Regenerative Medicine – WOMEN IN MEDICINE LTD

Posted on by

This site uses cookies to improve performance. If your browser does not accept cookies, you cannot view this site.

There are many reasons why a cookie could not be set correctly. Below are the most common reasons:

This site uses cookies to improve performance by remembering that you are logged in when you go from page to page. To provide access without cookies would require the site to create a new session for every page you visit, which slows the system down to an unacceptable level.

This site stores nothing other than an automatically generated session ID in the cookie; no other information is captured.

In general, only the information that you provide, or the choices you make while visiting a web site, can be stored in a cookie. For example, the site cannot determine your email name unless you choose to type it. Allowing a website to create a cookie does not give that or any other site access to the rest of your computer, and only the site that created the cookie can read it.

Link:
Regenerative Medicine - WOMEN IN MEDICINE LTD

Generative AI Is Totally Shameless. I Want to Be It – WIRED

Posted on by

AI has a lot of problems. It helps itself to the work of others, regurgitating what it absorbs in a game of multidimensional Mad Libs and omitting all attribution, resulting in widespread outrage and litigation. When it draws pictures, it makes the CEOs white, puts people in awkward ethnic outfits, and has a tendency to imagine women as elfish, with light-colored eyes. Its architects sometimes seem to be part of a death cult that semi-worships a Cthulu-like future AI god, and they focus great energies on supplicating to this immense imaginary demon (thrilling! terrifying!) instead of integrating with the culture at hand (boring, and you get yelled at). Even the more thoughtful AI geniuses seem OK with the idea that an artificial general intelligence is right around the corner, despite 75 years of failed precedentthe purest form of getting high on your own supply.

So I should reject this whole crop of image-generating, chatting, large-language-model-based code-writing infinite typing monkeys. But, dammit, I cant. I love them too much. I am drawn back over and over, for hours, to learn and interact with them. I have them make me lists, draw me pictures, summarize things, read for me. Where I work, weve built them into our code. Im in the bag. Not my first hypocrisy rodeo.

Theres a truism that helps me whenever the new big tech thing has every brain melting: I repeat to myself, Its just software. Word processing was going to make it too easy to write novels, Photoshop looked like it would let us erase history, Bitcoin was going to replace money, and now AI is going to ruin society, but its just software. And not even that much software: Lots of AI models could fit on a thumb drive with enough room left over for the entire run of Game of Thrones (or Microsoft Office). Theyre interdimensional ZIP files, glitchy JPEGs, but for all of human knowledge. And yet they serve such large portions! (Not always. Sometimes I ask the AI to make a list and it gives up. You can do it, I type. You can make the list longer. And it does! What a terrible interface!)

What I love, more than anything, is the quality that makes AI such a disaster: If it sees a space, it will fill itwith nonsense, with imagined fact, with links to fake websites. It possesses an absolute willingness to spout foolishness, balanced only by its carefree attitude toward plagiarism. AI is, very simply, a totally shameless technology.

As with most people on Earth, shame is a part of my life, installed at a young age and frequently updated with shame service packs. I read a theory once that shame is born when a child expects a reaction from their parentsa laugh, applauseand doesnt get it. Thats an oversimplification, but given all the jokes Ive told that have landed flat, it sure rings true. Social media could be understood, in this vein, as a vast shame-creating machine. We all go out there with our funny one-liners and cool pictures, and when no one likes or faves them we feel lousy about it. A healthy person goes, Ah well, didnt land. Felt weird. Time to move on.

AI is like having my very own shameless monster as a pet.

But when you meet shameless people they can sometimes seem like miracles. They have a superpower: the ability to be loathed, to be wrong, and yet to keep going. We obsess over themour divas, our pop stars, our former presidents, our political grifters, and of course our tech industry CEOs. We know them by their first names and nicknames, not because they are our friends but because the weight of their personalities and influence has allowed them to claim their own domain names in the collective cognitive register.

Are these shameless people evil, or wrong, or bad? Sure. Whatever you want. Mostly, though, theyre just big, by their own, shameless design. They contain multitudes, and we debate those multitudes. Do they deserve their fame, their billions, their Electoral College victory? We want them to go away but they dont care. Not one bit. They plan to stay forever. They will be dead before they feel remorse.

AI is like having my very own shameless monster as a pet. ChatGPT, my favorite, is the most shameless of the lot. It will do whatever you tell it to, regardless of the skills involved. Itll tell you how to become a nuclear engineer, how to keep a husband, how to invade a country. I love to ask it questions that Im ashamed to ask anyone else: What is private equity? How can I convince my family to let me get a dog? It helps me understand whats happening with my semaglutide injections. It helps me write codehas in fact renewed my relationship with writing code. It creates meaningless, disposable images. It teaches me music theory and helps me write crappy little melodies. It does everything badly and confidently. And I want to be it. I want to be that confident, that unembarrassed, that ridiculously sure of myself.

See the article here:

Generative AI Is Totally Shameless. I Want to Be It - WIRED

Posted in Ai

Your Path to Vibrant Living: Simple Habits for Lasting Wellness

Posted on by

Photo by Unsplash

Living a vibrant life is something we all aspire to, but it can be difficult to know where to start. With so much information out there about health and wellness, it can be overwhelming to try and figure out what habits will truly make a lasting impact. However, by focusing on simple, daily habits, we can create a foundation for vibrant living that will serve us well in the long run.

By incorporating these habits into our lives, we can improve our physical, mental, and emotional well-being and live life to the fullest. In today’s article from Immortality Medicine, we take a look at some proven techniques for facilitating whole-body wellness.

Don’t Overlook Eye Health

One of the most overlooked aspects of health is our vision. Regular visits to the eye doctor are crucial in maintaining good eye health and catching any potential issues early. Whether you wear glasses or contact lenses, annual eye exams can help detect eye diseases such as glaucoma and macular degeneration at an early stage. Additionally, these check-ups can also reveal other health problems like diabetes and high blood pressure, making them a vital part of your overall health care routine in your journey towards healthy living.

Consider Sermorelin Injections

Sermorelin injections offer numerous health benefits for adults seeking a healthier lifestyle. This therapy aids in the body's production of human growth hormone (hGH), which can be beneficial for those with hGH deficiency. Studies have suggested that sermorelin therapy is safe and effective for restoring normal growth hormone levels in adults. Not only does it improve brain function and mood, but it also enhances energy levels and the overall feeling of mental well-being. In terms of physical health, sermorelin therapy can improve exercise performance, increase muscle strength, and even contribute to heart health. It has also been associated with an increase in libido and lean body mass.

Manage Symptoms of Low Testosterone

Managing symptoms of low testosterone is a multifaceted approach that requires both lifestyle adjustments and potential medical intervention. Regular exercise, a balanced diet rich in lean proteins, fruits, vegetables and whole grains, and sufficient sleep are all crucial lifestyle habits that can help boost testosterone levels naturally. It's also important to maintain a healthy weight, as excess body fat can affect hormone production.

Additionally, stress management techniques such as meditation or yoga can help, as chronic stress is known to negatively impact hormone balance. In some cases, medical treatments may be necessary. These can include testosterone replacement therapy, which should be administered under the supervision of a healthcare provider. Always consult with a healthcare professional for personalized advice.

Learn to Love Your Stature

Height sensitivity can often impact one's self-esteem. However, it's important to remember that height doesn't define your worth or capabilities. Many successful individuals, including some of your favorite celebrities, may share your stature. Use this as a reminder that everyone is unique, and it's this uniqueness that makes us who we are. Embracing and accepting yourself, including your height, is a cornerstone of mental health, which plays a crucial role in the path of healthy living.

