Daily Archives: March 2, 2017

Artificial intelligence boosters predict a brave new world of flying cars and cancer cures. Detractors worry about a future where humans are enslaved to an evil race of robot overlords. Veteran AI scientist Eric Horvitz and Doomsday Clock guru Lawrence Krauss, seeking a middle ground, gathered a group of experts in the Arizona desert to discuss the worst that could possibly happen -- and how to stop it.

Their workshop took place last weekend at Arizona State University with funding from Tesla Inc. co-founder Elon Musk and Skype co-founder Jaan Tallinn.Officially dubbed "Envisioning and Addressing Adverse AI Outcomes,"it was a kind of AI doomsday games that organized some 40 scientists, cyber-security experts and policy wonks into groups of attackers -- the red team -- and defenders -- blue team -- playing out AI-gone-very-wrong scenarios, ranging from stock-market manipulation to global warfare.

Horvitz is optimistic -- a good thing because machine intelligence is his life's work -- but some other, more dystopian-minded backers of the project seemed to find his outlook too positive when plans for this event started about two years ago, said Krauss, a theoretical physicist who directs ASU's Origins Project, the program running the workshop. Yet Horvitz said that for these technologies to move forward successfully and to earn broad public confidence, all concerns must be fully aired and addressed.

"There is huge potential for AI to transform so many aspects of our society in so many ways. At the same time, there are rough edges and potential downsides, like any technology," said Horvitz, managing director of Microsoft's Research Lab in Redmond, Washington. ``To maximally gain from the upside we also have to think through possible outcomes in more detail than we have before and think about how wed deal with them."

Participants were given "homework"to submit entries for worst-case scenarios. They had to be realistic -- based on current technologies or those that appear possible -- and five to 25 years in the future. The entrants with the "winning" nightmares were chosen to lead the panels, which featured about four experts on each of the two teams to discuss the attack and how to prevent it.

Blue team, including Launchbury, Fisher and Krauss, in the War and Peace scenario

Tessa Eztioni, Origins Project at ASU

Turns outmany of these researchers can match science-fiction writers Arthur C. Clarke and Philip K. Dick for dystopian visions. In many cases, little imagination was required -- scenarios like technologybeing used to sway electionsor new cyber attacks using AI are being seen in the real world,or are at least technically possible. Horvitz cited research that shows how to alter the way a self-driving car sees traffic signs so that the vehicle misreads a "stop" sign as "yield.''

The possibility of intelligent, automated cyber attacks is the one that most worries John Launchbury, who directs one of the offices at the U.S.'s Defense Advanced Research Projects Agency, and Kathleen Fisher, chairwoman of the computer science department at Tufts University, who led that session. What happens if someone constructs a cyber weapon designed to hide itself and evade all attempts to dismantle it? Now imagine it spreads beyond its intended target to the broader internet. Think Stuxnet, the computer virus created to attack the Iranian nuclear program that got out in the wild, but stealthier and more autonomous.

"We're talking about malware on steroids that is AI-enabled," said Fisher, who is an expert in programming languages.Fisher presented her scenario under a slide bearing the words "What could possibly go wrong?" which could have also served as a tagline for the whole event.

How did the defending blue team fare on that one? Not well, said Launchbury. They argued that advanced AI needed for an attack would require a lot of computing power and communication, so it would be easier to detect. But the red team felt that it would be easy to hide behind innocuous activities, Fisher said. For example, attackers could get innocent users to play an addictive video game to cover up their work.

Exclusive insights on technology around the world.

Get Fully Charged, from Bloomberg Technology.

To prevent a stock-market manipulation scenario dreamed up by University of Michigan computer science professor Michael Wellman, blue team members suggested treating attackers like malware by trying to recognize them via a database on known types of hacks. Wellman, who has been in AI for more than 30 years and calls himself an old-timer on the subject, said that approach could be useful in finance.

