Monthly Archives: August 2017

OpenAI used the action at this years Dota 2 championships as an opportunity to show off its work by having top players lose repeatedly to its in-game bot.

Dotas normally a team game with a heavy emphasis on coordination and communication, but for players interested in beefing up their pure, technical ability, the game also has a 1v1 mode. Thats what tech company OpenAI used to show off its programming of a bot against one of the games most famous and beloved players, Danil Dendi Ishutin.

That mode has both players compete in the games mid-lane, with only the destruction of that first tower or two enemy kills earning either side a win. In addition, for purposes of this particular demonstration, specific items like Bottle and Soul Ring, which help players manage health and mana regeneration, were also restricted. Dendi decided to play Shadow Fiend, a strong but fragile hero who excels at aggressive plays, and to make it a mirror match the OpenAI bot did the same.

Rarely do you hear a crowd of people cheering over creep blocking, but thats what the fans in Key Arena did last night while watching the exhibition match. The earliest advantage in a 1v1 Dota face-off comes with one side slowing down their support wave of AI creeps enough to force the opponent farther into enemy territory and thats exactly what the bot managed to do within the first thirty seconds of the bout.

After that, things seemed to even out but Dendi, lacking a good read on his AI rival, played cautiously and ended up losing out on experience and gold as the bot was given space to land more last-hits. By three minutes in, OpenAI had already harassed Dendis tower and gained double the CS. The former TI winner suffered his first death as a result shortly after. At that point, with the AI unlikely to make a crucial mistake and Dendi falling further and further behind in experience points, the game match was all but over. The pro tried to change things around with last ditch attempt at a kill but he ended up sacrificing his own life to do it.

In a rematch, Dendi admitted that he was going to try and mimic the AIs strategy of pushing his lane early, explaining how the dynamic of a 1v1 fight in Dota is counter-intuitive since it relies on purely outplaying your opponent rather than trying to out think them. Switching sides from Radiant to Dire for game two, Dendi got off to an even worse. He and the opposing AI exchanged blows early, and within the first two minutes he as forced to retreat only to die along the way.

The OpenAI bot was trained, accroding to company CTO Greg Brockman, by playing many lifetimes worth of matches and only limited coaching along the way. Earlier in the week it had defeated other pros renowned for their technical play, including SumaiL and Arteezy, learning each time and improving itself. But these matches were more to test how far the bot had come than anything else. Self-playing was what got it to that point, with Brockman explaining in a blog post that the AIs learning style requires playing against opponents very close in skill level so it can make incremental adjustments to improve over time.

The company, funded in part by Elon Musk, is working on a number of different AI projects, including impersonating Reddit commenters, but games have always been an important part of designing and testing computer learning. From checkers and chess to StarCraft and now Dota, the well defined rule systems and clear win conditions are a natural fit.

And the 1v1 mode of Valves MOBA takes that logic even further, offering a way of limiting the number of variables operating in the form of other players. Rather than worry about what nine other people are doing and exponentially increasing the number of options and possibilities the AI has to contend with, 1v1 allows it to focus the games core elements, similar a beginner chess player practicing openings. The OpenAI teams ambitions dont stop there, however. The bots designers hope to see it perform in full-fledged 5v5 matches by next year.

You can watch the entire demo below.

Go here to read the rest:

An AI Is Beating Some Of The Best Dota Players In The World - Kotaku

Regulations, sanctions, rules they are not always pure evil as some might suggest. The regulations about how to keep a commuter rail safe or the sanctions the U.S. government uses to manage relations with foreign countries are necessary, not evil.

Yet, when it comes to AI, do we really need to worry?

Elon Musk has gone on the offensive attempting to convince us that AI needs to be more regulated because it could spin out of control. He tweeted that the dangers we face from AI and machine learning has vastly more risk than North Korea. He followed up that tweet by saying that everything thats a danger to the public is regulated, including cars and planes.

The problem with this line of thinking, of course, is that an AI is a piece of software. A plane weighs over 350,000 pounds and can fall out of the sky. Where are we in the continuum of machines taking over? In an infant stage not even crawling or walking. We might want to avoid hysterics.

Still, some of the reactions have been quite interesting.

One user said it was inappropriate to compare a nuclear threat to AI. One said the real danger is humans creating AI that doesnt work. Another pointed out the obvious if there is a nuclear war, it might not matter if the machines take over. Well all be dead.

The problem with the end of the world thanks to AI discussion is that we never get into specifics. Its a random tweet comparing machine intelligence to nuclear war. Its another random tweet talking about regulation. But what kind of AI should be regulated? By whom and where? What are the actual dangers? The problem with fear-mongering about AI is that there are no obvious examples of a machine actually causing mass destructionyet. We hear about failed automations, of cars driving themselves off the road, of a chatbot app crashing.