Incorporate Stretching into Your Daily Routine

Physical activity is an integral part of healthy living. Regular movement, even in small amounts, can greatly improve your health and well-being. A simple habit to start with is stretching throughout the day. Regular stretching can increase flexibility, improve posture, reduce stress, and boost circulation. Try incorporating a few minutes of stretching into your routine—when you wake up, during lunch breaks, or before going to bed. Every bit contributes to your overall wellness.

File Organization Is Key for Those with a Medical Condition

Keeping your medical files organized is crucial, especially if you have a medical condition that requires ongoing treatment. With multiple doctors, specialists, and healthcare providers involved in your care, it can be challenging to keep track of all the information. However, having your medical files organized and easily accessible can make a significant difference in the quality of care you receive.

One way to keep your files organized is by converting them to PDFs, which are easy to access, edit, and share, and can be stored on your computer or in a cloud-based storage system for safekeeping. You can try this page to find out more about making adjustments or even splitting your files.

The journey to sustainable wellness and healthy living lies in the accumulation of small, consistent habits. Remember, it's not about perfection, but progress. So, meet yourself where you are, use what you have, and do what you can. Your path to healthy living starts with a single step.

Article by Diane Harrison.

Google’s new LearnLM AI model focuses on education – The Verge

Posted on by

LearnLM, a family of AI models based on Googles other large language mode, Gemini, was built to be an expert in subjects, find and present examples in different ways like a photo or video, coach students while studying, and, in Googles words, inspire engagement.

Google has already integrated LearnLM into its products, bundling it with other services like Google Search, Android, YouTube, and the Gemini chatbot. For example, customers can use Circle to Search on Android to highlight a math or physics word problem, and LearnLM will help solve the question. On YouTube, while watching a lecture video, viewers can ask questions about the video, and the model will respond with an explanation.

Google says LearnLM was specifically fine-tuned to only respond and find answers based on educational research. In other words, LearnLM will not help someone plan a trip or find a restaurant.

Google says its working with educators in a new pilot program on Google Classroom so they can use LearnLM to simplify lesson planning. The company is also experimenting with Illuminate, a platform that will break down research papers into short audio clips with AI-generated voices. Ideally, this will help students understand complex information better.

Google also partnered with Columbias Teachers College, Arizona State University, NYU Tisch, and Khan Academy to provide feedback to LearnLM.

Read more from the original source:

Google's new LearnLM AI model focuses on education - The Verge

Posted in Ai

Verizon Expects to Double Network Thanks to AI Demand – PYMNTS.com

Posted on by

Verizon Communicationsis reportedly placing a significant bet on artificial intelligence (AI) as its next major growth driver.

The company, which is the largest retail wireless carrier in the United States, believes in the potential of AI to double its network over the next five years, Verizon Consumer CEOSowmyanarayan Sampathtold Bloomberg in aninterviewposted Tuesday (May 14).

While video traffic has been the primary driver of growth for Verizons network in recent years, Sampath foresees AI demand surpassing it soon, according to the report. The massive amounts of data generated by AI tools require efficient transmission from data centers to end-users for analysis and back to data centers. To meet this increased demand, Verizon aims to provide a robust network infrastructure.

Verizon also expects AI to transform various aspects of its operations, the report said.

The company has already integrated AI software fromGoogleinto its customer service operations. This technology has been deployed to empower 40,000 customer relations agents, enabling them to better serve customers, perthe report.

For instance, the AI tool listens in on representatives calls, tracking human sentiment and tone. If a customer appears frustrated, a supervisor can receive a notification and intervene to assist, according to the report.

In addition, routine processes like bill inquiries and payment details will be automated, allowing customer service agents to focus on more complex tasks, such as assisting customers with international travel or account changes, the report said.

Verizon also plans to leverage AI to offer personalized recommendations and curated offers to customers in real-time, per the report.

The company has strategically positioned its infrastructure to handle the anticipated surge in AI-related traffic, the report said. It has invested in spectrum acquisition to support the increased demand for data transmission.

During the companys most recent earnings call, which was held April 22, Verizon CEOHans Vestbergfocused in part on the companysAI strategy, saying that it needs to create an AI-centric revenue stream by commercializing the companys networks mobile edge computing capabilities.

Generative AI workloads represent a great long-term opportunity for us, Vestberg said. As we expand our network and increase our performance advantage, were also making Verizon a more efficient organization.

Original post:

Verizon Expects to Double Network Thanks to AI Demand - PYMNTS.com

Posted in Ai

Google I/O just showed me how to live the laziest life through AI – TechRadar

Posted on by

The dirty secret about Google Gemini (and probably all AI) is that it's built for lazy people like me. At the Google I/O keynote on Tuesday, Google spent hours showing us the myriad ways Gemini and its associated technologies could pick up the slack for us or, more specifically, me, the laziest person you'll ever meet.

I know what you're thinking, "No, no, Lance, you seem like a real hardworking guy." It's a lie. I've spent my decades-long career finding shortcuts - and AI is my white whale. Instead of riding a giant vessel into an uncertain fate, Google has handed me the whale...er...AI on a platter the size of a typical smartphone.

In Google's developing vision, there are so many things I no longer have to do for myself. This is laziness nirvana.

It starts simply. If I don't want to remember my license plate number (because why would I?), Google's new AskPhotos function can look through your massive photo library and use AI to identify your license plate for you. Walking around to the back of my car and checking the plate (or maybe just looking out the front window at the car parked in my driveway) is for people far more energetic than me.

Around 2011, I realized I had too much email to ever read, let alone understand. So much goes unread because I'm too lazy to go through it. My buddy Google Gemini slides up next to me, gives me a nudge and a grin, and promises to read it, summarize it, and create responses. This is the email of my lazy dreams.

As I write this, I'm about to head off on vacation. I hate planning vacations. I find it tedious and confusing, and I'm too lazy to get it done (my wife is the exact opposite). Gemini will let you tell it where you want to go, along with a few other details I may or may not have, and then it cooks up a shockingly complete travel itinerary (in a related vein, Gemini seems ready to manage an equally challenging move to, say, a new state).

I used to try to help my kids with their math problems and mostly failed, mainly because I was taught "old math" and they were taught "new math." Google's new Circle to Search for text problems, including math and formulas, would've soothed my lazy psyche into a blissful state of incompetence. Perhaps if my kids saw me doing this, simply just circling the thing I do not understand, they would've assumed I was being super helpful instead of what I am: super lazy.

I'm an "artist." I put it in quotes because my drawing skills are just above average. Worse, when faced with a drawing task, I often fight the impulse to not start at all. That's not just laziness, it's avoidance of possible disappointment because I couldn't create the thing. Google's latest image model, Imagine 3, is so good I realize there's no point in trying to create anything. I could never draw a wolf, wooden owl, yarn elephant, or people enjoying the golden hour sun that well. Enabling "lazy mode" and putting the pen down.

Hi, my name is Lance, and I'm a TikToker. For some reason I can't explain, TechRadar lets me shoot and post TikTok videos on its channel. Imagine Grandpa explaining tech, and you get the idea. It's a lot of work and sometimes, I'd rather visit our kitchen's snack bins. Google's Veo video model looks every bit as powerful as OpenAI's Sora. I bet that with a little training, it could create the TikToks for me and possibly include an AI-generated someone who looks vaguely like me (a thumb with glasses would do). Ooooh, Cheetos.

Now, this one's a bit dangerous. Google showed how you could pour all your research into Gemini, and it'll spit out at least an outline. You know reporting, which is hard work, by the way, is all research, right? Keep this lazy tool away from me.