Beyond actual solutions, organizers hope the doomsday workshop started conversations on what needs to happen, raised awareness and combined ideas from different disciplines. The Origins Project plans to make public materials from the closed-door sessions and may design further workshops around a specific scenario or two, Krauss said.

DARPA's Launchbury hopes the presence of policy figures among the participants will foster concrete steps, like agreements on rules of engagement for cyber war, automated weapons and robot troops.

Krauss, chairman of the board of sponsors of the group behind the Doomsday Clock, a symbolic measure of how close we are to global catastrophe, said some of what he saw at the workshop "informed" his thinking on whether the clock ought to shift even closer to midnight. But don't go stocking up on canned food and moving into a bunker in the wilderness just yet.

"Some things we think of as cataclysmicmay turn out to be just fine," he said.

Go here to read the rest:

AI Scientists Gather to Plot Doomsday Scenarios (and Solutions ... - Bloomberg

Artificial intelligence AI cant solve the Southern rail dispute, but it can help make services run more smoothly. Photograph: Yui Mok/PA

Artificial intelligence (AI) is likely to prove the most transformative technology of the 21st century. Those of us who work in the field whether in the public or private sector are at a frontier that is advancing at an ever-accelerating rate. Yet my work on tech policy at the Government Digital Service and the Home Office often left me in despair. At a time when the possibilities created by AI are multiplying rapidly, the government isnt really at the races.

The Governments Digital Strategy, published yesterday, and the governments Transformation Strategy, published a couple of weeks ago, are a case in point. It is fantastic that some more money is going into AI and robotics research in our universities, but treating AI as one for the future misses the opportunities of today.

At our own business, ASI, we work with organisations that are achieving radical improvements in efficiency from relatively simple applications of AI. A payments company that increases fraud detection by 93%. An airline that uses machine learning to predict demand for staff in real time, allowing them to cut the number of standby staff required by 33%. A train manufacturer that uses a predictive maintenance model to reduce the number of inspections an engineer needs to perform to find a fault in need of repair from 10,000 to two.

The opportunities are here and now. But the projects that could improve our public services and deliver value for money to the taxpayer were nowhere to be seen in the digital strategy. And government remains embarrassingly short of examples it can point to. In fact, at a conference on government data last week, the chief executive of the Civil Service resorted to praising a list of public toilets released as open data. We can do better than this.

The stakes are high. Even after seven years of austerity, the public sector spends more than 40% of GDP. Yet the services that we rely on are under ever greater pressure. The only way the government can continue to meet the expectations that people have of the NHS, transport or prisons is to find ways to radically improve efficiency.

The good news is that it is easy to imagine ways in which these services could benefit from AI with relatively little investment. It is encouraging that the justice secretary, Liz Truss, has made digital technology so central to her prisons and courts bill. Machine learning could play a big part in this. For example, Harvard researchers found that cell-sharing configurations can reduce reoffending rates by about 15% for drugs and theft offences in French prisons. It stands to reason that choices of cellmates matter, but even very experienced prison officers find it difficult to balance the bewildering array of factors that need to be taken into account. In contrast to humans, machine learning thrives in finding the patterns that matter in this kind of complexity. This could be done right now.

Weve all read about supercomputers that are able to read a million medical journals an hour and spot tumours more accurately than experienced doctors. But there are significant wins to be had from the much more prosaic matter of allocating resources in hospitals more efficiently. Hospitals are complex organisations dealing with unpredictable demands. Machine learning can help them run more smoothly. Recent trials modelled how long particular consultations and operations were likely to take and booked theatre resources accordingly. This hugely increased the utilisation rates of these valuable resources, and reduced the number of over-runs caused by the fixed-time slots.

Transport is another area that could hugely benefit. AI cant solve the Southern rail dispute, but it can help make services run more smoothly. A recent project by ASI built an adaptive scheduling system for a bus operator that modelled the complex ways in which traffic flows through a city. In just a few weeks this was able to make buses 38% more likely to show up at the right time. Cue happier passengers, less crowded busses, and big savings for the bus company.

