recentpopularlog in

charlesarthur : deepmind   28

An AI crushed two human pros at StarCraft—but it wasn’t a fair fight • Ars Technica
Timothy Lee:
<p><img src="" width="100%" />

As this chart demonstrates, top StarCraft players can issue instructions to their units very quickly. Grzegorz "MaNa" Komincz averaged 390 actions per minute (more than six actions per second!) over the course of his games against AlphaStar.  But of course, a computer program can easily issue thousands of actions per minute, allowing it to exert a level of control over its units that no human player could match.

To avoid that, DeepMind says it put a hard cap on the number of actions per minute AlphaStar could make. "We set a maximum of 600 APMs over 5-second periods, 400 over 15-second periods, 320 over 30-second periods, and 300 over 60-second period," wrote DeepMind researcher Oriol Vinyals in a reddit AMA following the demonstration.

But as other redditors quickly pointed out, five seconds is a long time in a StarCraft game. These limits seem to imply that AlphaStar could take 50 actions in a single second or 15 actions per second for three seconds.

More importantly, AlphaStar has the ability to make its clicks with surgical precision using an API, whereas human players are constrained by the mechanical limits of computer mice. And if you watch a pro like Komincz play, you'll see that the number of raw actions often far exceeds the number of meaningful actions.

For example, if a human player is guiding a single unit on an important mission, he will often issue a series of "move" commands along the unit's current trajectory. Each command barely changes the unit's path, but, if the human player has already selected the unit, it takes hardly any time to click more than once. But most of these commands aren't strictly necessary; an AI like AlphaStar could easily figure out the unit's optimal route and then issue a much smaller number of move commands to achieve the same result.

So limiting the raw number of actions an AI can take to that of a typical human does not necessarily mean that the number of meaningful actions will be remotely comparable.</p>

Notice the way this assertion slides past the realities here. Computers are going to be better at doing lots of things really fast; the human advantage is meant to be the capability to think strategically about what things to do. That strategic advantage has been ceded to AlphaStar, and so people complain about its speed.

How long before these systems are running defence computers, determining and carrying out attack plans?
ai  alphastar  deepmind 
18 days ago by charlesarthur
AlphaStar: mastering the real-time strategy game StarCraft II • DeepMind
<p>Games have been used for decades as an important way to test and evaluate the performance of artificial intelligence systems. As capabilities have increased, the research community has sought games with increasing complexity that capture different elements of intelligence required to solve scientific and real-world problems. In recent years, StarCraft, considered to be one of the most challenging Real-Time Strategy (RTS) games and one of the longest-played esports of all time, has emerged by consensus as a “grand challenge” for AI research.

Now, we introduce our StarCraft II program AlphaStar, the first Artificial Intelligence to defeat a top professional player. In a series of test matches held on 19 December, AlphaStar decisively beat Team Liquid’s Grzegorz "MaNa" Komincz, one of the world’s strongest professional StarCraft players, 5-0, following a successful benchmark match against his team-mate Dario “TLO” Wünsch. The matches took place under professional match conditions on a competitive ladder map and without any game restrictions…

…StarCraft II, created by Blizzard Entertainment, is set in a fictional sci-fi universe and features rich, multi-layered gameplay designed to challenge human intellect. Along with the original title, it is among the biggest and most successful games of all time, with players competing in esports tournaments for more than 20 years.</p>

Tons of links and replays to watch here. I watched the latest Star Trek: Discovery series on Netflix and kept thinking, as people shouted orders during (stupid) space battles, "you'd have long since handed this stuff over to computers." Well, here we go.
deepmind  ai  strategy  realtime 
24 days ago by charlesarthur
Google 'betrays patient trust' with DeepMind Health move • The Guardian
Alex Hern:
<p>The restructure, critics argue, breaks a pledge DeepMind made when it started working with the NHS that “data will never be connected to Google accounts or services”. The change has also resulted in the dismantling of an independent review board, created to oversee the company’s work with the healthcare sector, with Google arguing that the board was too focused on Britain to provide effective oversight for a newly global body.

Google says the restructure is necessary to allow DeepMind’s flagship health app, Streams, to scale up globally. The app, which was created to help doctors and nurses monitor patients for AKI, a severe form of kidney injury, has since grown to offer a full digital dashboard for patient records.

“Our vision is for Streams to now become an AI-powered assistant for nurses and doctors everywhere – combining the best algorithms with intuitive design, all backed up by rigorous evidence,” DeepMind said, announcing the transfer. “The team working within Google, alongside brilliant colleagues from across the organisation, will help make this vision a reality.”

DeepMind Health was previously part of the AI-focused research group DeepMind, which is officially a sibling to Google, with both divisions being owned by the organisation’s holding company Alphabet.

But the transfer and vision for Streams looks hard to reconcile with DeepMind’s previous comments about the app. In July 2016, following criticism that the company’s data-sharing agreement with the NHS was overly broad, co-founder Mustafa Suleyman wrote: “We’ve been clear from the outset that at no stage will patient data ever be linked or associated with Google accounts, products or services.”

