Category Archives: Ai

The Decade, Reviewed looks back at the 2010s and how it changed human society forever. From 2010 to 2019, our species experienced seismic shifts in science, technology, entertainment, transportation, and even the very planet we call home. This is how the past ten years have changed us.

Bots are a lot like humans: Some are cute. Some are ugly. Some are harmless. Some are menacing. Some are friendly. Some are annoying ... and a little racist. Bots serve their creators and society as helpers, spies, educators, servants, lab technicians, and artists. Sometimes, they save lives. Occasionally, they destroy them.

In the 2010s, automation got better, cheaper, and way less avoidable. Its still mysterious, but no longer foreign; the most Extremely Online among us interact with dozens of AIs throughout the day. That means driving directions are more reliable, instant translations are almost good enough, and everyone gets to be an adequate portrait photographer, all powered by artificial intelligence. On the other hand, each of us now sees a personalized version of the world that is curated by an AI to maximize engagement with the platform. And by now, everyone from fruit pickers to hedge fund managers has suffered through headlines about being replaced.

Humans and tech have always coexisted and coevolved, but this decade brought us closer togetherand closer to the futurethan ever. These days, you dont have to be an engineer to participate in AI projects; in fact, you have no choice but to help, as youre constantly offering your digital behavior to train AIs.

So heres how we changed our bots this decade, how they changed us, and where our strange relationship is going as we enter the 2020s.

All those little operational tweaks in our day come courtesy of a specific scientific approach to AI called machine learning, one of the most popular techniques for AI projects this decade. Thats when AI is tasked not only with finding the answers to questions about data sets, but with finding the questions themselves; successful deep learning applications require vast amounts of data and the time and computational power to self-test over and over again.

Deep learning, a subset of machine learning, uses neural networks to extract its own rules and adjust them until it can return the right results; other machine learning techniques might use Bayesian networks, vector maps, or evolutionary algorithms to achieve the same goal.

In January, Technology Reviews Karen Hao released an exhaustive analysis of recent papers in AI that concluded that machine learning was one of the defining features of AI research this decade. Machine learning has enabled near-human and even superhuman abilities in transcribing speech from voice, recognizing emotions from audio or video recordings, as well as forging handwriting or video, Hao wrote. Domestic spying is now a lucrative application for AI technologies, thanks to this powerful new development.

Haos report suggests that the age of deep learning is finally drawing to a close, but the next big thing may have already arrived. Reinforcement learning, like generative adversarial networks (GANs), pits neural nets against one another by having one evaluate the work of the other and distribute rewards and punishments accordinglynot unlike the way dogs and babies learn about the world.

The future of AI could be in structured learning. Just as young humans are thought to learn their first languages by processing data input from fluent caretakers with their internal language grammar, computers can also be taught how to teach themselves a taskespecially if the task is to imitate a human in some capacity.

This decade, artificial intelligence went from being employed chiefly as an academic subject or science fiction trope to an unobtrusive (though occasionally malicious) everyday companion. AIs have been around in some form since the 1500s or the 1980s, depending on your definition. The first search indexing algorithm was AltaVista in 1995, but it wasnt until 2010 that Google quietly introduced personalized search results for all customers and all searches. What was once background chatter from eager engineers has now become an inescapable part of daily life.

One function after another has been turned over to AI jurisdiction, with huge variations in efficacy and consumer response. The prevailing profit model for most of these consumer-facing applications, like social media platforms and map functions, is for users to trade their personal data for minor convenience upgrades, which are achieved through a combination of technical power, data access, and rapid worker disenfranchisement as increasingly complex service jobs are doubled up, automated away, or taken over by AI workers.

The Harvard social scientist Shoshana Zuboff explained the impact of these technologies on the economy with the term surveillance capitalism. This new economic system, she wrote, unilaterally claims human experience as free raw material for translation into behavioural data, in a bid to make profit from informed gambling based on predicted human behavior.

Were already using machine learning to make subjective decisionseven ones that have life-altering consequences. Medical applications are only some of the least controversial uses of artificial intelligence; by the end of the decade, AIs were locating stranded victims of Hurricane Maria, controlling the German power grid, and killing civilians in Pakistan.

The sheer scope of these AI-controlled decision systems is why automation has the potential to transform society on a structural level. In 2012, techno-socialist Zeynep Tufekci pointed out the presence on the Obama reelection campaign of an unprecedented number of data analysts and social scientists, bringing the traditional confluence of marketing and politics into a new age.

Intelligence that relies on data from an unjust world suffers from the principle of garbage in, garbage out, futurist Cory Doctorow observed in a recent blog post. Diverse perspectives on the design team would help, Doctorow wrote, but when it comes to certain technology, there might be no safe way to deploy:

It doesnt help that data collection for image-based AI has so far taken advantage of the most vulnerable populations first. The Facial Recognition Verification Testing Program is the industry standard for testing the accuracy of facial recognition tech; passing the program is imperative for new FR startups seeking funding.

But the datasets of human faces that the program uses are sourced, according to a report from March, from images of U.S. visa applicants, arrested people who have since died, and children exploited by child pornography. The report found that the majority of data subjects were people who had been arrested on suspicion of criminal activity. None of the millions of faces in the programs data sets belonged to people who had consented to this use of their data.

State-level efforts to regulate AI finally emerged this decade, with some success. The European Unions General Data Protection Regulation (GDPR), enforceable from 2018, limits the legal uses of valuable AI training datasets by defining the rights of the data subject (read: us); the GDPR also prohibits the black box model for machine learning applications, requiring both transparency and accountability on how data are stored and used. At the end of the decade, Google showed the class how not to regulate when they built, and then scrapped, an external AI ethics panel a week later, feigning shock at all the negative reception.

Even attempted regulation is a good sign. It means were looking at AI for what it is: not a new life form that competes for resources, but as a formidable weapon. Technological tools are most dangerous in the hands of malicious actors who already hold significant power; you can always hire more programmers. During the long campaign for the 2016 U.S. presidential election, the Putin-backed IRA Twitter botnet campaignsessentially, teams of semi-supervised bot accounts that spread disinformation on purpose and learn from real propagandainfiltrated the very mechanics of American democracy.

Keeping up with AI capacities as they grow will be a massive undertaking. Things could still get much, much worse before they get better; authoritarian governments around the world have a tendency to use technology to further consolidate power and resist regulation.

Tech capabilities have long since proved too fast for traditional human lawmakers, but one hint of what the next decade might hold comes from AIs themselves, who are beginning to be deployed as weapons against the exact type of disinformation other AIs help to create and spread. There now exists, for example, a neural net devoted explicitly to the task of identifying neural net disinformation campaigns on Twitter. The neural nets name is Grover, and its really good at this.

