions that are going to be asked because you can look at the questions that have been asked in the last three years and you know the answer to every question. Turns out, 90 per cent of chatbots can get away with simple question-and-answer pairs.

Any finite set of data can be fitted to a curve. For example, you can take previous years winning lottery numbers, and come up with a model that would have predicted them all perfectly. But the model will still be no better at predicting future winnings because the underlying mechanism is completely random.

The COVID-19 pandemic has been a prime example of how this happens in real life. There was no way to predict where lockdowns were going to lead to factory shutdowns, for example. As a result, companies saw a decline in the revenue gains they saw in many areas, according to McKinseys state of AI survey.

For example, 73 per cent of respondents saw revenue increases in strategy and corporate finance last year, while only 67 per cent did so this year. The difference was even more stark in supply chain management. Last year, 72 per cent saw revenue increases in this area but only 54 per cent did this year.

The fundamental characteristic of AI or machine learning is that youre using history to inform, says Donncha Carroll, partner in the revenue growth practice at Axiom Consulting Partners. You are wedded, chained, handcuffed by history. AI is good in circumstances where history is likely to repeat itself and youre okay with history repeating itself.

For example, he says, some of his clients have tried to use AI to predict future revenue. But often, revenue is influenced by factors that cant be predicted, that cant be controlled, and that the company doesnt have any data for. And if some of those factors have an outside impact on results, it can throw off the entire model.

Then it makes no sense to choose AI, he says. Are you going to invest hundreds of thousands of dollars in a solution that can be immediately made irrelevant by a change in one variable?

AI can still have a role here, he says, in helping to model various scenarios, or in surfacing insights that might not be otherwise apparent. Your likelihood of success goes up if your focus is more narrow.

AI will also fall short if the very presence of the AI changes the behavior of the system. For example, if AI is used to filter out hateful speech, people quickly learn what patterns the AI looks for and word things so that they get through the filters.

The best minds in the world have been trying to solve these problems and they have not been successful, Carroll says.

Kearney partner Bharath Thota once worked with a US$30-billion-plus global consumer products and goods conglomerate. The CFO leadership team wanted better visibility into the conglomerates financial metrics so they could see whether their growth was going up and down. The existing process was that they got PDFs of reports 30 days after the reporting period closed.

The data science team applied AI to forecast what the numbers would look like. They had good intent, says Thota. They wanted to provide the leadership with a futuristic view.

The mistake they made was in the financial data they were feeding into the algorithm. The financial analysts feeding in that data had to make a lot of assumptions, and so the data set wound up containing lots of individual biases.

The leadership was excited, says Thota. They had something forward-facing, not rearview-facing. But when the quarter ended, and they looked back at those predictions, they were completely off.

The entire project took months, says Thota. They had to figure out how to build this thing, do the architecture, research AI platforms, get everything to work together.

When a project like that fails, people lose interest and confidence in AI, he says. For this particular company, the solution was simply to build the CFO leadership team a financial dashboard that gave them the metrics they needed, when they needed them.

Eventually, Thota says, some AI was used as well, in the form of natural language generation, to automatically provide key insights into the data to the executives in plain-English terms.

It was a visibility problem, he says. And there was a simple solution to provide that visibility.

Read more on the next page...

Error: Please check your email address.

See more here:

How to know when AI is the right solution - ARNnet

Did you miss a session from the Future of Work Summit? Head over to ourFuture of Work Summit on-demand libraryto stream.

If you ask most people to name the key technologies of the metaverse, theyll usually focus on the eyewear and graphics engines. If theyre sophisticated, theyll also bring up 5G and blockchain. But those are the nuts and bolts of our immersive future. The technology that will pull the strings, creating and manipulating our experience, is AI.

Artificial intelligence will soon become one of the most important, and likely most dangerous, aspects of the metaverse. Im talking about agenda-driven artificial agents that look and act like any other users but are virtual simulations that will engage us in conversational manipulation, targeting us on behalf of paying advertisers.

This is especially dangerous when the AI algorithms have access to data about our personal interests, beliefs, habits and temperament, while also reading our facial expressions and vocal inflections. Such agents will be able to pitch us more skillfully than any salesman. And it wont just be to sell us products and services they could easily push political propaganda and targeted misinformation on behalf of the highest bidder.

And because these AI agents will look and sound like anyone else in the metaverse, our natural skepticism to advertising will not protect us. For these reasons, we need to regulate some aspects of the coming metaverse, especially AI-driven agents. If we dont, promotional AI-avatars will fill our lives, sensing our emotions in real time and quickly adjusting their tactics for a level of micro-targeting never before experienced.

But of all the technologies headed our way, its the Elf that could be the most subtle form of coercion. These Electronic Life Facilitators will be the next generation of digital assistants like Alexa and Siri. But they wont be disembodied voices; theyll be anthropomorphized personas customized for each user. And because the metaverse will ultimately be an augmentation layer on the real world, these digital elves will follow us everywhere, whether were shopping, working, or just hanging out.

And like the marketing agents described above, these elves could have an agenda, nudging us towards actions and activities, products and services, even views and beliefs on behalf of a paying advertiser. And they wont be like the crude chatbots of today, but embodied characters well come to think of as trusted figures in our life a mix between a familiar friend, a helpful advisor, and a caring therapist. But your elf will know you in ways no friend ever could, for it could monitor your daily life down to your blood pressure and respiration rate (via your smart watch).