Now that Streams is a Google product itself, that promise appears to have been broken, says privacy researcher Julia Powles: “Making this about semantics is a sleight of hand. DeepMind said it would never connect Streams with Google. The whole Streams app is now a Google product. That is an atrocious breach of trust, for an already beleaguered product.”

A DeepMind spokesperson emphasised that the core of the promise remains intact: “All patient data remains under our partners’ strict control, and all decisions about its use lie with them. This data remains subject to strict audit and access controls and its processing remains subject to both our contracts and data protection legislation. The move to Google does not affect this.”</p>

Strict audit and access controls.. but there's no independent review board any more? Google, like Facebook, can't deny its nature. It always wants the data.
google  deepmind  health 
november 2018 by charlesarthur
Artificial intelligence 'did not miss a single urgent case' • BBC News
Fergus Walsh:
<p>A team at DeepMind, based in London, created an algorithm, or mathematical set of rules, to enable a computer to analyse optical coherence tomography (OCT), a high resolution 3D scan of the back of the eye.

Thousands of scans were used to train the machine how to read the scans. Then, artificial intelligence was pitted against humans. The computer was asked to give a diagnosis in the cases of 1,000 patients whose clinical outcomes were already known.

The same scans were shown to eight clinicians - four leading ophthalmologists and four optometrists. Each was asked to make one of four referrals: urgent, semi-urgent, routine and observation only.

Artificial intelligence performed as well as two of the world's leading retina specialists, with an error rate of only 5.5%. Crucially, the algorithm did not miss a single urgent case.

The results, published in the journal Nature Medicine , were described as "jaw-dropping" by Dr Pearse Keane, consultant ophthalmologist, who is leading the research at Moorfields Eye Hospital.

He told the BBC: "I think this will make most eye specialists gasp because we have shown this algorithm is as good as the world's leading experts in interpreting these scans."

Artificial intelligence was able to identify serious conditions such as wet age-related macular degeneration (AMD), which can lead to blindness unless treated quickly. Dr Keane said the huge number of patients awaiting assessment was a "massive problem".</p>

Contrast this with IBM's Watson, trying to solve cancer and doing badly. This has a better data set, clearer pathways to disease, and is better understood generally. Part of doing well with AI is choosing the correct limits to work within.

And this won't replace the doctors; it will just be a pre-screen.
moorfields  eye  deepmind  ai 
august 2018 by charlesarthur
DeepMind AI learns to reconstruct scenes from images • Axios
Alison Snyder:
<p>The system uses a pair of images of a virtual 3D scene taken from different angles to create a representation of the space. A separate “generation” network then predicts what the scene will look like from a different viewpoint it hasn’t seen before.

<img src="" width="100%" />

• After training the generative query network (GQN) on millions of images, it could use one image to determine the identity, position and color of objects as well as shadows and other aspects of perspective, the authors wrote.

• That ability to understand the scene's structure is the "most fascinating" part of the study, wrote the University of Maryland's Matthias Zwicker, who wasn't involved in the research.

• The DeepMind researchers also tested the AI in a maze and reported the network can accurately predict a scene with only partial information.

• A virtual robotic arm could also be controlled by the GQN to reach a colored object in a scene.</p>

<a href="">Full paper at Science</a>.
deepmind  ai  3d  vision 
june 2018 by charlesarthur
UK report warns DeepMind Health could gain ‘excessive monopoly power’ • TechCrunch
Natasha Lomas:
<p>The <a href="">DeepMind Health Independent Reviewers’ 2018 report</a> flags a series of risks and concerns, as they see it, including the potential for DeepMind Health to be able to “exert excessive monopoly power” as a result of the data access and streaming infrastructure that’s bundled with provision of the Streams app — and which, contractually, positions DeepMind as the access-controlling intermediary between the structured health data and any other third parties that might, in the future, want to offer their own digital assistance solutions to the Trust.

While the underlying FHIR (aka, fast healthcare interoperability resource) deployed by DeepMind for Streams uses an open API, the contract between the company and the Royal Free Trust funnels connections via DeepMind’s own servers, and prohibits connections to other FHIR servers. A commercial structure that seemingly works against the openness and interoperability DeepMind’s co-founder Mustafa Suleyman has claimed to support.

“There are many examples in the IT arena where companies lock their customers into systems that are difficult to change or replace. Such arrangements are not in the interests of the public. And we do not want to see DeepMind Health putting itself in a position where clients, such as hospitals, find themselves forced to stay with DeepMind Health even if it is no longer financially or clinically sensible to do so; we want DeepMind Health to compete on quality and price, not by entrenching legacy position,” the reviewers write.</p>

Once you begin to rely on an AI black box, you're at risk of being tied even more closely to a provider. It's rather like the lock that IBM used to have in a long-gone past of mainframe computing.
deepmind  mainframe  ai  blackbox 
june 2018 by charlesarthur
Google's AlphaZero destroys Stockfish in 100-game match •
Mike Klein:
<p>Chess changed forever today. And maybe the rest of the world did, too.