Recently, we hired a new guy. It was a lot of work. Yes, worth the effort, but the work part I can do without. I had no idea Google would let me create a Virtual Teammate, one I can even name. He, she, or they can hang out in our chat rooms and engage like a real coworker. Gosh, this is so easy.

One of the big, overarching messages of Google I/O was to let Google do the Googling for you. In other words, if you need to perform a search, even a big, multi-part one, don't put much effort into it.

Don't tell Google, but I already do this. I'm usually too lazy to carefully parse out my prompt, so I type whatever is in my head into the search field and leave it to the search engine to figure it out.

It appears Google's been reading my lazy brain and just turned up the proactive search capabilities to 11. I mean, they are really owning the phrase, "Google will do the Googling for you." Music to my lazy heart.

I've done my share of tinkering and troubleshooting, and it doesn't always go well. What if I didn't? Google Project Astra is so peppy and proactive. All I have to do is film something, ask Google, "What the hell; is wrong with this?" and go take a nap while it spits out the answer.

An extension of this is Google Gemini's growing ability to explain everything I look at. Why should I expend the mental energy to interpret the images my eyes deliver to my brain? My Pixel 8 Pro has a "brain." I'll let it and Gemini doi it.

There are, it seems, no limits to what Google and its powerful AIs will do for you. If I want to find a couch and park myself there for a day or more, Google's upcoming AI Agents will be there to reason, plan, memorize, and think steps ahead. These little bits of AI are for you or rather me.

I'd share more, but, well, laziness.

Read the original:

Google I/O just showed me how to live the laziest life through AI - TechRadar

Posted in Ai

Google unveils Project Astra chatbot tech and brings ‘AI overview’ to search for all U.S. users – Fortune

Posted on by

Google showed off new AI chatbot technology dubbed Project Astra, along with a series of announcements infusing artificial intelligence throughout its catalogue of products, as company executives took the stage at its annual developers conference on Tuesday.

Alphabet CEO Sundar Pichai announced Tuesday that Google will roll out AI capabilities in its flagship search product to all U.S. users this week, while Demis Hassabis, the head of Googles DeepMind AI unit unveiled Project Astra, a universal AI agent that can understand the context of a users environment.

In a video demonstration of Astra, Google showed how users can point their phone camera to nearby objects and ask the AI agent relevant questions such as What neighborhood am I in? or Did you see where I left my glasses? Astra technology will come to the Gemini app later this year, the company said.

Speaking on stage near Alphabets headquarters in Mountain View, Calif., Pichai, Hassabis, and a parade of executives sought to show the companys progress in the high-stakes AI competition against BigTech rivals such as Microsoft, Meta, and Amazon, as well as richly-funded startups like OpenAI, Anthropic, and Perplexity.

Google said that its Gemini technology is now incorporated, in some way or another, in all of Googles key products those that boast more than 2 billion users, including YouTube and Search. And Google unveiled a new, standalone Gemini app for users to play with the latest AI features.

Google Search has already answered billions of queries with Gemini technology, Pichai said. We are encouraged not only to see an increase in search usage, but also in customer satisfaction.

In the coming weeks, Google will add multi-step reasoning in search, in which Gemini can answer long, multi-part questions. For example, it can find the best-rated yoga studios in Los Angeles, calculating the walking distance from each and offering the cost per class, all from one search query.

Google search is generative AI at the scale of human curiosity, Pichai said.

*Sixth generation of Trillium GPUs. Available to Google Cloud customers in late 2024, the new chips boast a 4.7x improvement in compute performance from the previous version.

*Veo, a generative video model available to use in VideoFX. Some Veo features will become available to select developers soon, and the wait list is open now.

*Google DeepMind and YouTube are building Music AI Sandbox, or a group of AI tools that can help artists create music.

*A new, dedicated Gemini smartphone app, which will offer all of the AI models features in one place. In the next few months, Gemini will also become available as an assistant on Android, as an overlay on whatever app a user is on, so they dont have to switch apps to use Gemini.

*Google is also rolling out a new feature to customize Gemini in the coming months. With what Google is calling Gems, users can curate what they want to see from the AI model.

*Gemini for Workspace side panel, an AI helper that lives on the side of the screen in Google Workspace applications, will be available next month.

*Data Q&A rolling out to Labs users in September, in which Gemini can help users organize spreadsheets.

Follow this link:

Google unveils Project Astra chatbot tech and brings 'AI overview' to search for all U.S. users - Fortune

Posted in Ai

Google will let you create personalized AI chatbots – The Verge

Posted on by

Google is adding a bunch of new features to its Gemini AI, and one of the most powerful is a personalization option called Gems that allows users to create custom versions of the Gemini assistant with varying personalities.

Gems lets you create iterations of chatbots that can help you with certain tasks and retain specific characteristics, kind of like making your own bot in Character.AI, the service that lets you talk to virtualized versions of popular characters and celebrities or even a fake psychiatrist. Google says you can make Gemini your gym buddy, sous-chef, coding partner, creative writing guide, or anything you can dream up. Gems feels similar to OpenAIs GPT Store that lets you make customized ChatGPT chatbots.

You can set up a gem by telling Gemini what to do and how to respond. For instance, you can tell it to be your running coach, provide you with a daily run schedule, and to sound upbeat and motivating. Then, in one click, Gemini will make a gem for you as youve described. The Gems feature is available soon to Gemini Advanced subscribers.

Continued here:

Google will let you create personalized AI chatbots - The Verge

Posted in Ai

Chinese firms make headway in producing high bandwidth memory for AI chipsets – Yahoo! Voices

Posted on by

By Fanny Potkin and Eduardo Baptista

SINGAPORE/BEIJING (Reuters) -Two Chinese chipmakers are in the early stages of producing high bandwidth memory (HBM) semiconductors used in artificial intelligence chipsets, according to sources and documents.

The progress in HBM - even if only in older versions of HBM - represents a major step forward in China's efforts to reduce its reliance on foreign suppliers amid tensions with Washington that have led to restrictions on U.S. exports of advanced chipsets to Chinese firms.

CXMT, China's top manufacturer of DRAM chips, has developed sample HBM chips in partnership with chip packaging and testing company Tongfu Microelectronics, according to three people briefed on the matter. The chips are being shown to clients, two of them said.

Tongfu Microelectronics' shares surged 8% in Wednesday trade.

In another example, Wuhan Xinxin is building a factory that will be able to produce 3,000 12-inch HBM wafers a month with construction slated to have begun in February this year, documents from corporate database Qichacha show.

CXMT and other Chinese chip firms have also been holding regular meetings with South Korean and Japanese semiconductor equipment firms to buy tools to develop HBM, said two of the people.

The sources were not authorised to speak on the matter and declined to be identified. Hefei-based CXMT or ChangXin Memory Technologies and Tongfu Microelectronics did not respond to requests for comment.

Wuhan Xinxin, which has flagged to regulators that it is interested in going public, and its parent company did not respond to requests for comment. The parent company is also the parent of NAND memory specialist YMTC or Yangtze Memory Technologies. YMTC said it did not have the capability to mass produce HBM.

Both CXMT and Wuhan Xinxin are private companies which have received local government funding to advance technologies as China pours capital into developing its chip sector.

Wuhan's local government also did not respond to requests for comment.

Separately, Chinese tech behemoth Huawei - which the U.S. has deemed a national security threat and is subject to sanctions - is aiming to produce HBM2 chips in partnership with other domestic companies by 2026, according to one of the sources and a separate person with knowledge of the matter.

The Information reported in April that a Huawei-led group of companies aiming to make HBM includes Fujian Jinhua Integrated Circuit, a memory chip maker also under U.S. sanctions.