Musk has noted before that we should regulate now before it gets out of hand. Again, he hasnt explained what should be regulated Microsoft Word? Chatbots? The subroutines in a home sensor that shuts off your sprinkler system? Satellites? Autonomous trucks? Lets get the subject out in the open and get into the specifics of regulation and see where that takes us, because my guess is that the companies making chatbots dont need to be regulated as much as they need to be told to make better and more useful bots with the funding they already have.

Or is this all about the laws of robotics? If thats the case, we get into a brand new problem what is a robot? Im sure Isaac Asimov never predicted that there would be a catbot that tells us the weather forecast (if he did, I apologize to all science fiction fans everywhere). Lets regulate the catbots before they get out of hand, right? Next up the dogbots.

The issue is pretty clear: When you start talking about specific regulations and dangers, they become a bit laughable. What are we really asking Congress to do anyway? And, when you start talking about machines taking over because they want to destroy humanitywell, its too late. Youre a piece of toast and the bots won. We need to get granular, not broad.

Do you agree? Disagree? If you have a reasonable argument to make about the dangers (or maybe the catbots) please send them to me. I promise to respond if youre interested in a civil discourse.

Continue reading here:

Sorry, Elon Musk. AI is not a bigger threat than North Korea - VentureBeat

I had just sat down to work one Saturday afternoon, when the familiar sound of the clothes washer, starting its spin cycle like an airplane taking off, started humming in the background. It was that sort of familiar noise that was both comforting and quickly drowns out the background, allowing me to sink quickly into a nice flow with some engineering work.

Suddenly, with a loud thump, the sound of a rattle, and something too awful to describe, the spinning machine came to a dramatic halt. I knew immediately it was the washing machine, as that peaceful hum was no longer softly blanketing the background. An uncomfortable silence was left in its void.

I walked over to the machine and made a quick inspection. Sure enough, there was a dim indicator on the front panel that readErras my clothes sat in a soapy swamp. My first indication was to go online and seek some machine first-aid atifixit.com. As an engineer myself, its almost a reflex to begin the troubleshooting process, no matter the medium.

Down the rabbit hole I went, educating myself on condenser units, evacuation pumps, controller computers, and the impressive array of components used to assemble these machines. Eventually some sort of alarm went off in my mind, and I was hit with the heavy reality of having wasted several hours attempting to gain expertise in a field I barely knew. So, I called the repair line and booked a repair.

Easy enough. In modern times, we have access to nearly immediate service only a phone call or screen-tap away. Though as I sat back down at my computer, I began to wonder if there was another, more efficient way to allow the manufacturer to diagnose and service my washer. After all, the selfish side of me reasoned it would save me, as the consumer, some additional TCO in the life of the appliance.

On the flip side, what if I wouldve just declared the device defective, irreparable, or obsolete? Would it have made its way to a recycling yard or trash heap as I enjoyed the delivery of a shiny new product? Appliance manufacturers are producing products withshorter lifespans than everand higher failure rates than their legacy counterparts.

This drives earlier whole-unit replacements and generates more waste. However, I would add that the millennial generation has a distaste for such environmental or corporate villainy, quickly sniffing out its presence and choosing the more sustainable option instead.

Consumer electronics are attractive, and the provocation of lust for the next best thing is always innate in their marketing strategies. However, what if we had another option where our devices could detect or predict failure, suggesting and even ordering replacement parts for us in the meantime? What if we could then be guided by the manufacturer through a mobile app, giving us the opportunity to save time and money by walking us through the replacement process?

If we were constrained for time, at least the manufacturer could realize savings, both environmentally and in labor cost, by invoking only one trip for the service technician. Even better, what if the device could fix itself?

Though the idea, at some levels, seems trivial, one critical piece to construct such a tool has been missing: visual cognition. Though computers have been getting better at recognizing individual parts, an ensemble of image recognition, cognition, and communication (chat-bot) are necessary for this type of automation.

Its at the intersection of these three that we can begin to create fully automated solutions where we can rapidly decrease our ejection of defective technology to landfills and simultaneously reduce our environmental impact, in the end saving money for both ourselves, and the producers of the products we enjoy.

Taking this one step further, with accurate failure reporting and detection, manufacturers can design products that perform their tasks more reliably and efficiently. Instead of ending up in the trash,as 70.8 percent of consumer electronics eventually do, fully functional hand me downs could enrich less affluent regions, with the maintenance cost also reduced through the above means.

Through I regularly defend the positive impact of AI to my friends who entertain a dystopian viewpoint, at the intersection of industries are possibilities that I gloss over on a daily basis. To a larger degree, even the receipt of defective machine parts for recycling could then be automated, allowing the return path to be optimized in a way that isnt currently possible.

My washing machine is now fixed, and I have some fresh, clean clothes. However, next time an appliance breaks and Im tempted to discard it, I would love to have Artificial Intelligence take care of the process, saving both environment and resources at the same time.