These are just three easy examples that could be implemented today. There are dozens of others across the entire public sector. But to help kickstart this kind of revolution it is vital that ministers, civil servants and frontline professionals become more familiar with what is possible. To achieve this, government should create a 20m fund for officials to bid into for projects that could demonstrate the value of AI.

Another thing the government could do to move the needle is to provide much better access to the data that is used to train these predictive models. Data.gov.uk has become a dumping ground for nugatory and obscure data sets. Why not require each public body to publish details describing its top 20 data sets that it uses for its own operations? That might help to ferment a proper debate about the new applications that the public might benefit from.

In the next two decades, AI will transform the way we live and work. There is no reason whatsoever why the government shouldnt be doing this too, but it is not. Adopting this technology is the most plausible way of delivering the public services people expect while making the savings we need.

Read the original:

Message to ministers: AI can transform the way we live right now - The Guardian

Artificial intelligence startup Ozlo thinks it has a solution for situations where virtual assistants fail in their responses: Getting the bots to ask questions back.

Ozlo is launching a trio of software packages for other companies to enhance the virtual assistants they build. Theyre aimed at making those assistants more sophisticated, including getting them to ask clarifying questions when they dont understand a user request.

Current virtual assistants, meaning conversational apps and bots like Apples Siri and Amazons Alexa, have this problem with being very brittle, Ozlo CEO Charles Jolley told Recode.

Hes referring to those moments when Siri says, I didnt quite get that, or where Google Assistant says, Sorry, I dont know how to do that yet, without addressing what part of the question the virtual assistants dont understand.

Its a problem Ozlo, with $14 million in funding from Greylock Partners and AME Cloud Ventures, set out to solve with its own mobile app released for public download on iOS last October and later made available on Android. The startup wouldnt share download data, but according to the Google Play store, it has only 100 Android downloads.

Jolley, who previously ran Facebook for Android, said the 30-person teams consumer app will continue to be offered but is really meant to test our service [the new products] in the real world. Opening the software up to companies has been part of the companys plan from the beginning, he said.

The three tools Ozlo is releasing include software for data analysis, interpretation of what a user intended to say, and for conversing with users. The last tool is supposed to help systems determine when to ask clarifying questions in response to a user request.

Ozlo is not alone in offering tools for companies to develop or improve upon virtual assistants. Google, IBM, Amazon and a smattering of startups also make tools to assist companies in building their own bots and enhance their software products.

An Ozlo rep said the company has three major customers signed on to use the services, a top consumer internet company, a top media organization and a top mobile app. All are names you would recognize, with products you likely use every day. Jolley said these customers are already building virtual assistants.

While the new products are being sold to companies, Jolley thinks consumers may be able to detect Ozlos use based on changes in how assistants work.

I think probably the most surprising thing will be when you ask your assistant something ambiguous and it asks you something back, he said.

Read the original:

This AI startup wants to help robot assistants ask people the questions - Recode

Visit just about any tennis club on a Saturday, and youre likely to witness otherwise sensible adults losing their minds over line calls. Players suffer complete meltdowns as they hurl insults. Parents morph into brooding teenagers. Friends become enemies. Postmatch beers can undo some of the damage, but the shame and resentment linger for days.

More civilized times may lie ahead. French inventor Grgoire Gentil has designed a $200 GoPro-size device that can be fastened to any net post and detect whether balls are in or out with surprising accuracy. Its called, reasonably enough, the In/Out. I was born in Paris and raised on clay, Gentil says. On clay, the ball leaves a mark, and he recalls many arguments over a blemish on the court. It was the starting point of this, I would say.