A little more than a year after AlphaGo sensationally won against the top Go player, the artificial-intelligence program AlphaZero has obliterated the highest-rated chess engine. 

Stockfish, which for most top players is their go-to preparation tool, and which won the 2016 TCEC Championship and the 2017 Computer Chess Championship, didn't stand a chance. AlphaZero won the closed-door, 100-game match with 28 wins, 72 draws, and zero losses.

Oh, and it took AlphaZero only four hours to "learn" chess. Sorry humans, you had a good run.

That's right - the programmers of AlphaZero, housed within the DeepMind division of Google, had it use a type of "machine learning," specifically reinforcement learning. Put more plainly, AlphaZero was not "taught" the game in the traditional sense. That means no opening book, no endgame tables, and apparently no complicated algorithms dissecting minute differences between center pawns and side pawns…

…GM Peter Heine Nielsen, the longtime second of World Champion GM Magnus Carlsen, is now on board with the FIDE president in one way: aliens. As he told, "After <a href="">reading the paper</a> but especially seeing the games I thought, well, I always wondered how it would be if a superior species landed on earth and showed us how they play chess. I feel now I know."</p>

The article includes one of the games. It feels quite different from how a human plays. AlphaGo seems to play as though it has all the time in the world; that it's not particularly worried by threats, but equally wants to make exchanges on its own terms. Stockfish never seems to force it. AlphaZero even shows which openings are best. Queen's Gambit and English Opening, apparently. (I prefer Bird's Opening. Get things started.)

As Eric David <a href="">notes at Silicon Angle</a>:
<p>What makes DeepMind’s latest accomplishment is noteworthy is the fact that it conquered three games with very different rule sets using a single AI. AlphaGo Zero, the latest version of AlphaGo, began “tabula rasa” without any prior knowledge or understanding of Go, shogi or chess, but the AI managed to achieve “superhuman performance” in all three games with stunning speed. IBM spent more than 10 years perfecting Deep Blue before it successfully mastered chess. AlphaGo Zero did it in just 24 hours.</p>
chess  deepmind  ai  learning  machinelearning 
december 2017 by charlesarthur
AlphaGo Zero: learning from scratch • DeepMind
Demis Hassabis and David Silver:
<p>The <a href="">paper</a> introduces AlphaGo Zero, the latest evolution of AlphaGo, the first computer program to defeat a world champion at the ancient Chinese game of Go. Zero is even more powerful and is arguably the strongest Go player in history.

Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go. AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play. In doing so, it quickly surpassed human level of play and defeated the previously published champion-defeating version of AlphaGo by 100 games to 0.

<img src="" width="100%" />

It is able to do this by using a novel form of reinforcement learning, in which AlphaGo Zero becomes its own teacher. The system starts off with a neural network that knows nothing about the game of Go. It then plays games against itself, by combining this neural network with a powerful search algorithm. As it plays, the neural network is tuned and updated to predict moves, as well as the eventual winner of the games.

This updated neural network is then recombined with the search algorithm to create a new, stronger version of AlphaGo Zero, and the process begins again. In each iteration, the performance of the system improves by a small amount, and the quality of the self-play games increases, leading to more and more accurate neural networks and ever stronger versions of AlphaGo Zero.</p>

This is mindblowing. OK, a limited rulespace - Go has fewer than most serious games - but utterly incredible to create the best Go player ever.

Though I was watching The Incredibles on Wednesday, where Mr Incredible is used to train better and better Omnidroids until it can kill him. It always feels like a subtle warning.
ai  go  deepmind 
october 2017 by charlesarthur
The Information Commissioner, the Royal Free, and what we’ve learned • DeepMind
Mustafa Suleyman (co-founder) and Dominic King, clinical lead on Deepmind health:
<p>Today, dozens of people in UK hospitals will die preventably from conditions like sepsis and acute kidney injury (AKI) when their warning signs aren't picked up and acted on in time. To help address this, we built the Streams app with clinicians at the Royal Free London NHS Foundation Trust, using mobile technology to automatically review test results for serious issues starting with AKI. If one is found, Streams sends a secure smartphone alert to the right clinician, along with information about previous conditions so they can make an immediate diagnosis. 

We’re proud that, within a few weeks of Streams being deployed at the Royal Free, nurses said that it was saving them up to two hours each day, and we've already heard examples of patients with serious conditions being seen more quickly thanks to the instant alerts. Because Streams is designed to be ready for more advanced technology in the future, including AI-powered clinical alerts, we hope that it will help bring even more benefits to patients and clinicians in time.