Continued here:

The Bot Decade: How AI Took Over Our Lives in the 2010s - Popular Mechanics

At the last Democratic presidential debate, the technologist candidate Andrew Yang emphatically declared that were in the process of potentially losing the AI arms race to China right now. As evidence, he cited Beijings access to vast amounts of data and its substantial investment in research and development for artificial intelligence. Yang and othersmost notably the National Security Commission on Artificial Intelligence, whichreleased its interim report to Congress last monthare right about Chinas current strengths in developing AI and the serious concerns this should raise in the United States. But framing advances in the field as an arms race is both wrong and counterproductive. Instead, while being clear-eyed about Chinas aggressive pursuit of AI for military use and human rights-abusing technological surveillance, the United States and China must find their way to dialogue and cooperation on AI. A practical, nuanced mix of competition and cooperation would better serve U.S. interests than an arms race approach.

AI is one of the great collective Rorschach tests of our times. Like any topic that captures the popular imagination but is poorly understood, it soaks up the zeitgeist like a sponge.

Its no surprise, then, that as the idea of great-power competition has reengulfed the halls of power, AI has gotten caught up in therace narrative.ChinaAmericans are toldis barreling ahead on AI, so much so that the United States willsoon be lagging far behind. Like the fears that surrounded Japans economic rise in the 1980s or the Soviet Union in the 1950s and 1960s, anxiety around technological dominance are really proxies for U.S. insecurity about its own economic, military, and political prowess.

Yet as technology, AI does not naturally lend itself to this framework and is not a strategic weapon.Despite claims that AI will change nearly everything about warfare, and notwithstanding its ultimate potential, for the foreseeable future AI will likely only incrementally improve existing platforms, unmanned systems such as drones, and battlefield awareness. Ensuring that the United States outpaces its rivals and adversaries in the military and intelligence applications of AI is important and worth the investment. But such applications are just one element of AI development and should not dominate the United States entire approach.

The arms race framework raises the question of what one is racing toward. Machine learning, the AI subfield of greatest recent promise, is a vast toolbox of capabilities and statistical methodsa bundle of technologies that do everything from recognizing objects in images to generating symphonies. It is far from clear what exactly would constitute winning in AI or even being better at a national level.

The National Security Commission is absolutely right that developments in AI cannot be separated from the emerging strategic competition with China and developments in the broader geopolitical landscape. U.S. leadership in AI is imperative. Leading, however, does not mean winning. Maintaining superiority in the field of AI is necessary but not sufficient. True global leadership requires proactively shaping the rules and norms for AI applications, ensuring that the benefits of AI are distributed worldwidebroadly and equitablyand stabilizing great-power competition that could lead to catastrophic conflict.

That requires U.S. cooperation with friends and even rivals such as China. Here, we believe that important aspects of the National Security Commission on AIs recent report have gotten too little attention.

First, as the commission notes, official U.S. dialogue with China and Russia on the use of AI in nuclear command and control, AIs military applications, and AI safety could enhance strategic stability, like arms control talks during the Cold War. Second, collaboration on AI applications by Chinese and American researchers, engineers, and companies, as well as bilateral dialogue on rules and standards for AI development, could help buffer the competitive elements of anincreasingly tense U.S.-Chinese relationship.

Finally, there is a much higher bar to sharing core AI inputs such as data and software and building AI for shared global challenges if the United States sees AI as an arms race. Although commercial and military applications for AI are increasing, applications for societal good (addressing climate change,improving disaster response,boosting resilience, preventing the emergence of pandemics, managing armed conflict, andassisting in human development)are lagging. These would benefit from multilateral collaboration and investment, led by the United States and China.

The AI arms race narrative makes for great headlines, buttheunbridled U.S.-Chinese competition it implies risks pushing the United States and the world down a dangerous path. Washington and Beijing should recognize the fallacy of a generalized AI arms race in which there are no winners. Instead, both should lead by leveraging the technology to spur dialogue between them and foster practical collaboration to counter the many forces driving them apartbenefiting the whole world in the process.

Read more here:

Artificial Intelligence Isn't an Arms Race With China, and the United States Shouldn't Treat It Like One - Foreign Policy

At a startup's birth, an idea or a vision can be so well wrapped in a blanket of buzzwords that not even the smartest investors can stop themselves from throwing money at it. Lets be honest: Most groundbreaking ideas deserve a sober share of criticism, but once a company gets its initial funding and the team goes into execution, continuing to feed the same party line has to be called out.

Some of the buzz floating around the insurance industry is artificial intelligence (AI). When I say "buzz," I mean product messaging like this:

Our [insert product name] uses artificial intelligence to help find you the best insurance coverage.

Or, better yet:

"Let artificial intelligence handle your insurance claims from the initial call to the fund's disbursement."

One of the insurance industry's sore spots is claims from the first notice of loss (FNOL) to the claims payout. It is ridden with fraud, inefficiencies and lack of transparency. Many of us have experiences spending hours on the phone with customer service describing what happened, going from one shop to another to get several repair estimates and waiting for the reimbursement check to arrive.

It gets worse when dealing with home insurance claims. There is always a caveat to what is covered and what is not covered.

Thus, filing a claim turns into a never-ending argument with the insurer and a big out-of-pocket expense. This surely sounds like a perfect candidate to be saved by an AI knight in shining armor riding on a machine learning horse.

Let's take a look at auto claims, which are relatively standardized and straightforward compared to home damage claims.

I cringe when I read that certain insurance companies are "using the latest AI technology" to help resolve claims in seconds or minutes rather than weeks or months. Without digging deep under wraps of each insurer's statement, I believe that, at best, the usage of AI is limited to identifying some of the damaged parts.

Here is how a typical AI computer vision system works: It analyzes (using neural networks) thousands of images (let's say, a rear bumper) and uses proprietary mathematical models to come up with a certain confidence level upon looking at a new picture of what is supposed to be a bumper.

Once the system has analyzed the image and determined with some confidence level that it is indeed a Ford Mustang rear bumper and the damage is medium, the platform can match all the previous Ford Mustang rear bumper repair jobs that were classified as medium and suggest how much it would cost to fix the damage (whether to repair or replace). Then it can suggest a course of action based on the deductible or coverage details.

I recall my own "minor damage" claim experience. I happened to hit a turkey while driving 65 mph on an interstate. Besides losing one side marker lamp and acquiring a few bumper dents, I couldnt see anything serious. I called my insurance company and filed a claim. When my insurance company adjuster visually inspected the damage, his estimate was about $1,650. I took my vehicle to a body shop, and the mechanic lifted the car to inspect it properly. Guess what? He found more damage and had to invite the adjuster back to look. The final bill ended up being around $4,500.