Huawei, which has seen demand soar for its Ascend AI chips, declined to comment. It is not clear where Huawei procures HBM. Fujian Jinhua did not respond to a request for comment.

LONG JOURNEY AHEAD

HBM - a type of DRAM standard first produced in 2013 in which chips are vertically stacked to save space and reduce power consumption - is ideal for processing massive amounts of data produced by complex AI applications and demand has soared amid the AI boom.

The market for HBM is dominated by South Korea's SK Hynix - until recently the sole HBM supplier to AI chip giant Nvidia according to analysts - as well as Samsung, and to a lesser extent U.S. firm Micron Technology. All three manufacture the latest standard - HBM3 chips - and are working to bring fifth-generation HBM or HMB3E to customers this year.

China's efforts are currently focused on HBM2, according to two of the sources and a separate person with direct knowledge of the matter.

The U.S. has not put restrictions on exports of HBM chips per se but HBM3 chips are made using American technology that many Chinese firms including Huawei are barred from accessing as part of the curbs.

Nori Chiou, an investment director at White Oak Capital and a former analyst who looked at the IT sector, estimates that Chinese chipmakers lag their global rivals by a decade in HBM.

"China faces a considerable journey ahead, as it currently lacks the competitive edge to rival its Korean counterparts even in the realm of traditional memory markets," he said.

"Nonetheless, (CXMT's) collaboration with Tongfu represents a significant opportunity for China to advance its capabilities in both memory and advanced packaging technologies within the HBM market."

Patents filed by CXMT, Tongfu and Huawei indicate that plans to develop HBM domestically date back at least three years when China's chip industry increasingly became the target of U.S. export controls.

CXMT has filed almost 130 patents in the United States, China, and Taiwan for different technical issues related to the manufacturing and functionalities of HBM chips, according to Anaqua's AcclaimIP database. Of those, 14 were published in 2022, 46 in 2023, and 69 in 2024.

One Chinese patent, published last month, shows the company is looking at advanced packaging techniques like hybrid bonding to create a more powerful HBM product. A separate filing shows that CXMT is also investing in developing technology needed to create HBM3.

(Reporting by Fanny Potkin in Singapore and Eduardo Baptista in Beijing; Additional reporting by Heekyong Yang and Joyce Lee in Seoul; Editing by Brenda Goh and Edwina Gibbs)

Read more:

Chinese firms make headway in producing high bandwidth memory for AI chipsets - Yahoo! Voices

Posted in Ai

Google’s Gemini AI is coming to the sidebar in Docs, Drive, Gmail, and more – The Verge

Posted on by

The right sidebar in Googles Workspace apps is now the center for a lot of Googles AI plans. The company announced today at its I/O developer conference that it is bringing Gemini 1.5 Pro, its latest mainstream language model, to the sidebar in Google Docs, Sheets, Slides, Drive, and Gmail. Itll be the same virtual assistant across all of those apps, and the key bit is that itll know about everything you have saved everywhere.

The idea seems to be to use Gemini to connect all the Workspace apps more seamlessly. According to Aparna Pappu, the general manager and VP of Workspace, Google users have long been trying to hack Gemini to do complicated, multi-app things: send an email based on the data theyre looking at in Sheets or add a reminder to respond to the email they were currently looking at. And since Gemini has access to all of your documents, emails, and files, it can answer questions without forcing you to switch apps.

In a briefing with press ahead of I/O, Pappu gave the example of searching for information about a New York Knicks game. I could ask something like, what time do doors open for the Knicks game, and Im not looking for information from the web, which is going to give me generic information. I want information from my ticket, which happens to be a PDF in my email somewhere. Gemini can find that information, and Pappu said early users are quickly learning to use it as a way to find things more quickly.

Among early testers, Pappu said, a popular use case has been receipts. Rather than dig through your email, your files, and everything, you can just ask Gemini to find and organize all your receipts from across your Google account. And lets say you hit on the prompt that says Put my expenses in a Drive folder, she says, from there Gemini can put them all into a Sheet.

In keeping with so many of Googles announcements at I/O, the Workspace team seems to be focused on using Gemini to help you get stuff done and do stuff on your behalf. Pappu talked about how popular Gmails Help me write feature has been, especially on mobile, where people dont want to type as much. By grounding the model in your data and not the entire internet, Google hopes it can also begin to mitigate the models tendency to hallucinate and make other mistakes.

At least for now, the new sidebar isnt for everyone: its available now to some early testers and will roll out to paid Gemini subscribers next month. But Pappu did say Google is looking at how it could use on-device models to bring the capabilities to more users over time, so your days of hunting through Google Drive to find that old PDF may finally be coming to an end. Eventually.

Link:

Google's Gemini AI is coming to the sidebar in Docs, Drive, Gmail, and more - The Verge

Posted in Ai

Google now offers ‘web’ search and an AI opt-out button – The Verge

Posted on by

This is not a joke: Google will now let you perform a web search. Its rolling out web searches now, and in my early tests on desktop, its looking like it could be an incredibly popular change to Googles search engine.

The optional setting filters out almost all the other blocks of content that Google crams into a search results page, leaving you with links and text and Google confirms to The Verge that it will block the companys new AI Overviews as well.

Isnt every search a web search? What is Google Search if not the web? you might rightfully ask.

But independent websites like HouseFresh and Retro Dodo have pointed out how their businesses have gotten buried deep beneath sponsored posts, Quora advice from 2016, best-of lists from big media sites, and no less than 64 Google Shopping product listings, in the words of HouseFresh managing editor Gisele Navarro.

Now, with one click, a bunch of those blockers seemingly disappear.

Search for best home arcade cabinets, one of Retro Dodos bread-and-butter queries, and its no longer buried it appears on page 1. (Drag our image slider to see the difference.)

HouseFresh still doesnt get page 1 billing for best budget air purifiers but its higher up, and youre no longer assaulted by an eye-popping number of Google Shopping results as you scroll:

If you search for Wyze cameras, youll now get a hint about their lax security practices on page 2 instead of page 3:

Im not sure its an improvement for every search, partly because Googles modules can be useful, and partly because the company isnt giving up on self-promotion just because you press the web button. Here, you can see Google still gives itself top billing for Google AR glasses either way, and its Top stories box is arguably a helpful addition:

Which of these results helps you better learn about the Maui wildfires? Im genuinely not sure:

And when you ask Google who wrote The Lord of the Rings, is there any reason you wouldnt want Googles full knowledge graph at your disposal?

Admittedly, its an answer that Google isnt likely to get wrong.

As far as I can tell, the order of Googles search results seem to be the same regardless of whether you pick web or all. It doesnt block links to YouTube videos or Reddit posts or SEO factories... and I still saw (smaller!) sponsored ads from Amazon and Verkada and Wyze push down my search results:

Web is just a filter that removes Googles knowledge panels and featured snippets and Shopping modules and Googles new AI Overviews as well, Google spokesperson Ned Adriance confirms to The Verge. AI Overviews are a feature in Search, just like a knowledge panel or a featured snippet, so they will not appear when someone uses the web filter for a search.

It doesnt magically fix some of the issues facing Googles search engine. But it is a giant opt-out button for people whove been aggravated by some of the companys seemingly self-serving moves, and a way to preserve the spirit of the 10 blue links even as Googles AI efforts try to leave them behind.

Danny Sullivan, Googles Public Liaison for Search, says hes been asking for something like this for years:

As a next step, Id like to see Google promote the button to make it more visible. Right now, the company warns that it may not always appear in the primary carousel on desktop at all you may need to click More first and then select Web.

Heres hoping this all works well on mobile, too; Im not seeing it on my phone yet.