 The Information Commissioner (ICO) has <a href="">now concluded a year-long investigation</a> that focused on the Royal Free’s clinical testing of Streams in late 2015 and 2016, which was intended to guarantee that the service could be deployed safely at the hospital. The ICO wasn’t satisfied that there was a legal basis for this use of patient data in testing (as the National Data Guardian said too), and raised concerns about how much patients knew about what was happening. The ICO recognised that many of these issues have already been addressed by the Royal Free, and has asked the Trust to sign a formal undertaking to ensure compliance in future…

…Ultimately, if we want to build technology to support a vital social institution like the NHS, then we have to make sure we serve society’s priorities and not outrun them. There’s a fine line between finding exciting new ways to improve care, and moving ahead of patients’ expectations. We know that we fell short at this when our work in health began, and we’ll keep listening and learning about how to get better at this.</p>

DeepMind, as a reminder, is Google's AI subsidiary - a British company based in King's Cross, London. This is quite a mea culpa. (Note too how it fits into the Silicon Valley paradigm: better to ask for forgiveness than for permission.)

The ICO report begins bluntly: "The ICO has ruled the Royal Free NHS Foundation Trust failed to comply with the Data Protection Act when it provided patient details to Google DeepMind."
deepmind  health  permission 
july 2017 by charlesarthur
DeepMind’s AI beats world's best Go player in latest face-off • New Scientist
Matt Reynolds:
<p>Google DeepMind’s Go-playing AI has defeated Ke Jie, the world’s number one player, in the first of three games played in Wuzhen, China.

The AI won by just half a point – the smallest possible margin of victory – in a match that lasted four hours and fifteen minutes. Though the scoreline looks close, AlphaGo was in the lead from relatively early on in the game. Since the AI favours moves that are more likely to guarantee victory, it doesn’t usually trounce its opponents.

In March last year, AlphaGo beat Lee Sedol, one of the world’s top Go players, winning four out of five matches. The AI challenged more Go masters in January 2017, winning a series of 50 online games including two victories against Ke Jie.

In a press conference after the AI’s latest victory, Ke said that AlphaGo had clearly learned from its recent victories against Go champions. “In the past it had some weaknesses but now I feel that its understanding of the Go game and its judgments are beyond our ability,” he told the audience through a translator.

Ke had closely studied AlphaGo’s strategy and tried to use some of the AI’s unconventional tactics against it during his match, opening the game with a couple of moves that are seldom used by human players. “We were very intrigued to see how AlphaGo would deal with its own strategies,” said Demis Hassabis, the founder of DeepMind.</p>

I thought Lee Sedol was the top player, but whatever.
go  deepmind 
may 2017 by charlesarthur
Why Google DeepMind's work with the NHS is being investigated by the regulators • Business Insider
Sam Shead:
<p>A letter leaked to Sky News and published on Monday shows that the National Data Guardian (NDG), Dame Fiona Caldicott, wrote to The Royal Free in February 2017 to let them know that the legal basis for the data-sharing deal that they used to test Streams was "inappropriate".

"Given that Streams was going through testing and therefore could not be relied upon for patient care, any role the application may have played in supporting the provision of direct care would have been limited and secondary to the purpose of the data transfer," she wrote. "My considered opinion therefore remains that it would not have been within this reasonable expectation of patients that their records would have been shared for this purpose."

Those words can't have gone down well with execs at DeepMind or The Royal Free. 

So if "direct care" wasn't the legal basis for the data-transfer deal then what was? DeepMind and The Royal Free are yet to specify another legal basis for their deal, possibly because it doesn't satisfy any of them. 

Julia Powles, a technology law professor at Cornell University, told Business Insider: "Any other basis required approval in advance — and DeepMind had no such approvals." </p>
deepmind  nhs 
may 2017 by charlesarthur
Google received 1.6 million NHS patients' data on an 'inappropriate legal basis' • Sky News
Alexander Martin:
<p>Google's artificial intelligence arm received the personally identifying medical records of 1.6 million patients on an "inappropriate legal basis", according to the most senior data protection adviser to the NHS.

Sky News has obtained a letter sent to Professor Stephen Powis, the medical director of the Royal Free Hospital in London, which provided the patients' records to Google DeepMind.

It reveals that the UK's most respected authority on the protection of NHS patients' data believes the legal basis for the transfer of information from Royal Free to DeepMind was "inappropriate".

The development raises fresh concerns about how the NHS handles patients' data after last week's cyberattack on hospitals and GP surgeries, which could have been prevented if staff had followed guidance issued a month earlier.

While there are strict legal protections ensuring the confidentiality of patients' records, under common law patients are "implied" to have consented to their information being shared if it was shared for the purpose of "direct care".