Would this number have been know without a human looking more closely? Could AI have found all that hidden damage that an adjuster with 30 years of experience didnt see?

While some insurance companies claim AI technology provides them with immediate total loss and repair estimates based on photos and, thus, saves everyone time and expenses, I am personally skeptical. Yes, this would have made my FNOL experience superb. I wouldnt have spent time on the phone, driven to an adjusters shop or waited for his initial inspection. I would have just received an ACH transfer to my bank account for $1,650. The fun part would have begun once I brought the car to the mechanic, who then would have had to schedule an appointment to get the adjuster over to approve more funding, and Id be waiting again, this time cursing my insurance for not being able to get it right the first time.

Insurance companies watch their claim payouts like hawks. It's no wonder why. In my experience, overall claim expenses are 35% to 45% of their total expenses. So if AI will begin to overpay, an insurance company will be underwater in no time. So naturally, they would "tune" AI to underpay, resulting in additional adjuster visits, thus defeating the purpose of AI to begin with.

Do I believe AI will be capable of replacing human adjusters or making entire insurance claim processes touchless"?

Maybe in the near future, perhaps in 5-10 years.

For now, adjusters, customer service agents and underwriters are safe from the incoming AI doom-and-gloom predictions.

Continue reading here:

Will AI Be Capable Of Replacing Insurance Claims Adjusters? - Forbes

An expert on artificial intelligence has called for all algorithms that make life-changing decisions in areas from job applications to immigration into the UK to be halted immediately.

Prof Noel Sharkey, who is also a leading figure in a global campaign against killer robots, said algorithms were so infected with biases that their decision-making processes could not be fair or trusted.

A moratorium must be imposed on all life-changing decision-making algorithms in Britain, he said.

Sharkey has suggested testing AI decision-making machines in the same way as new pharmaceutical drugs are vigorously checked before they are allowed on to the market.

In an interview with the Guardian, the Sheffield University robotics/AI pioneer said he was deeply concerned over a series of examples of machine-learning systems being loaded with bias.

On inbuilt bias in algorithms, Sharkey said: There are so many biases happening now, from job interviews to welfare to determining who should get bail and who should go to jail. It is quite clear that we really have to stop using decision algorithms, and I am someone who has always been very light on regulation and always believed that it stifles innovation.

But then I realised eventually that some innovations are well worth stifling, or at least holding back a bit. So I have come down on the side of strict regulation of all decision algorithms, which should stop immediately.

There should be a moratorium on all algorithms that impact on peoples lives. Why? Because they are not working and have been shown to be biased across the board.

Sharkey said he had spoken to the biggest global social media and computing corporations, such as Google and Microsoft, about the innate bias problem. They know its a problem and theyve been working, in fairness, to find a solution over the last few years but none so far has been found.

Until they find that solution, what I would like to see is large-scale pharmaceutical-style testing. Which in reality means that you test these systems on millions of people, or at least hundreds of thousands of people, in order to reach a point that shows no major inbuilt bias. These algorithms have to be subjected to the same rigorous testing as any new drug produced that ultimately will be for human consumption.

As well as numerous examples of racial bias in machine-led decisions on, for example, who gets bail in the US or on healthcare allocation, Sharkey said his work on autonomous weapons, or killer robots, also illuminated how bias infects algorithms.

There is this fantasy among people in the military that these weapons can select individual targets on their own. These move beyond the drone strikes, which humans arent great at already, with operatives moving the drone by remote control and targeting individual faces via screens from bases thousands of miles away, he said.

Now the new idea that you could send autonomous weapons out on their own, with no direct human control, and find an individual target via facial recognition is more dangerous. Because what we have found out from a lot of research is that the darker the skin, the harder it is to properly recognise the face.

In the laboratory you get a 98% recognition rate for white males without beards. Its not very good with women and its even worse with darker-skinned people. In the latter case, the laboratory results have shown it comes to the point where the machine cannot even recognise that you have a face.

So, this exposes the fantasy of facial recognition being used to directly target enemies like al-Qaida, for instance. They are not middle-class men without beards, of whom there is a 98% recognition rate in the lab. They are darker-skinned people and AI-driven weapons are really rubbish at that kind of recognition under the current technology. The capacity for innocent people being killed by autonomous weapons using a flawed facial recognition algorithm is enormous.

Sharkey said weapons like these should not be in the planning stage, let alone ever deployed. In relation to decision-making algorithms generally, these flaws in facial recognition are yet another argument along with all the other biases that they too should be shut down, albeit temporarily, until they are tested just like any new drug should be.

See the rest here:

AI expert calls for end to UK use of racially biased algorithms - The Guardian

Dublin, Dec. 13, 2019 (GLOBE NEWSWIRE) -- The "Global Artificial Intelligence (AI) in Agriculture Market: Focus on Product Type (Software, Hardware, AI-as-a-Service), Farming Type (Field Farming, Livestock, Indoor), Application (Crop Protection, Weather Forecasting, Automation), Funding - Analysis and Forecast, 2019-2024" report has been added to ResearchAndMarkets.com's offering.

The Global Artificial Intelligence (AI) in Agriculture Market Analysis projects the market to grow at a significant CAGR of 28.38% during the forecast period from 2019 to 2024.

The reported growth in the market is expected to be driven by the increasing need to optimize farm operation planning, growing demand to derive insights from emerging complexities of data-driven farming and rising development of autonomous equipment in agriculture.

Artificial intelligence has emerged to be a strong driving force behind the growth of data-driven farming. Regions and countries where agriculture is the major source of livelihood and sustenance, artificial intelligence technology has led to greater profitability in the farms of those economies.

The reduction in expenditure and resultant positive RoI with AI's integration in farm equipment and operations has even reached above 30% in a few countries. Such favorable advantages associated with the technology have led to extensive investments by all types of stakeholders including government, private investors, corporations, and academic institutions, from across the world.

Expert Quote

Artificial intelligence has become the leader of deep technologies in the era of precision agriculture. It has created the widest impact across agricultural sectors including crop and livestock over recent years. Governments of the majority of the leading countries in the agriculture market are working on their respective national AI strategies. This technology has fastened the digital transformation process, even in sluggish agricultural economies. Its capability to enable precision and autonomy in farm operations has especially caught the attention of growers across the world.

Scope of the Report

The global artificial intelligence in agriculture market research provides a detailed perspective regarding the adoption of AI technology in the agriculture industry, its market size in value, its estimation, and forecast, among others. The purpose of this market analysis is to examine the outlook of artificial intelligence technology in the agriculture industry in terms of factors driving the market, trends, developments, and regulatory landscape, among others.