Gentil, 44, now lives in Palo Alto and built the In/Out in his living room lab. The device monitors both sides of a tennis court using a pair of cameras similar to those found in smartphones. After attaching the In/Out to the net with a plastic strap, a player pushes a button on its screen, and it scans the court to find the lines using open-source artificial intelligence software. AI also helps the device track the balls flight, pace, and spin. This would not have been possible five years ago, Gentil says.

The In/Outs dual cameras map the lines of a tennis court, and the device beeps to signal missed shots.

Source: In/Out

In a test at Stanford, Gentil and I played for an hour, and the In/Out beeped whenever one of his shots sailed long or wide. (I dont remember missing any.) On close calls, we rushed over to watch a video replay on the In/Out screen. At hours end, Gentil whipped out a tablet and connected to the In/Out app, which showed where all our shots had landed and provided some other stats.

Although equipment like the In/Out has been around for years, Gentils is the only one that costs about as little as a decent racket. Top tournaments, including the Grand Slams, use Hawk-Eye, a Sony Corp.-owned system of superaccurate cameras that customers say costs $60,000 or more to set up on each court. Given the price, its typically reserved for show courts. Sony didnt respond to requests for comment.

PlaySight Interactive Ltd., a startup in Israel, makes a six-camera system thats less accurate than Hawk-Eye but costs a mere $10,000 per court, plus a monthly fee to collect data that can be reviewed online or in an app. PlaySights setup also includes a large screen that lets players see line calls and ball speed without interrupting the game. The company has sold its gear mostly to tennis clubs and universities.

Its screen can show video replays.

Source: In/Out

Chris Edwards heads the product testing work done by retailer Tennis Warehouse and has tried all three tracking systems. The In/Out doesnt bring the same depth of insight as PlaySight, he says. But as far as a portable, cheap device goes, the In/Out has the potential to be the best by far. I havent seen anything else like this.

Over the past decade, Gentil has made a dozen products. He sold a software company to Cisco Systems Inc., designed an augmented-reality motorcycle helmet, and built a hand-size drone that can follow a person around. He spent two years developing the In/Out, tuning the software, even 3D-printing a plastic tennis ball-shaped case for it. Its been a tumultuous process, Gentil says. You get an algorithm working on the tennis court one day and think you will sell hundreds of thousands of units, and the day after, nothing is working.

The most important business stories of the day.

Get Bloomberg's daily newsletter.

Gentil acknowledges his machines limits. The In/Out has a 20-millimeter margin of error, compared with about 3mm for the Hawk-Eye, and can get confused during doubles matches if the extra players block its line of sight. Gentil says he hopes to improve the devices accuracy and recommends that two In/Outs be used for doubles. As for the possibility that Sony or PlaySight might sue him over the concept of his invention, hes filed some patents himself, he says. If Hawk-Eye is coming after me tomorrow morning, they are going against innovation and against the tennis community. I think I might have the tennis community with me.

The bottom line: Gentils $200 line-calling AI isnt as accurate as rival products, but unlike them, its affordable enough for mass adoption.

See original here:

This $200 AI Will End Tennis Club Screaming Matches - Bloomberg

Share

Share

Share

Share

Email

Line Corp.,owner of Japans most popular messaging service, is getting into the artificial intelligence market in a big way by outlining an ambitious plan that pits it against the likes of Google, Facebook and Amazon.

According to a report inBloomberg News, Line Corp. is gearing up to launch a suite of AI softwaretools that will enable a digital assistant thatspeaks in Japanese and Korean. The assistant will be able to converse with users and provide weather and news via a dedicated smartphone app or a speaker that sits on the table and is called Wave, similar to Amazons Echo.

Line Corp., which unveiled the strategy during the Mobile World Congress in Barcelona, Spain, this week, said both the app and the speaker will come to the market between April and June. While Line faces a lot of competition, the companythinks it can stand out from the pack because of its local knowledge about the markets in which it is operating, including South Korea, Taiwan, Thailand and Indonesia.