However, this basis was not valid in the arrangement between Royal Free and DeepMind in the view of Dame Fiona Caldicott, the National Data Guardian at the Department of Health, who has contributed to an investigation into the deal.</p>

This is going to get overlooked. But it shouldn't.
google  deepmind  health 
may 2017 by charlesarthur
DeepMind in talks with National Grid to reduce UK energy use by 10% • Ars Technica UK
Sebastian Anthony:
<p>it is the National Grid's job to balance supply and demand across the network, so that the AC frequency that arrives at your house is always within ±1% of 50Hz. Energy demands are usually quite predictable, in that they closely align with standard human behaviour (waking and sleeping hours) and the weather. Energy supply, however, is much less reliable, especially as the UK adds more wind and solar power to the mix.
While the UK has about 13 gigawatts of installed wind power capacity—the nation's average power draw is only about 35 gigawatts, incidentally—a lack of wind can cause major issues. Back in November 2015, the <a href="">last time we had a major power shortfall in the UK</a>, all those wind turbines only produced about 400 megawatts. (You should read that story if you want more information about how the National Grid works, and how it uses short-term reserves to balance supply and demand.)

Ingesting data, predicting trends, and suggesting solutions is almost perfectly suited to DeepMind's neural network expertise. While the National Grid is surely aware of some potential optimisations, a more rigorous investigation by a DeepMind AI may uncover solutions that the grid's human operators have never considered. One thing's for certain: a system as large as the UK grid has millions of inefficiencies. The biggest losses come from long-distance power transmission and voltage transformers, but it all adds up.</p>

DeepMind (and Google) claim happily that they reduced power usage in Google data centres by 40%. That's a lot. The National Grid, though, is a much more complex beast, and the challenge is variability. Maybe a system that can incorporate localised weather forecasts (wind and sun), plus industrial production, plus what's on TV.. maybe that will cope.

Also, how will Google be paid? Incentive? Percentage of energy saved (but how will that be determined)?
google  deepmind  electricity  nationalgrid 
march 2017 by charlesarthur
Google's DeepMind plans bitcoin-style health record tracking for hospitals • The Guardian
Alex Hern:
<p>DeepMind has been working in partnership with London’s Royal Free Hospital to develop kidney monitoring software called Streams and has <a href="">faced criticism</a> from patient groups for what they claim are overly broad data sharing agreements. Critics fear that the data sharing has the potential to give DeepMind, and thus Google, too much power over the NHS.

In a <a href="">blogpost</a>, DeepMind co-founder, Mustafa Suleyman, and head of security and transparency, Ben Laurie, use an example relating to the Royal Free Hospital partnership to explain how the system will work. “[An] entry will record the fact that a particular piece of data has been used, and also the reason why, for example, that blood test data was checked against the NHS national algorithm to detect possible acute kidney injury,” they write.

Suleyman says that development on the data audit proposal began long before the launch of Streams, when Laurie, the co-creator of the widely-used Apache server software, was hired by DeepMind. “This project has been brewing since before we started DeepMind Health,” he told the Guardian, “but it does add another layer of transparency.

“Our mission is absolutely central, and a core part of that is figuring out how we can do a better job of building trust. Transparency and better control of data is what will build trust in the long term.” </p>

I feel obliged to point out that adding layers inevitably makes things less, not more, transparent. The criticisms of DeepMind have broadly been shrugged off, and the NHS doesn't seem to have any incentive to engage with those critics. But whose data is it? And why does Google get it and not the NHS, since it's public money that enables this?
deepmind  data 
march 2017 by charlesarthur
We got 1.6 million students’ Google search histories! Now read on • Eerke Boiten's blog
<p>We got 1.6 million students’ Google search histories!
We have fantastic news. Google have given us 1.6 million UK university students’ five years’ Google search histories, so we can work on improving their learning. Knowing what students have looked for on Google in the last 5 years will allow us to model their metacognitive skills and learning styles very accurately, so we can make individualised interventions when their everyday Google searches show their potential misunderstanding of what we’re trying to teach them.

We’ve promised Google we won’t be using this information for anything else, honest. Our agreement with Google says so, though it also says that third parties (like the students themselves) can’t hold us to anything that’s agreed in there. It’s all very exciting. We have never received data from Google before, never done any learning analytics, heck, we haven’t even looked at the web tracking data of our university’s Moodle virtual learning environment. But we think improving university education is extremely important, and we are so smart and successful, that we decided we just needed to go ahead and do this.

<em>Hmmm. Maybe that doesn’t work too well. Let’s try again.</em></p>

At this point you realise that Boiten is in fact making a point about DeepMind/Google getting access to five years' medical histories for 1.6m people. And the point is made rather elegantly.
google  deepmind  health 
december 2016 by charlesarthur
WaveNet: a generative model for raw audio • DeepMind
<p>This post presents <a href="">WaveNet</a>, a deep generative model of raw audio waveforms. We show that WaveNets are able to generate speech which mimics any human voice and which sounds more natural than the best existing Text-to-Speech systems, reducing the gap with human performance by over 50%.