Continued here:

AI, responsible sustainability, and my broken washing machine - TNW

LONDON (Reuters) - Artificial intelligence robots are turbo-charging the race to find new drugs for the crippling nerve disorder ALS, or motor neurone disease.

The condition, also known as Lou Gehrig's disease, attacks and kills nerve cells controlling muscles, leading to weakness, paralysis and, ultimately, respiratory failure.

There are only two drugs approved by the U.S. Food and Drug Administration to slow the progression of ALS (amyotrophic lateral sclerosis), one available since 1995 and the other approved just this year. About 140,000 new cases are diagnosed a year globally and there is no cure for the disease, famously suffered by cosmologist Stephen Hawking.

"Many doctors call it the worst disease in medicine and the unmet need is huge," said Richard Mead of the Sheffield Institute of Translational Neuroscience, who has found artificial intelligence (AI) is already speeding up his work.

Such robots - complex software run through powerful computers - work as tireless and unbiased super-researchers. They analyze huge chemical, biological and medical databases, alongside reams of scientific papers, far quicker than humanly possible, throwing up new biological targets and potential drugs.

One candidate proposed by AI machines recently produced promising results in preventing the death of motor neurone cells and delaying disease onset in preclinical tests in Sheffield.

Mead, who aims to present the work at a medical meeting in December, is now assessing plans for clinical trials.

He and his team in northern England are not the only ones waking up to the ability of AI to elucidate the complexities of ALS.

In Arizona, the Barrow Neurological Institute last December found five new genes linked to ALS by using IBM's Watson supercomputer. Without the machine, researchers estimate the discovery would have taken years rather than only a few months.

Mead believes ALS is ripe for AI and machine-learning because of the rapid expansion in genetic information about the condition and the fact there are good test-tube and animal models to evaluate drug candidates.

That is good news for ALS patients seeking better treatment options. Famous sufferers include Lou Gehrig, the 1923-39 New York Yankees baseball player; actor and playwright Sam Shepard, who died last month; and Hawking, a rare example of someone living for decades with the condition.

If the research goes on to deliver new medicines, it would mark a notable victory for AI in drug discovery, bolstering the prospects of a growing batch of start-up companies focused on the technology.

Those firms are based on the premise that while AI robots won't replace scientists and clinicians, they should save time and money by finding drug leads several times faster than conventional processes.

Mead from Sheffield is working with BenevolentAI, one of a handful of British "unicorns" - private companies with a market value above $1 billion, in this case $1.7 billion - which is rapidly expanding operations at its offices in central London.

Others in the field include Scotland's Exscientia and U.S.-based firms Berg, Numerate, twoXAR, Atomwise and InSilico Medicine - the last of which recently launched a drug discovery platform geared specifically to ALS.

"What we are trying to do is find relationships that will give us new targets in disease," said Jackie Hunter, a former drug hunter at GlaxoSmithKline (GSK) who now heads Benevolent's pharma business.

"We can do things so much more dynamically and be really responsive to what essentially the information is telling us."

Unlike humans, who may have pet theories, AI scans through data and generates hypotheses in an unbiased way.

Conventional drug discovery remains a hit-and-miss affair and Hunter believes the 50 percent failure rates seen for experimental compounds in mid- and late-stage clinical trials due to lack of efficacy is unsustainable, forcing a shift to AI.

A key test will come with a Phase IIb study by Benevolent to assess a previously unsuccessful compound from Johnson & Johnson in a new disease area - this time for treating Parkinson's disease patients with excessive daytime sleepiness.

Big pharmaceutical companies like GSK, Sanofi and Merck are now exploring the potential of AI through deals with start-ups.

They are treading cautiously, given the failure of "high throughput screening" in the early 2000s to improve efficiency by using robots to test millions of compounds. Yet AI's ability to learn on the job means things may be different this time.

CPR Asset Management fund manager Vafa Ahmadi, for one, believes it is a potential game-changer.

"Using artificial intelligence is going to really accelerate the way we produce much better targeted molecules. It could have a dramatic impact on productivity, which in turn could have a major impact on the valuation of pharmaceutical stocks," he said.

Drugmakers and start-ups are not the only ones chasing that value. Technology giants including Microsoft, IBM and Google's parent Alphabet are also setting up life sciences units to explore drug R&D.

For Benevolent's Hunter, today's attempts to find new drugs for ALS and other difficult diseases marks an important test-bed for the future of AI, which is already being deployed in other high-tech areas such as autonomous cars.

"The aim is to show that we can deliver in a very difficult and complex area. I believe if you can do it in drug discovery and development, you can show the power of AI anywhere."

Reporting by Ben Hirschler; Editing by Pravin Char

Read the original here:

How AI robots hunt new drugs for crippling nerve disease - Reuters