There is a shift toward toward post-smartphone, post-touch technologies, Chief Executive Officer Takeshi Idezawa said in an interview with Bloomberg. These connected devices will permeate even deeper into our daily lives and therefore must even closer match the local needs, languages and cultures.

According to the report,Lines AI software platform was developed with its parent company Naver Corp., which operates a search engine. While Line is mainly a messaging app, customers use it to read the news, get a taxi ride and find part-time work. All of that content and interaction in local languages provides Line with an edge over larger rivals, noted the report, with Idezawa arguing that the AI experience is only as good as the data its trained on.

Share

Share

Share

Share

Email

See the article here:

Japan's Line Corp. To Launch AI App, Speaker - PYMNTS.com

One billion people use Facebook Messenger every month. And no matter how bad the current perceptions of the bot scene are, that number is hard to ignore. Octane AIis counting on celebrity content creators to build conversational experiences that people actually want to have. With its public launch and the rollout of its content creation platformConvos, the Octane team is taking a gamble on the medium. If it catches on, the company will have a community-driven head start that will make anyquips about the lack of sophistication in the typicalbottech stack inconsequential.

The Convo platform lets anyone create tree-like stories that otherscan engage with.Building the conversations is as easy ashaving a slightly delusional conversation with yourself the barrier being creativity rather than the underlying mechanics.

Once published, users have the power to makedecisions thatchange narrativeswithin a givenchat. This happens via embedded bubbles, almost like a multiple choice question for each response. The pre-defined structure takes variability out of the experience and removes the need for natural language processing.

There aremultiple categories of bots that willall do well, explained Ben Parr, co-founder of Octane AI.But we dont see anyone doing a great job withcontent.

Octanes private beta began back in November. Through a series of pilots, including one with Maroon 5, the team noticed abnormally high conversionrates. Though Octane wasnt able to provide specifics, the takeaway here is that bots created on the companys platform might be able to drive more traffic with a smaller dedicated following than something like a traditional Facebook Page.

Were not heavy on machine learning yet, but personalization ofConvos could be huge, saidParr.

Parr, who waspreviously Editor-At-Large at Mashable, is joined byLeif K-Brooks, the former founder of OmegleandMatt Schlicht, founder ofChatbots Magazine, to make up the founding team of Octane AI. Schlicht received criticism back in November for failing to disclose Octanes ambitions when soliciting pitch desks from the bot community to pass on to investors. Parr believes that, The bot community has gotten even stronger in the time sinceSchlichts apology.The company raised $1.5 million in a November 2016 seed round led byGeneral Catalyst.

View post:

Octane AI boldly bets that Convos are the future of content - TechCrunch

It is no mystery why poker is such a popular pastime: the dynamic card game produces drama in spades as players are locked in a complicated tango of acting and reacting that becomes increasingly tense with each escalating bet. The same elements that make poker so entertaining have also created a complex problem for artificial intelligence (AI). A study published today in Science describes an AI system called DeepStack that recently defeated professional human players in heads-up, no-limit Texas holdem poker, an achievement that represents a leap forward in the types of problems AI systems can solve.

DeepStack, developed by researchers at the University of Alberta, relies on the use of artificial neural networks that researchers trained ahead of time to develop poker intuition. During play, DeepStack uses its poker smarts to break down a complicated game into smaller, more manageable pieces that it can then work through on the fly. Using this strategy allowed it to defeat its human opponents.

For decades scientists developing artificial intelligence have used games to test the capabilities of their systems and benchmark their progress. Twenty years ago game-playing AI had a breakthrough when IBMs chess-playing supercomputer Deep Blue defeated World Chess Champion Garry Kasparov. Last year Google DeepMinds AlphaGo program shocked the world when it beat top human pros in the game of go. Yet there is a fundamental difference between games such as chess and go and those like poker in the amount of information available to players. Games of chess and go are perfect information games, [where] you get to see everything you need right in front of you to make your decision, says Murray Campbell, a computer scientist at IBM who was on the Deep Blue team but not involved in the new study. In poker and other imperfect-information games, theres hidden informationprivate information that only one player knows, and that makes the games much, much harder.