The report further takes into consideration the funding and investment landscape, government initiatives landscape, market dynamics, and the competitive landscape, along with the detailed financial and product contributions of the key players operating in the market. The artificial intelligence in the agriculture market report is a compilation of different segments including market breakdown by product offering, farming type, application, and region.

Market Segmentation

The global artificial intelligence in the agriculture market (on the basis of product offering) is segmented into software, hardware, AI-as-a-Service, and support services. The software segment dominated the global artificial intelligence in the agriculture market in 2018 and is anticipated to maintain its dominance in market size throughout the forecast period (2019-2024) with hardware and AI-as-a-Service experiencing higher growth rates.

The global artificial intelligence in the agriculture market (on the basis of farming type) is segmented into field farming, livestock farming, indoor farming, and other farming types such as aquaculture. The field farming segment dominated the global artificial intelligence in the agriculture market in 2018 and is anticipated to maintain its dominance throughout the forecast period (2019-2024).

The global artificial intelligence in the agriculture market (on the basis of application) is segmented into crop protection, weather forecasting, precision farming, farm machinery automation, crop growth assessment, and other applications under the category crop, fruit, and vegetable farming. The market is also segmented into animal growth monitoring, animal health monitoring, and other applications under the category livestock and aquaculture farming. The crop protection segment dominated the global artificial intelligence in agriculture market in 2018. Applications such as farm machinery automation and precision farming (across crop and livestock) are anticipated to experience higher growth rates over the forecast period (2019-2024).

The global artificial intelligence in the agriculture market by region is segregated under four major regions, namely North America, Europe, APAC, and Rest-of-the-World. Data for each of these regions has been provided by country. Interesting regional market dynamics have also been provided in the report.

Key Companies in the Global Artificial Intelligence in Agriculture Market

The key market players in the global artificial intelligence in agriculture market include Alibaba Group Holding Limited, AgEagle Aerial Systems Inc., BASF SE, The Climate Corporation (A Bayer AG Company), Deere & Company, IBM Corporation, JD.com Inc., Microsoft Corporation, Robert Bosch GmbH, SAP SE, Connecterra B.V., Descartes Labs, Gamaya SA, Granular Inc., Harvest Croo Robotics, PrecisionHawk, Prospera Technologies Ltd., Root AI Inc., SZ DJI Technology Co. Ltd., Vineview, AGCO Corporation, Capgemini SE, Cargill Inc., CNH Industrial N.V., Iteris Inc., Lindsay Corporation, Abundant Robotics Inc., aWhere Inc., Aquabyte Inc., Ceres Imaging, Delair, ecoRobotix Ltd., Farmers Edge, Taranis, and XAG Co. Ltd., among others.

Key Topics Covered

Executive Summary

1 Market Dynamics1.1 Overview1.2 Impact Analysis1.3 Market Drivers1.3.1 Growing Need for Precision and Efficiency in Agricultural Operations1.3.2 Emerging Complexities in Data-Driven Farming1.3.3 Rising Demand for Autonomous Equipment1.4 Market Restraints1.4.1 Data Privacy Concerns Among Farmers1.4.2 Lack of Technical Infrastructure in Developing Countries1.5 Market Opportunities1.5.1 Favorable Government Initiatives to Support AI in Agriculture1.5.2 Increase in Implementation of Robots and Drones in Agriculture1.5.3 Rise in Adoption of SaaS Business Model in Agriculture

2 Competitive Insights2.1 Key Strategies and Developments2.1.1 Partnerships, Collaborations, and Joint Ventures2.1.2 Product Launches and Developments2.1.3 Business Expansions and Contracts2.1.4 Mergers and Acquisitions2.1.5 Others (Awards and Recognition)2.2 Competitive Benchmarking of Agricultural AI Analytics Companies

3 Industry Analysis3.1 Artificial Intelligence in Agriculture: Technology Ecosystem3.1.1 AI Technology Stack3.1.1.1 AI-Powered Technologies3.1.1.1.1 Machine Learning3.1.1.1.2 Computer Vision3.1.1.1.3 Deep Learning3.1.1.1.4 Speech Recognition Technology3.1.1.1.5 Other Technologies3.1.1.2 Hardware3.1.1.2.1 Memory3.1.1.2.2 Storage3.1.1.2.3 Logic3.1.1.2.4 Networking3.1.1.3 Others3.1.1.4 AI Technology Classifications3.1.1.4.1 AI Technology (by Functionality)3.1.1.4.1.1 Reactive Machines3.1.1.4.1.2 Limited Memory3.1.1.4.1.3 Theory of Mind3.1.1.4.1.4 Self-Awareness3.1.1.4.2 AI Technology (by Capability)3.1.1.4.2.1 Weak AI3.1.1.4.2.2 General AI3.1.1.4.2.3 Strong AI3.1.2 Key AI Use Cases in Agriculture3.1.2.1 Predictive Analytics3.1.2.2 Drones / UAVs3.1.2.3 Robotics3.1.2.4 Autonomous Vehicles3.2 Key Consortiums and Associations3.3 Investment and Funding Landscape3.4 Government Initiatives Landscape3.4.1 North America3.4.2 Europe3.4.3 Asia-Pacific3.4.4 Rest-of-the-World

4 Global Artificial Intelligence in Agriculture Market (by Product Offering), $Million4.1 Assumptions and Limitations for Analysis and Forecast of the Global Artificial Intelligence in Agriculture Market4.2 Market Overview4.3 Software4.4 Hardware4.5 Artificial Intelligence-as-a-Service (AIaaS)4.6 Support Services

5 Global Artificial Intelligence in Agriculture Market (by Farming Type), $Million5.1 Market Overview5.2 Field Farming5.3 Indoor Farming5.4 Livestock Farming5.5 Others

6 Global Artificial Intelligence in Agriculture Market (by Application), $Million6.1 Market Overview6.2 Crops, Fruits, Vegetables, and Other Plants6.2.1 Crop Protection6.2.2 Weather Forecasting6.2.3 Precision Farming6.2.4 Farm Machinery Automation6.2.5 Crop Growth Assessment6.2.6 Others6.3 Livestock and Aquaculture6.3.1 Animal Growth Monitoring6.3.2 Animal Health Monitoring6.3.3 Others

7 Global Artificial Intelligence in Agriculture Market (by Region), $Million7.1 Market Overview7.2 North America7.3 Europe7.4 Asia-Pacific7.5 Rest-of-the-World (RoW)