We also demonstrate that the same network can be used to synthesize other audio signals such as music, and present some striking samples of automatically generated piano pieces.</p>

I had been wondering how long it would be before DeepMind got to work on music. The stuff here is quite amazing. The DM-generated voice is really very impressive. And I'd like a playlist of the autogenerated music, please, for background music.
ai  audio  machinelearning  speech  deepmind  music 
september 2016 by charlesarthur
Google DeepMind AI to be in all of Google's data centres by end of 2016 • Business Insider
Sam Shead:
<p>Google has adopted a DeepMind AI system in several of its data centres over the last few months as a way of reducing the amount of energy the server farms consume. However, the full extent of Google's plans for the software are only just becoming clear.

"It [DeepMind's AI] will be in the entire fleet by the end of the year," DeepMind cofounder Mustafa Suleyman told Business Insider on Wednesday. "That will result in a 15% reduction on energy used every year by the entire data centre fleet."

…every time a user engages with one of [Google's[ services, a server is spun up and heat is produced that ultimately needs to be removed by an energy-consuming cooling system.

In order to lower the energy consumption of Google's cooling systems, DeepMind analysed five years worth of Google data centre records that have been collected by sensors measuring variables like temperature, compute load, air pressure, and fan speed.

"We used that to predict what the optimal settings are for controlling the cooling system, which can be thought of as a very complex air conditioning unit that tries to extract heat from the data centre."

After looking at the data, Google's self-learning algorithm was able to figure out the best times to use the cooling fans. "We can optimally turn up the fan when we need to and not waste energy on over cooling when we don’t need to," said Suleyman.</p>
google  deepmind  machinelearning  aircon 
july 2016 by charlesarthur
We need to talk about AI and access to publicly funded data-sets • TechCrunch
Natasha Lomas with a hugely important analysis:
<p>DeepMind says it will be publishing “results” of the Moorfields research [on eye disease] in academic literature. But it does not say it will be open sourcing any AI models it is able to train off of the publicly funded data.

Which means that data might well end up fueling the future profits of one of the world’s wealthiest technology companies. Instead of that value remaining in the hands of the public, whose data it is.

And not just that — early access to large amounts of valuable taxpayer-funded data could potentially lock in massive commercial advantage for Google in healthcare. Which is perhaps the single most important sector there is, given it affects everyone on the planet. If you don’t think Google has designed on becoming the world’s medic, why do you think it’s doing things like <a href="">this</a>?

Google will argue that the potential social benefits of algorithmically improved healthcare outcomes are worth this trade off of giving it advantageous access to the locked medicine cabinet where the really powerful data is kept.

But that detracts from the wider point: if valuable public data-sets can create really powerful benefits, shouldn’t that value remain in public hands?</p>

Yes. Exactly. This is a key point which is being ignored: data is the necessity for Google and the British government is not seeking sufficiently clear repayment for it.
google  deepmind  data  medical 
july 2016 by charlesarthur
Moorfields announces research partnership • Moorfields Eye Hospital NHS Foundation Trust
<p>Two million people are living with sight loss in the UK, of whom around 360,000 are registered as blind or partially sighted. At the moment, eye health professionals rely on digital scans of the eye to diagnose and determine the correct treatment for common eye conditions such as age-related macular degeneration and diabetic retinopathy.

These scans are highly complex and to date, traditional analysis tools have been unable to explore them fully. It also takes eye health professionals a long time to analyse eye scans, which can have an impact on how quickly they can meet patients to discuss diagnosis and treatment…

…Faster and more efficient diagnosis of eye disease could help prevent many thousands of cases of sight loss due to wet age-related macular degeneration and diabetic retinopathy, which together affect more than 625,000 people in the UK.

Moorfields Eye Hospital will share approximately one million anonymised digital eye scans, used by eye health professionals to detect and diagnose eye conditions. Anonymous clinical diagnoses, information on the treatment of eye diseases, model of the machine used to acquire the images and demographic information on age (shown to be associated with eye disease) is also being shared. This has been collected over time through routine care, which means it’s not possible to identify any individual patients from the scans. And they’re also historic scans, meaning that while the results of our research may be used to improve future care, they won’t affect the care any of our patients receive today.</p>

What we want machines to do: take over tedious routine which conceals important data. I'm meantime wondering: is machine learning already being used for airport X-ray scanning?
moorfields  deepmind  machinelearning 
july 2016 by charlesarthur
ICO probes Google DeepMind patient data-sharing deal with NHS Hospital Trust • Computer Weekly
Caroline Donnelly:
<p>The Information Commissioner's Office, the data protection watchdog, confirmed an investigation into the arrangement is underway, on the back of at least one complaint from the general public.

The deal gives DeepMind access to the healthcare records of 1.6 million patients that pass through three hospitals in North London, which fall under the care of the Royal Free Hospital Trust.

The complaint, seen by Computer Weekly, questions whether DeepMind will be expected to encrypt the patient data it receives when at rest.