Artificial intelligence researchers have been working on poker for a long timein fact, AI programs from all over the world have squared off against humans in poker tournaments, including the Annual Computer Poker Competition, now in its 10th year. Heads-up, no-limit Texas holdem presents a particularly daunting AI challenge: As with all imperfect-information games, it requires a system to make decisions without having key information. Yet it is also a two-person version of poker with no limit on bet size, resulting in a massive number of possible game scenarios (roughly 10160, on par with the 10170 possible moves in go). Until now poker-playing AIs have attempted to compute how to play in every possible situation before the game begins. For really complex games like heads-up, no-limit, they have relied on a strategy called abstraction in which different scenarios are lumped together and treated the same way. (For example, a system might not differentiate between aces and kings.) Abstraction simplifies the game, but it also leaves holes that opponents can find and exploit.

With DeepStack, study author Michael Bowling, a professor of machine learning, games and robotics, and colleagues took a different approach, adapting the AI strategies used for perfect-information games like go to the unique challenges of heads-up, no-limit. Before ever playing a real game DeepStack went through an intensive training period involving deep learning (a type of machine learning that uses algorithms to model higher-level concepts) in which it played millions of randomly generated poker scenarios against itself and calculated how beneficial each was. The answers allowed DeepStacks neural networks (complex networks of computations that can learn over time) to develop general poker intuition that it could apply even in situations it had never encountered before. Then, DeepStack, which runs on a gaming laptop, played actual online poker games against 11 human players. (Each player completed 3,000 matches over a four-week period.)

DeepStack used its neural network to break up each game into smaller piecesat a given time, it was only thinking between two and 10 steps ahead. The AI solved each mini game on the fly, working through millions of possible scenarios in about three seconds and using the outcomes to choose the best move. In some sense this is probably a lot closer to what humans do, Bowling says. Humans certainly dont, before they sit down and play, precompute how theyre going to play in every situation. And at the same time, humans cant reason through all the ways the poker game would play out all the way to the end. DeepStack beat all 11 professional players, 10 of them by statistically significant margins.

Campbell was impressed by DeepStacks results. They're showing what appears to be a quite a general approach [for] dealing with these imperfect-information games, he says, and demonstrating them in a pretty spectacular way. In his view DeepStack is an important step in AI toward tackling messy, real-world problems such as designing security systems or performing negotiations. He adds, however, that even an imperfect-info game like poker is still much simpler than the real world, where conditions are continuously changing and our goals are not always clear.

DeepStack is not the only AI system that has enjoyed recent poker success. In January a system called Libratus, developed by a team at Carnegie Mellon University, beat four professional poker players (the results have not been published in a scientific journal). Unlike DeepStack, Libratus does not employ neural networks. Instead, the program, which runs off a supercomputer, relies on a sophisticated abstraction technique early in the game and shifts to an on-the-fly reasoning strategy similar to that used by DeepStack in the games later stages. Campbell, who is familiar with both technologies, says it is not clear which is superior, pointing out that whereas Libratus played more elite professionals, DeepStack won by larger margins. Michael Wellman, a computer scientist at the University of Michigan who was also not involved in the work, considers both successes significant milestone[s] in game computation.

Bowling sees many possible directions for future AI research, some related to poker (such as systems that can compete in six-player tournaments) and others that extend beyond it. I think the interesting problems start to move into what happens if were playing a game where we dont even know the rules, he says. We often have to make decisions where were not exactly sure how things actually work, he adds, which will involve building agents that can cope with that and learn to play those games, getting better as they interact with the world.

Link:

Time to Fold, Humans: Poker-Playing AI Beats Pros at Texas Hold'em - Scientific American