Yes, this sounds creepy and invasive, which is why platform providers will likely make them cute and non-threatening, with innocent features that seem more like a magical character than a human-sized assistant following you around. This is why I prefer the word elf to describe them, as they might appear to you as a fairy or gremlin, hovering over your shoulder a small character that can whisper in your ear or fly out in front of you to draw attention to things in your augmented world it wants you to focus on.

There are many positive uses of such technology, but when controlled by for-profit corporations, AI agents can too easily coerce us, steering us towards products and services without us even realizing. After all, the metaverse itself is designed to fool our senses when combined with the power of AI the dangers are very real. I raise these issues in hope the industry pushes for meaningful regulation before the problems become so ingrained that we accept them as inevitable. After all, we deserve a magical metaverse, free of excessive monitoring and hidden manipulation.

Louis B. Rosenberg is a computer scientist, entrepreneur, and prolific inventor. Thirty years ago while working as a researcher at Stanford and Air Force Research Laboratory, Rosenberg developed the first functional augmented reality system. He then founded one of the early virtual reality companies (Immersion Corp) and one of the early augmented reality companies (Outland Research). Hes currently founder and CEO of swarm intelligence company Unanimous AI.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn More

See the original post here:

The danger of AI micro-targeting in the metaverse - VentureBeat

Written by Jackson Barnett Jan 26, 2022 | FEDSCOOP

The Department of Defenses new top artificial intelligence office will reach initial operational capability February 1, the Navys chief data officer said Wednesday.

The new chief digital and AI officer will oversee the DODs chief data officer, Joint Artificial Intelligence Center and Defense Digital Service and serve as an integrator for their efforts. Navy CDO Tom Sasala said that the new office will help the DOD push forward on data initiatives and be helpful to the work his team has been doing to bring data-based decision making to the Navy.

Ithink thats a really important component, he said of the office during an webinar presented by C4ISRnet.

Sasala said he met Tuesday with DOD Chief Data Officer Dave Spirk and Jim Mitre, the official tasked with standing up the office, to offer specific recommendations on the new role.

The DOD has been shuffling its management of AI and emerging technology initiatives over the past few years. Most recently, the JAIC was moved to report directly to the deputy secretary of defense from the CIOs office by the fiscal 2021 National Defense Authorization Act. The CDAO role is being created to ensure AI and data initiatives receive more attention, given the vast portfolio of the departments chief operating officer.

The CDAO will serve as the Departments senior official responsible for strengthening and integrating data, artificial intelligence, and digital solutions in the Department, the memo creating the positions signed by the deputy secretary states.

The CDAOis expected to reach full operational capability by June 2022.

A DOD spokesperson said: We are looking both internally and external to the Department and actively reviewing potential candidates to ensure we have the right person to drive the kind of progress we need.

Thisstory was updated to include commentfrom the Department ofDefense.

The rest is here:

New DOD Chief Digital and AI Office to start work by Feb. 1 - FedScoop

President Vladimir Putin declared 2021 to be the Year of Science and Technology in Russia, with November named as the month of artificial intelligence (AI), signaling Russian leaderships strong interest in this umbrella term. The Russian defense sector is particularly captivated by the opportunities associated with AI-based technologies. In recent years, AI, robotics, as well as the further integration of automation and autonomy into weapons systems and military decision-making have all been highlighted as priorities for modernizing the Russian armed forces.

In 2017, Putin famously said that artificial intelligence is the future, not only for Russia, but for all humankind Whoever becomes the leader in this sphere will become the ruler of the world. Quoting these words, analysts often attribute Russias development, testing, and use of weaponized AI to the necessity of competing in the so-called global AI race or the global tech race with the current leading AI developers: the United States and China.

While the perceived need to compete and catch up is part of Russias motivations, its interest in military AI should not only be attributed to a quest for relative power. Understanding the depth and complexity of the debates surrounding AI, autonomy, and automation in Russia requires an examination of discussions about its strategic implications for the Russian army, the perceived benefits and risks of autonomy, and, more broadly, the importance of technological modernization and innovation for Russias place in the world.

This report aims to provide an overview of the different conceptions and motivations, both oriented towards domestic and international audiences, that have been and are guiding Russian political and military leaderships in their ambitions to pursue weaponized AI. First, it outlines the various factors, both external and internal, behind the quest for pursuing AI, autonomy, and automation in the Russian military. Second, it presents some of the Russian plans in this area, what is known about their capabilities, and the challenges to strengthening them. Third, it dives into Russian debates on autonomy, and especially autonomous weapons systems, as well as discussions on the ethics of developing so-called killer robots, or autonomous combat robots ( , or ), a term often used in the Russian-language literature.

The analysis is based on a survey of open-source materials, including media reports, press releases, official statements and speeches, peer-reviewed articles and think tank reports, as well as publications in Russian military journals. The author would like to present it as the first step in an ongoing doctoral research project, as well as a contribution to the emerging English-language literature on how weaponized AI is perceived in Russia.

Continue Reading

Read the original post:

Russian Perceptions of Military AI, Automation, and Autonomy - Foreign Policy Research Institute