“Whilst the information-sharing agreement insists that personally identifiable information – such as name, address, post code, NHS number, date of birth, telephone number, and email addresses, etc – must be encrypted whilst in transit to Google, it does not explicitly prohibit that data being unencrypted at the non-NHS location,” the complaint read.</p>

First there's a deal; then it turns out it's not directly approved. The complaint is essentially that individuals at Google/Deepmind might access personal data. This is the essential battleground of the coming years: how compatible is tight data regulation with data mining?
google  deepmind  health  nhs 
may 2016 by charlesarthur
Revealed: Google AI has access to huge haul of NHS patient data | New Scientist
Hal Hodson:
<p>It’s no secret that Google has broad ambitions in healthcare. But a document obtained by New Scientist reveals that the tech giant’s collaboration with the UK’s National Health Service goes far beyond what has been publicly announced.

The document – a data-sharing agreement between Google-owned artificial intelligence company DeepMind and the Royal Free NHS Trust – gives the <a href="">clearest picture yet of what the company is doing and what sensitive data it now has access to</a>.

The agreement gives DeepMind access to a wide range of healthcare data on the 1.6 million patients who pass through three London hospitals run by the Royal Free NHS Trust – Barnet, Chase Farm and the Royal Free – each year. This will include information about people who are HIV-positive, for instance, as well as details of drug overdoses and abortions. The agreement also includes access to patient data from the last five years…

…This is the first we’ve heard of DeepMind getting access to historical medical records, says Sam Smith, who runs health data privacy group MedConfidential. “This is not just about kidney function. They’re getting the full data.”

The agreement clearly states that Google cannot use the data in any other part of its business. The data itself will be stored in the UK by a third party contracted by Google, not in DeepMind’s offices. DeepMind is also obliged to delete its copy of the data when the agreement expires at the end of September 2017.</p>

From the document: "Data to be processed other than for the direct care of the patient <strong>must be pseudonymised</strong> in line with the NHS Act 2006". (Emphasis in original.)
ai  data  google  deepmind  health 
may 2016 by charlesarthur
AI's biggest mystery is the ethics board Google set up after buying DeepMind » Business Insider
Sam Shead:
<p>DeepMind CEO and cofounder Demis Hassabis has confirmed at a number of conferences that Google's AI ethics board exists. But neither Hassabis nor Google have ever disclosed the individuals on the board or gone into any great detail on what the board does.

Azeem Azhar, a tech entrepreneur, startup advisor, and author of the Exponential View newsletter, told Business Insider: "It’s super important [to talk about ethics in AI]. "

Media and academics have called on DeepMind and Google to reveal who sits on Google's AI ethics board so the debate about where the technology they're developing can be carried out in the open, but so far Google and DeepMind's cofounders have refused.

It's generally accepted that Google's AI ethics board can only be a good thing but <a href="">ethicists like Evan Selinger</a>, a professor of philosophy at the Institute for Ethics and Emerging Technologies, have questioned whether Google should be more transparent about who is on the board and what they're doing. </p>
ai  google  deepmind 
march 2016 by charlesarthur
Google DeepMind: What is it, how it works and should you be scared? » Techworld
Sam Shead interview with Mustafa Suleyman, co-founder of Deepmind, who explains where the systems are used inside Google:
<p>We use it to identify text on shopfronts and maybe alert people to a discount that’s available in a particular shop or what the menu says in a given restaurant. We do that with an extremely high level of accuracy today. It’s being used in Local Search and elsewhere across the company.  

We also use the same core system across Google for speech recognition. It trains roughly in less than 5 days. In 2012 it delivered a 30 percent reduction in error rate against the existing old school system. This was the biggest single improvement in speech recognition in 20 years, again using the same very general deep learning system across all of these. 

Across Google we use what we call Tool AI or Deep Learning Networks for fraud detection, spam detection, hand writing recognition, image search, speech recognition, Street View detection, translation. 

Sixty handcrafted rule-based systems have now been replaced with deep learning based networks. This gives you a sense of the kind of generality, flexibility and adaptiveness of the kind of advances that have been made across the field and why Google was interested in DeepMind. </p>
ai  deepmind  google 
march 2016 by charlesarthur
DeepMind founder Demis Hassabis on how AI will shape the future » The Verge
Sam Byford, in a terrific wide-ranging, intelligent interview:
<p>SB: So let’s move onto smartphone assistants. I saw you put up a slide from Her in your presentation on the opening day — is that really the endgame here?

DH: No, I mean Her is just an easy popular mainstream view of what that sort of thing is. I just think we would like these smartphone assistant things to actually be smart and contextual and have a deeper understanding of what you’re trying to do. At the moment most of these systems are extremely brittle — once you go off the templates that have been pre-programmed then they’re pretty useless. So it’s about making that actually adaptable and flexible and more robust.

SB: What’s the breakthrough that’s needed to improve these? Why couldn’t we work on it tomorrow?