Read more:

Google now offers 'web' search and an AI opt-out button - The Verge

Posted in Ai

Project Astra is the future of AI at Google – The Verge

Posted on by

Ive had this vision in my mind for quite a while, says Demis Hassabis, the head of Google DeepMind and the leader of Googles AI efforts. Hassabis has been thinking about and working on AI for decades, but four or five years ago, something really crystallized. One day soon, he realized, We would have this universal assistant. Its multimodal, its with you all the time. Call it the Star Trek Communicator; call it the voice from Her; call it whatever you want. Its that helper, Hassabis continues, thats just useful. You get used to it being there whenever you need it.

At Google I/O, the companys annual developer conference, Hassabis showed off a very early version of what he hopes will become that universal assistant. Google calls it Project Astra, and its a real-time, multimodal AI assistant that can see the world, knows what things are and where you left them, and can answer questions or help you do almost anything. In an incredibly impressive demo video that Hassabis swears is not faked or doctored in any way, an Astra user in Googles London office asks the system to identify a part of a speaker, find their missing glasses, review code, and more. It all works practically in real time and in a very conversational way.

Astra is just one of many Gemini announcements at this years I/O. Theres a new model, called Gemini 1.5 Flash, designed to be faster for common tasks like summarization and captioning. Another new model, called Veo, can generate video from a text prompt. Gemini Nano, the model designed to be used locally on devices like your phone, is supposedly faster than ever as well. The context window for Gemini Pro, which refers to how much information the model can consider in a given query, is doubling to 2 million tokens, and Google says the model is better at following instructions than ever. Googles making fast progress both on the models themselves and on getting them in front of users.

Going forward, Hassabis says, the story of AI will be less about the models themselves and all about what they can do for you. And that story is all about agents: bots that dont just talk with you but actually accomplish stuff on your behalf. Our history in agents is longer than our generalized model work, he says, pointing to the game-playing AlphaGo system from nearly a decade ago. Some of those agents, he imagines, will be ultra-simple tools for getting things done, while others will be more like collaborators and companions. I think it may even be down to personal preference at some point, he says, and understanding your context.

Astra, Hassabis says, is much closer than previous products to the way a true real-time AI assistant ought to work. When Gemini 1.5 Pro, the latest version of Googles mainstream large language model, was ready, Hassabis says he knew the underlying tech was good enough for something like Astra to begin to work well. But the model is only part of the product. We had components of this six months ago, he says, but one of the issues was just speed and latency. Without that, the usability isnt quite there. So, for six months, speeding up the system has been one of the teams most important jobs. That meant improving the model but also optimizing the rest of the infrastructure to work well and at scale. Luckily, Hassabis says with a laugh, Thats something Google does very well!

A lot of Googles AI announcements at I/O are about giving you more and easier ways to use Gemini. A new product called Gemini Live is a voice-only assistant that lets you have easy back-and-forth conversations with the model, interrupting it when it gets long-winded or calling back to earlier parts of the conversation. A new feature in Google Lens allows you to search the web by shooting and narrating a video. A lot of this is enabled by Geminis large context window, which means it can access a huge amount of information at a time, and Hassabis says its crucial to making it feel normal and natural to interact with your assistant.

Know who agrees with that assessment, by the way? OpenAI, which has been talking about AI agents for a while now. In fact, the company demoed a product strikingly similar to Gemini Live barely an hour after Hassabis and I chatted. The two companies are increasingly fighting for the same territory and seem to share a vision for how AI might change your life and how you might use it over time.

How exactly will those assistants work, and how will you use them? Nobody knows for sure, not even Hassabis. One thing Google is focused on right now is trip planning it built a new tool for using Gemini to build an itinerary for your vacation that you can then edit in tandem with the assistant. There will eventually be many more features like that. Hassabis says hes bullish on phones and glasses as key devices for these agents but also says there is probably room for some exciting form factors. Astra is still in an early prototype phase and only represents one way you might want to interact with a system like Gemini. The DeepMind team is still researching how best to bring multimodal models together and how to balance ultra-huge general models with smaller and more focused ones.

Were still very much in the speeds and feeds era of AI, in which every incremental model matters and we obsess over parameter sizes. But pretty quickly, at least according to Hassabis, were going to start asking different questions about AI. Better questions. Questions about what these assistants can do, how they do it, and how they can make our lives better. Because the tech is a long way from perfect, but its getting better really fast.

Go here to see the original:

Project Astra is the future of AI at Google - The Verge

Posted in Ai

Today’s AI models are impressive. Teams of them will be formidable – The Economist

Posted on by

On May 13th OpenAI unveiled its latest model, GPT-4o. Mira Murati, the companys chief technology officer, called it the future of interaction between ourselves and the machines, because users can now speak to the AI and it will talk back in an expressive, human-like way.

The upgrade is part of wider moves across the tech industry to make chatbots and other artificial-intelligence, or AI, products into more useful and engaging assistants for everyday life. Show GPT-4o pictures or videos of art or food that you enjoy and it could probably furnish you with a list of museums, galleries and restaurants you might like. But it still has some way to go before it can become a truly useful AI assistant. Ask the model to plan a last-minute trip to Berlin for you based on your leisure preferencescomplete with details of which order to do everything, given how long each one takes and how far apart they are and which train tickets to buy, all within a set budgetand it will disappoint.

Read the original:

Today's AI models are impressive. Teams of them will be formidable - The Economist

Posted in Ai

Google demos out AI video generator Veo with the help of Donald Glover – Mashable

Posted on by

Google, with the help of creative renaissance man Donald Glover, has demoed an AI video generator to compete with OpenAI's Sora. The model is called Veo, and while no clear launch date or rollout plan has been announced, the demo does appear to show a Sora-like product, apparently capable of generating high-quality, convincing video.

What's "cool" about VEO? "You can make a mistake faster," Glover said in a video shown during Google's I/O 2024 livestream. "That's all you really want at the end of the day at least in art is just to make mistakes fast."

Credit: Mashable screenshot from a Google promo

Speaking onstage in Hawaii at Google I/O, Google Deepmind CEO Demis Hassabis said, "Veo creates high quality 1080p videos from text image and video prompts." This makes Veo the same type of tool, with the same resolution as Sora on its highest setting. A slider shown in the demo shows a Veo video length being stretched out to a little over one minute, also the approximate length of a Sora video.

Since Veo and Sora are both unreleased products, there's very little use trying to compare them in detail at this point. However, according to Hassabis, the interface will allow Veo users to "further edit your videos using additional prompts." This would be a function that Sora doesn't currently have according to creators who have been given access.

Mashable Light Speed

What was Veo trained on? That's not currently clear. About a month ago, YouTube CEO Neal Mohan told Bloomberg that if OpenAI used YouTube videos to train Sora, that would be a "clear violation" of the YouTube terms of service. However, YouTube's parent company Alphabet also owns Google, which made Veo. Mohan strongly implied in that Bloomberg interview that YouTube does feed content to Google's AI models, but only, he claims, when users sign off on it.

What we do know about the creation of Veo is that, according to Hassabis, this model is the culmination of Google and Deepmind's many similar projects, including Deepmind's Generative Query Network (GQN) research published back in 2018, last year's VideoPoet,Google's rudimentary video generator Phenaki, and Google's Lumiere, which was demoed earlier this year.

Glover's specific AI-enabled filmmaking project hasn't been announced. According to the video at I/O, Glover says he's "been interested in AI for a couple of years now," and that he reached out to Google and apparently not the other way around. "We got in contact with some of the people at Google and they had been working on something of their own, so we're all meeting," Glover says in Google's Veo demo video.