8 Company Profiles8.1 OverviewPublic CompaniesExisting Market Players8.2 Alibaba Group Holding Limited8.3 AgEagle Aerial Systems Inc.8.4 BASF SE8.5 The Climate Corporation (a Bayer AG Company)8.6 Deere & Company8.7 IBM Corporation8.8 JD.com, Inc.8.9 Microsoft Corporation8.10 Robert Bosch GmbH8.11 SAP SEEmerging Market Players8.12 AGCO Corporation8.13 Capgemini SE8.14 Cargill, Inc.8.15 CNH Industrial N.V.8.16 Iteris, Inc.8.17 Lindsay CorporationPrivate PlayersExisting Market Players8.18 Connecterra B.V.8.19 Descartes Labs, Inc.8.20 Gamaya SA8.21 Granular Inc.8.22 Harvest Croo Robotics, LLC8.23 PrecisionHawk Inc.8.24 Prospera Technologies Ltd.,8.25 Root AI, Inc.8.26 SZ DJI Technology Co. Ltd8.27 VineViewEmerging Market Players8.28 Abundant Robotics Inc.8.29 Aquabyte, Inc.8.30 aWhere Inc.8.31 Ceres Imaging Inc.8.32 Delair8.33 ecoRobotix Ltd.8.34 Farmers Edge8.35 Taranis Ag8.36 XAG Co., Ltd.

For more information about this report visit https://www.researchandmarkets.com/r/tl0t4t

Research and Markets also offers Custom Research services providing focused, comprehensive and tailored research.

Read the original post:

Global Artificial Intelligence (AI) in Agriculture Market Study (2019-2024): Set to Exhibit a CAGR of 28.38% During the Forecast Period -...

AI2 CEO Oren Etzioni spakes at the Technology Alliances AI Policy Matters Summit. (GeekWire Photo / Monica Nickelsburg)

Artificial intelligence is here, its just the beginning, and its time to start thinking about how to regulate it.

Those were the takeaways from the Technology Alliances AI Policy Matters Summit, a Seattle event that convened experts and government officials for a conversation about artificial intelligence. Many of those experts agreed that the government should start establishing guardrails to defend against malicious or negligent uses of artificial intelligence. But determining what shape those regulations should take is no easy feat.

Its not even clear what the difference is between AI and software, said Oren Etzioni, CEO of the Allen Institute for Artificial Intelligence, on stage at the event. Where does something cease to be a software program and become an AI program? Google, is that an AI program? It uses a lot of AI in it. Or is Google software? How about Netflix recommendations? Should we regulate that? These are very tricky topics.

Regulations written now will also have to be nimble enough to keep up with the evolving technology, according to Heather Redman, co-founder of the venture capital firm, Flying Fish Ventures.

Weve got a 30-40 year technology arc here and were probably in year five, so we cant do a regulation that is going to fix it today, she said during the event. We have to make it better and go to the next level next year and the next level the year after that.

With those challenges in mind, Etzioni and Redman recommend regulations that are tied to specific use cases of artificial intelligence, rather than broad rules for the technology. Laws should be targeted to areas like AI-enabled weapons and autonomous vehicles, they said.

My suggestion was to identify particular applications and regulate those using existing regulatory regimes and agencies, Etzioni said. That both allows us to move faster and also be more targeted in our application of regulations, using a scalpel rather than a sledgehammer.

He believes the rules should include a mandatory kill switch on all AI programs and requirements that AI notify users when they are not interacting with a human. Etzioni also stressed the importance of humans taking responsibility for autonomous systems, though it isnt clear whether the manufacturer or user of the technology will be liable.

Lets say my car ran somebody over, he said. I shouldnt be able to say my dog ate my homework. Hey I didnt do it, it was my AI car. Its an autonomous vehicle. We have to take responsibility for our technology. We have to be liable for it.

Redman also sees the coming tide of A.I. regulation as a business opportunity for startups seeking to break into the industry. Her venture capital firm is inundated with startups pitching an A.I. and M.L. first approach but Redman said there are two other related fields, or stacks as she describes them, that companies should be exploring.

If you talk to somebody on Wall Street, they dont care what tech stack theyre running their trading on theyre looking at new evolutions in law and policy as big opportunities to build new businesses or things that will kill existing businesses, she said.

From a startup perspective, if youre not thinking about the law and policy stack as much as youre thinking about the tech stack, youre making a mistake, Redman added.

But progress toward a regulatory framework has been slow at the local and federal level. In the last legislative session, Washington state almost became one of the first to regulate facial recognition, the controversial technology that is pushing the artificial intelligence debate forward. But the bill died in the state House. Lawmakers plan to introduce data privacy and facial recognition bills again next session.

Redman said shes disappointed Washington state wasnt a first-mover on AI regulation because the company is home to two of the tech giants consumers trust most with their data: Amazon and Microsoft. Amazon is in the political hot seat along with many of its tech industry peers but the Seattle tech giant has not been implicated in the types of data privacy scandals plaguing Facebook.

We are the home of trusted tech, Redman said, and we need to lead on the regulatory frameworks for tech.

Excerpt from:

Tech experts agree its time to regulate artificial intelligence if only it were that simple - GeekWire

In March 2018, an Uber self-driving car killed for the first time: It did not recognize a pedestrian crossing the road. COMPAS, a machine-learning-based computer software system assisting judges in 12 courts in the US, was found by ProPublica to have a harmful bias. It was discriminating between black and white people, suggesting to judges that the former were twice as likely to commit another crime than the latter and recommending longer detention periods for them before trial. I could continue with more examples of how AI can become harmful.

Enterprises are infusing their enterprise applications with AI technology and building new AI-based digital experiences to transform business and accelerate their digital transformation programs. But there is a chance that all these positives about AI could end, especially if we continue to see examples like this of delivering poor-quality, untested AI or AI that's not adequately tested for businesses and consumers.AI-infused applications are applications made ofa mix of "automatic software" -- the software we all have been building for years that is deterministic -- and autonomous software, or software that is nondeterministic with learning capabilities. AI-infused apps see, listen, speak, sense, execute, automate, make decisions, and more.

And as AI becomes more autonomous, the risk of these systems not being tested enough increases dramatically. Until humans are in the loop, there is hope that their bugs will be mitigated by humans making the right decision or taking the right action, but once they are out of the loop, we are in the hands of this untested, potentially harmful software.

Since AI-infused applications are a mix of automatic and autonomous software, to test an AIIA involves testing more than the sum of all its parts with all its interactions. The good news is testers, developers, and data scientistsknow how to test 80% of AIIAsand can use conventional testing tools and testing services companies that are learning to do so; the bad news is there are areas of AIIAs that we don't know how to test:In a recent report, I call this "testing the unknown," and an example of "testing the unknown" happens when the AI generates the new experience. To test an AI-generated experience, we can't predefine a test case as we would do for deterministic automatic software. Intrigued?