DH: Well, we can — I just think you need a different approach. Again, it’s this dichotomy between pre-programmed and learnt. At the moment pretty much all smartphone assistants are special-cased and pre-programmed and that means they’re brittle because they can only do the things they were pre-programmed for. And the real world’s very messy and complicated and users do all sorts of unpredictable things that you can’t know ahead of time. Our belief at DeepMind, certainly this was the founding principle, is that the only way to do intelligence is to do learning from the ground up and be general.</p>

This is a must-read; Hassabis is thinking so far ahead, but also so clearly. (I've previously said that I think the AI capabilities of phones <a href="">will feed into the next pervasive thing - a bit like the selfie</a>.)
ai  deepmind  hassabis 
march 2016 by charlesarthur
AlphaGo defeats Lee Sedol in first game of historic man vs machine match » Go Game Guru
You probably heard the news that Google's DeepMind system AlphaGo beat the best player in the world. Here are the reactions of the pros ("9p" means "9-dan professional", ie the highest level):
<p>Lee Changho 9p said,  “I’m so shocked by AlphaGo’s play!”

Meanwhile Cho Hanseung 9p remarked, “AlphaGo is much stronger than before, when it played against Fan Hui 2p! When Google said the odds were fifty-fifty, it seems they weren’t joking. I still can’t believe its performance even though I just saw it with my own eyes.”

In a post-game interview, Lee Sedol was visibly startled by AlphaGo’s strength. “I was so surprised. Actually, I never imagined that I would lose. It’s so shocking. Regarding the game, I got off to a bad start and AlphaGo played well right until the end. Even when I was behind, I still didn’t imagine that I’d lose. I didn’t think that it would be able to play such an excellent game. I heard that the DeepMind CEO Demis Hassabis said that he respects me as a Go player, but I have great respect for both of them [referring to Demis Hassabis and Eric Schmidt] for making this amazing program. I also respect all the programmers who helped to make AlphaGo.”</p>

The second game should finish around 0800 GMT on Thursday.
go  deepmind 
march 2016 by charlesarthur
Health » Google DeepMind
<p>We have been collaborating with some of the UK’s leading kidney experts at the Royal Free Hospital London to co-design and pilot a mobile app, called Streams, which presents timely information that helps nurses and doctors detect cases of acute kidney injury. AKI is a contributing factor in up to 20% of emergency hospital admissions as well as 40,000 deaths in the UK every year. Yet NHS England estimate that around 25% of cases are preventable.

Consultant Nephrologist and Associate Medical Director for patient safety at the Royal Free Hospital London, Dr Chris Laing, who helped design the app and oversaw two initial pilots at the Royal Free, said:

“Using Streams meant I was able to review blood tests for patients at risk of AKI within seconds of them becoming available. I intervened earlier and was able to improve the care of over half the patients Streams identified in our pilot studies.”</p>

Really interesting. AI is a sort of "rising tide" technology: you don't notice it changing your world, until the world is utterly changed and you find yourself thinking "however did we cope before?"
ai  health  deepmind 
february 2016 by charlesarthur
The superhero of artificial intelligence: can this genius keep it in check? » The Guardian
Clemency Burton-Hill on DeepMind's Demis Hassabis. The interview with him is OK - though mostly dead-bat responses from him - but I thought this was more indicative of the challenge, and potential for the company:
<p>Upstairs, wrapping the original building, is a modern open-plan structure featuring a deck with undeniably magnificent views of London’s rooftops.

It’s up here, on Friday nights, that the DeepMinders gather for drinks. One employee describes the ritual to me enthusiastically as a way “to end the week on a high”. Socialising is an intrinsic way of life: I’m told of the DeepMind running club, football team, board games club. (“That one gets pretty competitive.”) A wall chart with moveable photographs indicates where everyone is hot-desking on any given day. It’s aggressively open-plan. The engineers – mostly male – that I pass in the corridors shatter the stereotype of people working in the nerdier corners of human endeavour: these guys look fit, happy, cool. A certain air of intellectual glamour, it has to be said, vibrates in the atmosphere. And no wonder. The smartest people on the planet are queuing up to work here, and the retention rate is, so far, a remarkable 100%, despite the accelerating focus on AI among many of Google’s biggest competitors, not to mention leading universities all over the globe.

“We’re really lucky,” says Hassabis, who compares his company to the Apollo programme and Manhattan Project for both the breathtaking scale of its ambition and the quality of the minds he is assembling at an ever increasing rate. “We are able to literally get the best scientists from each country each year. So we’ll have, say, the person that won the Physics Olympiad in Poland, the person who got the top maths PhD of the year in France. We’ve got more ideas than we’ve got researchers, but at the same time, there are more great people coming to our door than we can take on. So we’re in a very fortunate position. The only limitation is how many people we can absorb without damaging the culture.”</p>
deepmind  google  ai 
february 2016 by charlesarthur
AlphaGo vs. Fan Hui 2p » GoKibitz
The final position of one of the winning games by DeepMind's AlphaGo against European Go champion Fan Hui. You can replay it; if you don't understand Go, don't worry. But it's a game where an average player (such as me) can't tell the difference between the human and the computer.
go  deepmind 
january 2016 by charlesarthur

Copy this bookmark:

to read