There's currently no way for the general public to try Veo, but there is a waitlist signup page.

Visit link:

Google demos out AI video generator Veo with the help of Donald Glover - Mashable

Posted in Ai

The SF Bay Area Has Become The Undisputed Leader In AI Tech And Funding Dollars – Crunchbase News

Posted on by

Theres been much talk of a resurgent San Francisco with the new technology wave of artificial intelligence washing over the software world. Indeed, Crunchbase funding data as well as interviews with startup investors and real estate industry professionals show the San Francisco Bay Area has become the undisputed epicenter of artificial intelligence.

Last year, more than 50% of all global venture funding for AI-related startups went to companies headquartered in the Bay Area, Crunchbase data shows, as a cluster of talent congregates in the region.

Beginning in Q1 2023, when OpenAIs ChatGPT reached 100 million users within months of launching, the amount raised by Bay Area startups in AI started trending up. That accelerated with OpenAI raising $10 billion from Microsoft marking the largest single funding deal ever for an AI foundation model company. In that quarter, more than 75% of AI funding went to San Francisco Bay Area startups.

AI-related companies based in the Bay Area went on to raise more than $27 billion in 2023, up from $14 billion in 2022, when the regions companies raised 29% of all AI funding.

From a deal count perspective, Bay Area companies raised 17% of global rounds in this sector in 2023 making the region the leading metro area in the U.S. That is an increase over 13% in 2022.

The figure also represents more than a third of AI deal counts in the U.S., and means the Bay Area alone had more AI-related startup funding deals than all countries outside of the U.S.

Leading Bay Area-based foundation model companies OpenAI, Anthropic and Inflection AI have each raised more than $1 billion or much more and have established major real estate footprints in San Francisco.

OpenAI has closed on 500,000 square feet of office space in the citys Mission Bay district and Anthropic around 230,000 square feet in the Financial District.

From a leasing standpoint, [AI] is the bright spot in San Francisco right now, said Derek Daniels, a regional director of research in San Francisco for commercial real estate brokerage Colliers, who has been following the trends closely.

By contrast, big tech has been pulling back and reassessing space needs, he said.

According to Daniels, the citys commercial real estate market bottomed out in the second half of 2023. While the San Francisco office space market still faces challenges, there is quality sublet space which is also seeing some demand for smaller teams, he said. And some larger tenants who have been out of the picture for office space of 100,000 square feet or more are starting to come back.

Fifty percent of startups that graduated from the prestigious startup accelerator Y Combinators April batch were AI-focused companies.

Many of the founders who came to SF for the batch have decided to make SF home for themselves, and for their companies, Garry Tan, president and CEO of Y Combinator, said in an announcement of the accelerators winter 2024 batch.

YC itself has expanded its office space in San Franciscos Dogpatch neighborhood adjacent to Mission Bay. We are turning San Franciscos doom loop into a boom loop, Tan added.

Of the batch 34 companies that graduated in March from 500 Global, another accelerator, 60% are in AI. Its next batch is closer to 80% with an AI focus, said Clayton Bryan, partner and head of the global accelerator fund.

Around half of the companies in the recently graduated 500 Global batch are from outside the U.S., including Budapest, London and Singapore. But many want to set up shop in the Bay Area for the density of talent, events and know-how from hackathons, dinners and events, he said.

Startup investors also see the Bay Area as the epicenter for AI.

In the more recent crop of AI companies there is a real center of gravity in the Bay Area, said Andrew Ferguson, a partner at Databricks Ventures, which has been actively investing in AI startups such as Perplexity AI, Unstructured Technologies, Anomalo, Cleanlaband Glean.

The Bay Area does not have a lock on good talent. But theres certainly a nucleus of very strong talent, he said.

Databricks Ventures, the venture arm of AI-enhanced data analytics unicorn Databricks, has made five investments in AI companies in the Bay Area in the past six months. In total, the firm has made around 25 portfolio company investments since the venture arm was founded in 2022, largely in the modern data stack.

Freed from in-person office requirements during the pandemic, many young tech workers decamped from the expensive Bay Area to travel or work remotely in less expensive locales. Now, some are moving back to join the San Francisco AI scene.

Many young founders are just moving back to the Bay Area, even if they were away for the last couple of years, in order to be a part of immersing themselves in the middle of the scene, said Stephanie Zhan, a partner at Sequoia Capital. Its great for networking, for hiring, for learning about whats going on, what other products people are building.

Coincidentally, Sequoia Capital subleased space to OpenAI in its early days, in an office above Dandelion Chocolates in San Franciscos Mission District.

Zhan presumes that many nascent AI companies arent yet showing up in funding data, as they are still ideating or at pre-seed or seed funding, and will show up in future funding cycles.

While the Bay Area dominates for AI funding, its important to note the obvious: Much of that comes from a few massive deals to the large startups based in the region, including OpenAI, Anthropic and Inflection AI.

There is a lot of AI startup and research activity elsewhere as well, Zhan noted, with researchers coming out of universities around the globe, including cole Polytechnique in Paris, ETH Zrich and the University of Cambridge and Oxford University in the U.K., to name a few. Lead researchers from the University of Toronto and University of Waterloo have also fed into generative AI technology in San Francisco and in Canada, Bryan said.

While the U.S. has a strong lead, countries that are leading funding totals for AI-related startups outside of the U.S. are China, the U.K., Germany, Canada and France, according to Crunchbase data.

London-based Stability AI kicked off the generative AI moment before ChatGPT with its text-to-image models in August 2022. Open source model developer Mistral AI, based in Paris, has raised large amounts led by Bay Area-based venture capital firms Lightspeed Venture Partners and Andreessen Horowitz.

And in China, foundation model company Moonshot AI based in Beijing has raised more than $1 billion.

Still, the center of gravity in the Bay Area is driven by teams coming out of Big Tech or UC Berkeley and Stanford University who have a history of turning those ideas into startups, said Ferguson.

The unique congregation of Big Tech companies, research, talent and venture capital in the Bay Area has placed the region at the forefront of AI.

The valuation of the AI companies and some of the revenue by the top end of the AI companies is driving that population migration, said 500 Globals Bryan. At a recent AI event at Hana House in Palo Alto, California, he found it interesting that most people were not originally from the Bay Area. Everyone now wants a direct piece or an indirect piece of that value that is going into AI.

Illustration: Li-Anne Dias

Stay up to date with recent funding rounds, acquisitions, and more with the Crunchbase Daily.

See original here:

The SF Bay Area Has Become The Undisputed Leader In AI Tech And Funding Dollars - Crunchbase News

Posted in Ai

Android is getting an AI-powered scam call detection feature – The Verge

Posted on by

Google is working on new protections to help prevent Android users from falling victim to phone scams. During its I/O developer conference on Tuesday, Google announced that its testing a new call monitoring feature that will warn users if the person theyre talking to is likely attempting to scam them and encourage them to end such calls.

Google says the feature utilizes Gemini Nano a reduced version of the companys Gemini large language model for Android devices that can run locally and offline to look for fraudulent language and other conversation patterns typically associated with scams. Users will then receive real-time alerts during calls where these red flags are present.

Some examples of what could trigger these alerts include calls from bank representatives who make requests that real banks are unlikely to make, such as asking for personal information like your passwords or card PINs, requesting payments via gift cards, or asking users to urgently transfer money to them. These new protections are entirely on-device, so the conversations monitored by Gemini Nano will remain private, according to Google.

Theres no word on when the scam detection feature will be available, but Google says users will need to opt in to utilize it and that itll share more information later this year.