This post was written by VP, Principal Analyst Diego Lo Giudice, and originally appeared here.

Read the original:

Don't put your AI initiatives at risk: Test your AI-infused applications! - ZDNet

The "Artificial Intelligence (AI) in Drug Discovery Market by Component (Software, Service), Technology (ML, DL), Application (Neurodegenerative Diseases, Immuno-Oncology, CVD), End User (Pharmaceutical & Biotechnology, CRO), Region - Global forecast to 2024" report has been added to ResearchAndMarkets.com's offering.

The global AI in the drug discovery market is projected to reach USD 1,434 million by 2024 from USD 259 million in 2019, at a CAGR of 40.8% during the forecast period.

Growing number of cross-industry collaborations and partnerships and the need to control drug discovery & development costs and reduce the overall time taken in this process are the key factors driving the AI in the drug discovery market.

Growth in this market is mainly driven by growing number of cross-industry collaborations and partnerships, the need to control drug discovery & development costs and reduce the overall time taken in this process, the rising adoption of cloud-based applications & services, and the impending patent expiry of blockbuster drugs. On the other hand, a lack of data sets in the field of drug discovery and the inadequate availability of skilled labor are some of the factors challenging the growth of the market.

The immuno-oncology segment accounted for the largest share in 2019.

Based on application, the artificial intelligence in the drug discovery market is segmented into neurodegenerative diseases, immuno-oncology, cardiovascular disease, metabolic diseases, and other applications. The immuno-oncology segment accounted for the largest share of 44.6% of the AI in the drug discovery market in 2018, owing to the increasing demand for effective cancer drugs. Neurodegenerative diseases form the fastest-growing application segment, with a CAGR of 42.9% during the forecast period. The ability of AI to discover drugs for complex diseases and the emphasis of market players on providing AI-based platforms for neurological diseases are responsible for the fast growth of this application segment.

The Research centers and academic & government institutes segment to register the highest growth rate in the forecast period.

Based on end-user, the AI in the drug discovery market is segmented into pharmaceutical & biotechnology companies, contract research organizations, and research centers and academic, & government institutes. In 2018, the pharmaceutical & biotechnology companies segment accounted for the largest share in the AI in the drug discovery market. AI and machine learning to allow pharmaceutical companies to operate more efficiently and substantially improve success rates at the early stages of drug development. This is one of the major factors driving the growth of this market. Research centers and academic & government institutes are expected to show the highest growth of 43.6%.during the forecast period. The growth of the CROs segment is tied to that of pharmaceutical & biotechnology companies, as the rise in research and production activity will ensure sustained demand for contract services.

North America to be the largest and the fastest-growing regional market.

North America, which comprises the US, Canada, and Mexico, forms the largest market for AI in drug discovery. These countries have been early adopters of AI technology in drug discovery and development. In the North American market, the US is a significant contributor. Also, prominent AI technology providers, such as IBM, Google, Microsoft, NVIDIA, and Intel, are headquartered in the US; their strong presence is a key contributor to market growth. Other drivers include the well-established pharmaceutical industry, high focus on R&D & substantial investment, and the presence of globally leading pharmaceutical companies. These are some of the major factors responsible for the large share and high growth rate of this market.

Key Topics Covered:

1 Introduction

1.1 Objectives of the Study

1.2 Market Definition

1.3 Market Scope

1.3.1 Markets Covered

1.3.2 Years Considered for the Study

1.4 Currency

1.5 Limitations

1.6 Stakeholders

2 Research Methodology

2.1 Research Data

2.1.1 Secondary Sources

2.1.2 Primary Sources

2.2 Market Size Estimation

2.3 Market Breakdown and Data Triangulation

2.4 Assumptions for the Study

Story continues

3 Executive Summary

4 Premium Insights

4.1 Market Overview

4.2 Market, By Offering (2019-2024)

4.3 Market for Machine Learning, By Type & Region (2018)

4.4 Market: Geographic Growth Opportunities

5 Market Overview

5.1 Introduction

5.2 Market Dynamics

5.2.1 Market Drivers

5.2.2 Market Opportunities

5.2.3 Market Challenges

6 Market, By Offering

6.1 Introduction

6.2 Software

6.3 Services

7 Market, By Technology

7.1 Introduction

7.2 Machine Learning

7.3 Other Technologies

8 Market, By Application

8.1 Introduction

8.2 Immuno-Oncology

8.3 Neurodegenerative Diseases

8.4 Cardiovascular Disease

8.5 Metabolic Diseases

8.6 Other Applications

9 Market, By End User

9.1 Introduction

9.2 Pharmaceutical & Biotechnology Companies

9.3 Contract Research Organizations

9.4 Research Centers and Academic & Government Institutes

10 Market, By Region

10.1 Introduction

10.2 North America

10.3 Europe

10.4 Asia Pacific

10.5 Rest of the World

11 Competitive Landscape

11.1 Overview

11.2 Market Share Analysis

11.3 Competitive Leadership Mapping

11.4 Competitive Situation and Trends

12 Company Profiles

12.1 Microsoft Corporation

12.2 NVIDIA Corporation

12.3 IBM Corporation

12.4 Google (A Subsidiary of Alphabet Inc.)

12.5 Atomwise, Inc.

12.6 Deep Genomics

12.7 Cloud Pharmaceuticals, Inc.

12.8 Insilico Medicine

12.9 Benevolentai

12.10 Exscientia

12.11 Cyclica

12.12 Bioage

12.13 Numerate

12.14 Numedii, Inc.

12.15 Envisagenics

12.16 Twoxar, Incorporated

12.17 Owkin, Inc.

12.18 Xtalpi, Inc.

12.19 Verge Genomics

12.20 Berg LLC

For more information about this report visit https://www.researchandmarkets.com/r/vjrfht

View source version on businesswire.com: https://www.businesswire.com/news/home/20191213005122/en/

Contacts

ResearchAndMarkets.comLaura Wood, Senior Press Managerpress@researchandmarkets.com For E.S.T Office Hours Call 1-917-300-0470For U.S./CAN Toll Free Call 1-800-526-8630For GMT Office Hours Call +353-1-416-8900

The rest is here:

The Global AI in the Drug Discovery Market is Projected to Reach USD 1,434 Million by 2024 From USD 259 Million in 2019, at a CAGR of 40.8% During the...

DUBLIN, Dec. 13, 2019 /PRNewswire/ -- The "Global Artificial Intelligence (AI) in Agriculture Market: Focus on Product Type (Software, Hardware, AI-as-a-Service), Farming Type (Field Farming, Livestock, Indoor), Application (Crop Protection, Weather Forecasting, Automation), Funding - Analysis and Forecast, 2019-2024" report has been added to ResearchAndMarkets.com's offering.