So, while the candidates who might find such tech useful are vast, compatibility could limit its applicability. Gemini Nano is only currently supported on the Google Pixel 8 Pro and Samsung S24 series, according to its developer support page.

Visit link:

Android is getting an AI-powered scam call detection feature - The Verge

Posted in Ai

GPT-4o delivers human-like AI interaction with text, audio, and vision integration – AI News

Posted on by

OpenAI has launched its new flagship model, GPT-4o, which seamlessly integrates text, audio, and visual inputs and outputs, promising to enhance the naturalness of machine interactions.

GPT-4o, where the o stands for omni, is designed to cater to a broader spectrum of input and output modalities. It accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs, OpenAI announced.

Users can expect a response time as quick as 232 milliseconds, mirroring human conversational speed, with an impressive average response time of 320 milliseconds.

The introduction of GPT-4o marks a leap from its predecessors by processing all inputs and outputs through a single neural network. This approach enables the model to retain critical information and context that were previously lost in the separate model pipeline used in earlier versions.

Prior to GPT-4o, Voice Mode could handle audio interactions with latencies of 2.8 seconds for GPT-3.5 and 5.4 seconds for GPT-4. The previous setup involved three distinct models: one for transcribing audio to text, another for textual responses, and a third for converting text back to audio. This segmentation led to loss of nuances such as tone, multiple speakers, and background noise.

As an integrated solution, GPT-4o boasts notable improvements in vision and audio understanding. It can perform more complex tasks such as harmonising songs, providing real-time translations, and even generating outputs with expressive elements like laughter and singing. Examples of its broad capabilities include preparing for interviews, translating languages on the fly, and generating customer service responses.

Nathaniel Whittemore, Founder and CEO of Superintelligent, commented: Product announcements are going to inherently be more divisive than technology announcements because its harder to tell if a product is going to be truly different until you actually interact with it. And especially when it comes to a different mode of human-computer interaction, there is even more room for diverse beliefs about how useful its going to be.

That said, the fact that there wasnt a GPT-4.5 or GPT-5 announced is also distracting people from the technological advancement that this is a natively multimodal model. Its not a text model with a voice or image addition; it is a multimodal token in, multimodal token out. This opens up a huge array of use cases that are going to take some time to filter into the consciousness.

GPT-4o matches GPT-4 Turbo performance levels in English text and coding tasks but outshines significantly in non-English languages, making it a more inclusive and versatile model. It sets a new benchmark in reasoning with a high score of 88.7% on 0-shot COT MMLU (general knowledge questions) and 87.2% on the 5-shot no-CoT MMLU.

The model also excels in audio and translation benchmarks, surpassing previous state-of-the-art models like Whisper-v3. In multilingual and vision evaluations, it demonstrates superior performance, enhancing OpenAIs multilingual, audio, and vision capabilities.

OpenAI has incorporated robust safety measures into GPT-4o by design, incorporating techniques to filter training data and refining behaviour through post-training safeguards. The model has been assessed through a Preparedness Framework and complies with OpenAIs voluntary commitments. Evaluations in areas like cybersecurity, persuasion, and model autonomy indicate that GPT-4o does not exceed a Medium risk level across any category.

Further safety assessments involved extensive external red teaming with over 70 experts in various domains, including social psychology, bias, fairness, and misinformation. This comprehensive scrutiny aims to mitigate risks introduced by the new modalities of GPT-4o.

Starting today, GPT-4os text and image capabilities are available in ChatGPTincluding a free tier and extended features for Plus users. A new Voice Mode powered by GPT-4o will enter alpha testing within ChatGPT Plus in the coming weeks.

Developers can access GPT-4o through the API for text and vision tasks, benefiting from its doubled speed, halved price, and enhanced rate limits compared to GPT-4 Turbo.

OpenAI plans to expand GPT-4os audio and video functionalities to a select group of trusted partners via the API, with broader rollout expected in the near future. This phased release strategy aims to ensure thorough safety and usability testing before making the full range of capabilities publicly available.

Its hugely significant that theyve made this model available for free to everyone, as well as making the API 50% cheaper. That is a massive increase in accessibility, explained Whittemore.

OpenAI invites community feedback to continuously refine GPT-4o, emphasising the importance of user input in identifying and closing gaps where GPT-4 Turbo might still outperform.

(Image Credit: OpenAI)

See also: OpenAI takes steps to boost AI-generated content transparency

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

Tags: ai, api, artificial intelligence, benchmarks, chatgpt, coding, developers, development, gpt-4o, Model, multimodal, openai, performance, programming

Original post:

GPT-4o delivers human-like AI interaction with text, audio, and vision integration - AI News

Posted in Ai

Apples AI research suggests features are coming for Siri, artists, and more. – The Verge

Posted on by

It would be easy to think that Apple is late to the game on AI. Since late 2022, when ChatGPT took the world by storm, most of Apples competitors have fallen over themselves to catch up. While Apple has certainly talked about AI and even released some products with AI in mind, it seemed to be dipping a toe in rather than diving in headfirst.

But over the last few months, rumors and reports have suggested that Apple has, in fact, just been biding its time, waiting to make its move.There have been reports in recent weeks that Apple is talking to both OpenAI and Google about powering some of its AI features, and the company has also been working on its own model, called Ajax.

If you look through Apples published AI research, a picture starts to develop of how Apples approach to AI might come to life. Now, obviously, making product assumptions based on research papers is a deeply inexact science the line from research to store shelves is windy and full of potholes. But you can at least get a sense of what the company is thinking about and how its AI features might work when Apple starts to talk about them at its annual developer conference, WWDC, in June.

I suspect you and I are hoping for the same thing here: Better Siri. And it looks very much like Better Siri is coming! Theres an assumption in a lot of Apples research (and in a lot of the tech industry, the world, and everywhere)that large language models will immediately make virtual assistants better and smarter. For Apple, getting to Better Siri means making those models as fast as possible and making sure theyre everywhere.

In iOS 18, Apple plans to have all its AI features running on an on-device, fully offline model, Bloomberg recently reported. Its tough to build a good multipurpose model even when you have a network of data centers and thousands of state-of-the-art GPUs its drastically harder to do it with only the guts inside your smartphone. So Apples having to get creative.

In a paper called LLM in a flash: Efficient Large Language Model Inference with Limited Memory (all these papers have really boring titles but are really interesting, I promise!), researchers devised a system for storing a models data, which is usually stored on your devices RAM, on the SSD instead. We have demonstrated the ability to run LLMs up to twice the size of available DRAM [on the SSD], the researchers wrote, achieving an acceleration in inference speed by 4-5x compared to traditional loading methods in CPU, and 20-25x in GPU. By taking advantage of the most inexpensive and available storage on your device, they found, the models can run faster and more efficiently.

Apples researchers also created a system called EELBERT that can essentially compress an LLM into a much smaller size without making it meaningfully worse. Their compressed take on Googles Bert model was 15 times smaller only 1.2 megabytes and saw only a 4 percent reduction in quality. It did come with some latency tradeoffs, though.

In general, Apple is pushing to solve a core tension in the model world: the bigger a model gets, the better and more useful it can be, but also the more unwieldy, power-hungry, and slow it can become. Like so many others, the company is trying to find the right balance between all those things while also looking for a way to have it all.

A lot of what we talk about when we talk about AI products is virtual assistants assistants that know things, that can remind us of things, that can answer questions, and get stuff done on our behalf. So its not exactly shocking that a lot of Apples AI research boils down to a single question: what if Siri was really, really, really good?