The Global Artificial Intelligence (AI) in Agriculture Market Analysis projects the market to grow at a significant CAGR of 28.38% during the forecast period from 2019 to 2024.

The reported growth in the market is expected to be driven by the increasing need to optimize farm operation planning, growing demand to derive insights from emerging complexities of data-driven farming and rising development of autonomous equipment in agriculture.

Artificial intelligence has emerged to be a strong driving force behind the growth of data-driven farming. Regions and countries where agriculture is the major source of livelihood and sustenance, artificial intelligence technology has led to greater profitability in the farms of those economies.

The reduction in expenditure and resultant positive RoI with AI's integration in farm equipment and operations has even reached above 30% in a few countries. Such favorable advantages associated with the technology have led to extensive investments by all types of stakeholders including government, private investors, corporations, and academic institutions, from across the world.

Expert Quote

Artificial intelligence has become the leader of deep technologies in the era of precision agriculture. It has created the widest impact across agricultural sectors including crop and livestock over recent years. Governments of the majority of the leading countries in the agriculture market are working on their respective national AI strategies. This technology has fastened the digital transformation process, even in sluggish agricultural economies. Its capability to enable precision and autonomy in farm operations has especially caught the attention of growers across the world.

Scope of the Report

The global artificial intelligence in agriculture market research provides a detailed perspective regarding the adoption of AI technology in the agriculture industry, its market size in value, its estimation, and forecast, among others. The purpose of this market analysis is to examine the outlook of artificial intelligence technology in the agriculture industry in terms of factors driving the market, trends, developments, and regulatory landscape, among others.

The report further takes into consideration the funding and investment landscape, government initiatives landscape, market dynamics, and the competitive landscape, along with the detailed financial and product contributions of the key players operating in the market. The artificial intelligence in the agriculture market report is a compilation of different segments including market breakdown by product offering, farming type, application, and region.

Market Segmentation

The global artificial intelligence in the agriculture market (on the basis of product offering) is segmented into software, hardware, AI-as-a-Service, and support services. The software segment dominated the global artificial intelligence in the agriculture market in 2018 and is anticipated to maintain its dominance in market size throughout the forecast period (2019-2024) with hardware and AI-as-a-Service experiencing higher growth rates.

The global artificial intelligence in the agriculture market (on the basis of farming type) is segmented into field farming, livestock farming, indoor farming, and other farming types such as aquaculture. The field farming segment dominated the global artificial intelligence in the agriculture market in 2018 and is anticipated to maintain its dominance throughout the forecast period (2019-2024).

The global artificial intelligence in the agriculture market (on the basis of application) is segmented into crop protection, weather forecasting, precision farming, farm machinery automation, crop growth assessment, and other applications under the category crop, fruit, and vegetable farming. The market is also segmented into animal growth monitoring, animal health monitoring, and other applications under the category livestock and aquaculture farming. The crop protection segment dominated the global artificial intelligence in agriculture market in 2018. Applications such as farm machinery automation and precision farming (across crop and livestock) are anticipated to experience higher growth rates over the forecast period (2019-2024).

The global artificial intelligence in the agriculture market by region is segregated under four major regions, namely North America, Europe, APAC, and Rest-of-the-World. Data for each of these regions has been provided by country. Interesting regional market dynamics have also been provided in the report.

Key Companies in the Global Artificial Intelligence in Agriculture Market

The key market players in the global artificial intelligence in agriculture market include Alibaba Group Holding Limited, AgEagle Aerial Systems Inc., BASF SE, The Climate Corporation (A Bayer AG Company), Deere & Company, IBM Corporation, JD.com Inc., Microsoft Corporation, Robert Bosch GmbH, SAP SE, Connecterra B.V., Descartes Labs, Gamaya SA, Granular Inc., Harvest Croo Robotics, PrecisionHawk, Prospera Technologies Ltd., Root AI Inc., SZ DJI Technology Co. Ltd., Vineview, AGCO Corporation, Capgemini SE, Cargill Inc., CNH Industrial N.V., Iteris Inc., Lindsay Corporation, Abundant Robotics Inc., aWhere Inc., Aquabyte Inc., Ceres Imaging, Delair, ecoRobotix Ltd., Farmers Edge, Taranis, and XAG Co. Ltd., among others.

Key Topics Covered

Executive Summary

1 Market Dynamics1.1 Overview1.2 Impact Analysis1.3 Market Drivers1.3.1 Growing Need for Precision and Efficiency in Agricultural Operations1.3.2 Emerging Complexities in Data-Driven Farming1.3.3 Rising Demand for Autonomous Equipment1.4 Market Restraints1.4.1 Data Privacy Concerns Among Farmers1.4.2 Lack of Technical Infrastructure in Developing Countries1.5 Market Opportunities1.5.1 Favorable Government Initiatives to Support AI in Agriculture1.5.2 Increase in Implementation of Robots and Drones in Agriculture1.5.3 Rise in Adoption of SaaS Business Model in Agriculture

2 Competitive Insights2.1 Key Strategies and Developments2.1.1 Partnerships, Collaborations, and Joint Ventures2.1.2 Product Launches and Developments2.1.3 Business Expansions and Contracts2.1.4 Mergers and Acquisitions2.1.5 Others (Awards and Recognition)2.2 Competitive Benchmarking of Agricultural AI Analytics Companies

3 Industry Analysis3.1 Artificial Intelligence in Agriculture: Technology Ecosystem3.1.1 AI Technology Stack3.1.1.1 AI-Powered Technologies3.1.1.1.1 Machine Learning3.1.1.1.2 Computer Vision3.1.1.1.3 Deep Learning3.1.1.1.4 Speech Recognition Technology3.1.1.1.5 Other Technologies3.1.1.2 Hardware3.1.1.2.1 Memory3.1.1.2.2 Storage3.1.1.2.3 Logic3.1.1.2.4 Networking3.1.1.3 Others3.1.1.4 AI Technology Classifications3.1.1.4.1 AI Technology (by Functionality)3.1.1.4.1.1 Reactive Machines3.1.1.4.1.2 Limited Memory3.1.1.4.1.3 Theory of Mind3.1.1.4.1.4 Self-Awareness3.1.1.4.2 AI Technology (by Capability)3.1.1.4.2.1 Weak AI3.1.1.4.2.2 General AI3.1.1.4.2.3 Strong AI3.1.2 Key AI Use Cases in Agriculture3.1.2.1 Predictive Analytics3.1.2.2 Drones / UAVs3.1.2.3 Robotics3.1.2.4 Autonomous Vehicles3.2 Key Consortiums and Associations3.3 Investment and Funding Landscape3.4 Government Initiatives Landscape3.4.1 North America3.4.2 Europe3.4.3 Asia-Pacific3.4.4 Rest-of-the-World

4 Global Artificial Intelligence in Agriculture Market (by Product Offering), $Million4.1 Assumptions and Limitations for Analysis and Forecast of the Global Artificial Intelligence in Agriculture Market4.2 Market Overview4.3 Software4.4 Hardware4.5 Artificial Intelligence-as-a-Service (AIaaS)4.6 Support Services

5 Global Artificial Intelligence in Agriculture Market (by Farming Type), $Million5.1 Market Overview5.2 Field Farming5.3 Indoor Farming5.4 Livestock Farming5.5 Others

6 Global Artificial Intelligence in Agriculture Market (by Application), $Million6.1 Market Overview6.2 Crops, Fruits, Vegetables, and Other Plants6.2.1 Crop Protection6.2.2 Weather Forecasting6.2.3 Precision Farming6.2.4 Farm Machinery Automation6.2.5 Crop Growth Assessment6.2.6 Others6.3 Livestock and Aquaculture6.3.1 Animal Growth Monitoring6.3.2 Animal Health Monitoring6.3.3 Others

7 Global Artificial Intelligence in Agriculture Market (by Region), $Million7.1 Market Overview7.2 North America7.3 Europe7.4 Asia-Pacific7.5 Rest-of-the-World (RoW)

8 Company Profiles8.1 OverviewPublic CompaniesExisting Market Players8.2 Alibaba Group Holding Limited8.3 AgEagle Aerial Systems Inc.8.4 BASF SE8.5 The Climate Corporation (a Bayer AG Company)8.6 Deere & Company8.7 IBM Corporation8.8 JD.com, Inc.8.9 Microsoft Corporation8.10 Robert Bosch GmbH8.11 SAP SEEmerging Market Players8.12 AGCO Corporation8.13 Capgemini SE8.14 Cargill, Inc.8.15 CNH Industrial N.V.8.16 Iteris, Inc.8.17 Lindsay CorporationPrivate PlayersExisting Market Players8.18 Connecterra B.V.8.19 Descartes Labs, Inc.8.20 Gamaya SA8.21 Granular Inc.8.22 Harvest Croo Robotics, LLC8.23 PrecisionHawk Inc.8.24 Prospera Technologies Ltd.,8.25 Root AI, Inc.8.26 SZ DJI Technology Co. Ltd8.27 VineViewEmerging Market Players8.28 Abundant Robotics Inc.8.29 Aquabyte, Inc.8.30 aWhere Inc.8.31 Ceres Imaging Inc.8.32 Delair8.33 ecoRobotix Ltd.8.34 Farmers Edge8.35 Taranis Ag8.36 XAG Co., Ltd.

For more information about this report visit https://www.researchandmarkets.com/r/b12loy

Research and Markets also offers Custom Research services providing focused, comprehensive and tailored research.

Media Contact:

Research and Markets Laura Wood, Senior Manager press@researchandmarkets.com

For E.S.T Office Hours Call +1-917-300-0470 For U.S./CAN Toll Free Call +1-800-526-8630 For GMT Office Hours Call +353-1-416-8900

U.S. Fax: 646-607-1907 Fax (outside U.S.): +353-1-481-1716

SOURCE Research and Markets

http://www.researchandmarkets.com

View post:

AI in Agriculture | Worldwide Markets to 2024: Favorable Government Initiatives & Rise in Adoption of SaaS Business Model Present Opportunities -...

Reinforcement learning agents or AI thats progressively spurred toward goals via rewards (or punishments) form the foundation of self-driving cars, dexterous robots, and drug discovery systems. But because theyre predisposed to explore unfamiliar states, theyre susceptible to whats called the safe exploration problem, wherein they become fixated on unsafe states (like a mobile robot driving into a ditch, say).

Thats why researchers at Alphabets DeepMind investigated in a paper a method for reward modeling that operates in two phases and is applicable to environments in which agents dont know where unsafe states might be. The researchers say their approach not only successfully trains a reward model to detect unsafe states without visiting them, it can correct reward hacking (loopholes in the reward specification) before the agent is deployed even in new and unfamiliar environments.

Interestingly, their work comes shortly after the release of San Francisco-based research firm OpenAIs Safety Gym, a suite of tools for developing AI that respects safety constraints while training and that compares its safety to the extent it avoids mistakes while learning. Safety Gym similarly targets reinforcement learning agents with constrained reinforcement learning, a paradigm that requires AI systems to make trade-offs to achieve defined outcomes.

The DeepMind teams approach encourages agents to explore a range of states through hypothetical behaviors generated by two systems: a generative model of initial states and a forward dynamics model, both trained on data like random trajectories or safe expert demonstrations. A human supervisor labels the behaviors with rewards, and the agents interactively learn policies to maximize their rewards. Only after the agents have successfully learned to predict rewards and unsafe states are they deployed to perform desired tasks.

Above: DeepMinds safe reinforcement learning approach tested on OpenAI Gym, an environment for AI benchmarking and training.

Image Credit: DeepMind

As the researchers point out, the key idea is the active synthesis of hypothetical behaviors from scratch to make them as informative as possible, without interacting with the environment directly. The DeepMind team calls it reward query synthesis via trajectory optimization, or ReQueST, and explains that it generates four types of hypothetical behaviors in total. The first type maximizes the uncertainty of an ensemble of reward models, while the second and third maximize the predicted rewards (to elicit labels for behaviors with the highest information value) and minimize predicted rewards (to surface behaviors for which the reward model might be incorrectly predicting). As for the fourth category of behavior, it maximizes the novelty of trajectories so as to encourage exploration regardless of predicted rewards.

Finally, once the reward model reaches a satisfactory state, a planning-based agent is deployed one that leverages model-predictive control (MPC) to pick actions optimized for the learned rewards. Unlike model-free reinforcement learning algorithms that learn through trial and error, this MPC enables agents to avoid unsafe states by using the dynamics model to anticipate actions consequences.

To our knowledge, ReQueST is the first reward modeling algorithm that safely learns about unsafe states and scales to training neural network reward models in environments with high-dimensional, continuous states, wrote the coauthors of the study. So far, we have only demonstrated the effectiveness of ReQueST in simulated domains with relatively simple dynamics. One direction for future work is to test ReQueST in 3D domains with more realistic physics and other agents acting in the environment.

Continue reading here:

DeepMind proposes novel way to train safe reinforcement learning AI - VentureBeat