A group of Apple researchers has been working on a way to use Siri without needing to use a wake word at all; instead of listening for Hey Siri or Siri, the device might be able to simply intuit whether youre talking to it. This problem is significantly more challenging than voice trigger detection, the researchers did acknowledge, since there might not be a leading trigger phrase that marks the beginning of a voice command. That might be why another group of researchers developed a system to more accurately detect wake words. Another paper trained a model to better understand rare words, which are often not well understood by assistants.

In both cases, the appeal of an LLM is that it can, in theory, process much more information much more quickly. In the wake-word paper, for instance, the researchers found that by not trying to discard all unnecessary sound but, instead, feeding it all to the model and letting it process what does and doesnt matter, the wake word worked far more reliably.

Once Siri hears you, Apples doing a bunch of work to make sure it understands and communicates better. In one paper, it developed a system called STEER (which stands for Semantic Turn Extension-Expansion Recognition, so well go with STEER) that aims to improve your back-and-forth communication with an assistant by trying to figure out when youre asking a follow-up question and when youre asking a new one. In another, it uses LLMs to better understand ambiguous queries to figure out what you mean no matter how you say it. In uncertain circumstances, they wrote, intelligent conversational agents may need to take the initiative to reduce their uncertainty by asking good questions proactively, thereby solving problems more effectively. Another paper aims to help with that, too: researchers used LLMs to make assistants less verbose and more understandable when theyre generating answers.

Whenever Apple does talk publicly about AI, it tends to focus less on raw technological might and more on the day-to-day stuff AI can actually do for you. So, while theres a lot of focus on Siri especially as Apple looks to compete with devices like the Humane AI Pin, the Rabbit R1, and Googles ongoing smashing of Gemini into all of Android there are plenty of other ways Apple seems to see AI being useful.

One obvious place for Apple to focus is on health: LLMs could, in theory, help wade through the oceans of biometric data collected by your various devices and help you make sense of it all. So, Apple has been researching how to collect and collate all of your motion data, how to use gait recognition and your headphones to identify you, and how to track and understand your heart rate data. Apple also created and released the largest multi-device multi-location sensor-based human activity dataset available after collecting data from 50 participants with multiple on-body sensors.

Apple also seems to imagine AI as a creative tool. For one paper, researchers interviewed a bunch of animators, designers, and engineers and built a system called Keyframer that enable[s] users to iteratively construct and refine generated designs. Instead of typing in a prompt and getting an image, then typing another prompt to get another image, you start with a prompt but then get a toolkit to tweak and refine parts of the image to your liking. You could imagine this kind of back-and-forth artistic process showing up anywhere from the Memoji creator to some of Apples more professional artistic tools.

In another paper, Apple describes a tool called MGIE that lets you edit an image just by describing the edits you want to make. (Make the sky more blue, make my face less weird, add some rocks, that sort of thing.) Instead of brief but ambiguous guidance, MGIE derives explicit visual-aware intention and leads to reasonable image editing, the researchers wrote. Its initial experiments werent perfect, but they were impressive.

We might even get some AI in Apple Music: for a paper called Resource-constrained Stereo Singing Voice Cancellation, researchers explored ways to separate voices from instruments in songs which could come in handy if Apple wants to give people tools to, say, remix songs the way you can on TikTok or Instagram.

Over time, Id bet this is the kind of stuff youll see Apple lean into, especially on iOS. Some of it Apple will build into its own apps; some it will offer to third-party developers as APIs. (The recent Journaling Suggestions feature is probably a good guide to how that might work.) Apple has always trumpeted its hardware capabilities, particularly compared to your average Android device; pairing all that horsepower with on-device, privacy-focused AI could be a big differentiator.

But if you want to see the biggest, most ambitious AI thing going at Apple, you need to know about Ferret. Ferret is a multi-modal large language model that can take instructions, focus on something specific youve circled or otherwise selected, and understand the world around it. Its designed for the now-normal AI use case of asking a device about the world around you, but it might also be able to understand whats on your screen. In the Ferret paper, researchers show that it could help you navigate apps, answer questions about App Store ratings, describe what youre looking at, and more. This has really exciting implications for accessibility but could also completely change the way you use your phone and your Vision Pro and / or smart glasses someday.

Were getting way ahead of ourselves here, but you can imagine how this would work with some of the other stuff Apple is working on. A Siri that can understand what you want, paired with a device that can see and understand everything thats happening on your display, is a phone that can literally use itself. Apple wouldnt need deep integrations with everything; it could simply run the apps and tap the right buttons automatically.

Again, all this is just research, and for all of it to work well starting this spring would be a legitimately unheard-of technical achievement. (I mean, youve tried chatbots you know theyre not great.) But Id bet you anything were going to get some big AI announcements at WWDC. Apple CEO Tim Cook even teased as much in February, and basically promised it on this weeks earnings call. And two things are very clear: Apple is very much in the AI race, and it might amount to a total overhaul of the iPhone. Heck, you might even start willingly using Siri! And that would be quite the accomplishment.

Original post:

Apples AI research suggests features are coming for Siri, artists, and more. - The Verge

Posted in Ai

Providing further transparency on our responsible AI efforts – Microsoft On the Issues – Microsoft

Posted on by

The following is the foreword to the inaugural edition of our annual Responsible AI Transparency Report. The FULL REPORT is available at this link.

We believe we have an obligation to share our responsible AI practices with the public, and this report enables us to record and share our maturing practices, reflect on what we have learned, chart our goals, hold ourselves accountable, and earn the publics trust.

In 2016, our Chairman and CEO, Satya Nadella, set us on a clear course to adopt a principled and human-centered approach to our investments in artificial intelligence (AI). Since then, we have been hard at work building products that align with our values. As we design, build, and release AI products, six values transparency, accountability, fairness, inclusiveness, reliability and safety, and privacy and security remain our foundation and guide our work every day.

To advance our transparency practices, in July 2023, we committed to publishing an annual report on our responsible AI program, taking a step that reached beyond the White House Voluntary Commitments that we and other leading AI companies agreed to. This is our inaugural report delivering on that commitment, and we are pleased to publish it on the heels of our first year of bringing generative AI products and experiences to creators, non-profits, governments, and enterprises around the world.

As a company at the forefront of AI research and technology, we are committed to sharing our practices with the public as they evolve. This report enables us to share our maturing practices, reflect on what we have learned, chart our goals, hold ourselves accountable, and earn the publics trust. Weve been innovating in responsible AI for eight years, and as we evolve our program, we learn from our past to continually improve. We take very seriously our responsibility to not only secure our own knowledge but also to contribute to the growing corpus of public knowledge, to expand access to resources, and promote transparency in AI across the public, private, and non-profit sectors.

In this inaugural annual report, we provide insight into how we build applications that use generative AI; make decisions and oversee the deployment of those applications; support our customers as they build their own generative applications; and learn, evolve, and grow as a responsible AI community. First, we provide insights into our development process, exploring how we map, measure, and manage generative AI risks. Next, we offer case studies to illustrate how we apply our policies and processes to generative AI releases. We also share details about how we empower our customers as they build their own AI applications responsibly. Last, we highlight how the growth of our responsible AI community, our efforts to democratize the benefits of AI, and our work to facilitate AI research benefit society at large.

There is no finish line for responsible AI. And while this report doesnt have all the answers, we are committed to sharing our learnings early and often and engaging in a robust dialogue around responsible AI practices. We invite the public, private organizations, non-profits, and governing bodies to use this first transparency report to accelerate the incredible momentum in responsible AI were already seeing around the world.

Click here to read the full report.

Tags: AI, generative ai, Responsible AI, Responsible AI Transparency Report, transparency, White House Voluntary Commitments

See original here:

Providing further transparency on our responsible AI efforts - Microsoft On the Issues - Microsoft

Posted in Ai