Learning to read "Alice in Wonderland" may be the key to teaching machines to understanding the way we speak.

At least that is the tactic Facebook is using to impove the language capabilities of M, the company's virtual assistance.

Facebook began testing M in October, and the early reports of it were ecstatic. People have used M to do everything from purchasing tickets to ordering food. But perhaps the reason M is so good at what it does is because it's not pure AI like Apple's Siri. 

Right now, Facebook's M is part-human, part-AI.

But Facebook's team is trying to improve M's language capabilities so it can better cater to users' needs. Mark Zuckerberg put up a Facebook post Thursday night explaining that the company is training a computer to predict the missing words in children's books for that very purpose.

"For this research, our team taught the computer to look at the context of a sentence and much more accurately predict those more difficult words — nouns and names — which are often the most important parts of sentences," Zuckerberg wrote in the post.

So far, the Facebook research team has discovered that its AI is best at predicting words when it's given just the right amount of context. "We call this 'The Goldilocks Principle,'" Zuckerberg explained.

"We still have a long way to go before machines can understand language the way people do, but this research takes us closer to building helpful services like M, our digital assistant in Messenger," Zuckerberg wrote.

Join the conversation about this story »

NOW WATCH: This simple math problem for kids is frustrating everyone on Facebook

The future is supposed to be a glorious place where robot butlers cater to our every need and the four-hour work day is a reality.

But the true picture could be much bleaker.

Top computer scientists in the US warned over the weekend that the rise of artificial intelligence (AI) and robots in the workplace could cause mass unemployment and dislocated economies, rather than simply unlocking productivity gains and freeing us all up to watch TV and play sports.

And a recent report from Citi, produced in conjunction with the University of Oxford, highlights how increased automation could lead to greater inequality.

'If machines are capable of doing almost any work humans can do, what will humans do?'

The rise of robots and AI in the workplace seems almost inevitable at the moment. At a conference on financial technology last week, pretty much every startup presenting included AI in some form or another and the World Economic Forum made "The Fourth Industrial Revolution" the topic of its Davos conference this year.

But The Financial Times reports that Moshe Vardi, a computer science professor at Rice University in Texas, told the American Association for the Advancement of Science over the weekend:

We are approaching the time when machines will be able to outperform humans at almost any task. Society needs to confront this question before it is upon us: if machines are capable of doing almost any work humans can do, what will humans do?

A typical answer is that we will be free to pursue leisure activities. [But] I do not find the prospect of leisure-only life appealing. I believe that work is essential to human well-being.

Professor Vardi is far from the first scientist to warn about the potential negative effects of AI and robotics on humanity. Tesla founder Elon Musk co-founded a non-profit that will "advance digital intelligence in the way that is most likely to benefit humanity as a whole" and Professor Stephen Hawking told the BBC in December 2014:"The development of full artificial intelligence could spell the end of the human race."

The World Economic Forum also backed up Professor Vardi's fears with a report released last month warning that the rise of robots will lead to a net loss of over 5 million jobs in 15 major developed and emerging economies by 2020.

'Increased leisure time may only become a reality for the under-employed or unemployed'

All these findings are fears are echoed in a recent research note put out by Citibank and co-authored by two co-directors and a research fellow of the University of Oxford's policy school, the Oxford Martin School.

Citi's global equity product head Robert Garlick writes in the report:

Could automation increase leisure time further whilst also maintaining a good standard of living for everyone? The risk is that this increased leisure time may only become a reality for the under-employed or unemployed.

The report, released last month and titled "Technology at work: V2.0", concludes that 35% of jobs in the UK are at risk of being replaced by automation, 47% of US jobs are at risk, and across the OECD as a whole an average of 57% of jobs are at risk. In China, the risk of automation is as high as 77%.

Most of the jobs at risk are low-skilled service jobs like call centres or in manufacturing industries. But increasingly skilled jobs are at risk of being replaced. The next big thing in financial technology at the moment is "roboadvice"— algorithms that can recommend savings and investment products to someone in the same way a financial advisor would. If roboadvisors take off it could lead to huge upheavals in that high-skilled profession.

Garlick writes:

The big data revolution and improvements in machine learning algorithms means that more occupations can be replaced by technology, including tasks once thought quintessentially human such as navigating a car or deciphering handwriting.

Of course, these are theoretical risks — technology exists or is in within reach that means these jobs could be done by robots and machines, but it doesn't necessarily mean they will be. And the report is, in general, optimistic about the future of automation and robotics in the workplace.

But Citi says governments and populations are going to have to prepare for these changes, which are going to hit the world of work faster than technology advances have in the past.

The report predicts that many workers will have to retrain in their lifetime as jobs are replaced by machines. Citi recommends investment in education as the single biggest factor that could help mitigate the impact of increased automation and AI.

'Inequality between the 1% and the 99% may widen as workforce automation continues'

But within that recommendation Citi hints at the biggest issue associated with rising robotics and automation in the workplace — inequality.

Citi's Garlick says that "unlike innovation in the past, the benefits of technological change are not being widely shared — real median wages have fallen behind growth in productivity and inequality has increased."

He writes later:

The European Centre for the Development of Vocational Training (Cedefop) estimated that in the EU nearly half of the new job opportunities will require highly skilled workers. Today’s technology sectors have not provided the same opportunities, particularly for less educated workers, as the industries that preceded them.

Not only is technology set to destroy low-skilled jobs, it will replace them with high-skilled jobs, meaning the biggest burden is on the hardest hit. The onus will be on low-earning, under-educated people to retrain for high-skilled technical jobs — a big ask both financially and politically.

Carl Benedikt Frey, co-director for the Oxford Martin Programme on Technology and Employment, writes in the Citi report (emphasis ours):

The expanding scope of automation is likely to further exacerbate income disparities between US cities. Cities that exhibited both higher average levels of income in 2000 as well as the average income growth between 2000 and 2010, are less exposed to recent trends in automation.

Thus, cities with higher incomes, and the ones experiencing more rapid income growth, have fewer jobs that are amenable to automation. Similarly, cities with a higher share of top-1% income earners are less susceptible to automation, implying that inequality between the 1 percent and the 99 percent may widen as workforce automation continues. In contrast, cities with a larger share of middle class workers also are more at risk of computerisation.

Hence, new jobs have emerged in different locations from the ones where old jobs are likely to disappear, potentially exacerbating the ongoing divergence between US cities. Looking forward, this trend will require workers to relocate from contracting to expanding cities.

And not only will the less well-off be forced to make the most changes in the robot revolution — reeducating and relocating — those that do retrain will be competing for fewer and fewer jobs.

Here's the Citi report:

This downward trend in new job creation in new technology industries is particularly evident starting in the Computer Revolution of the 1980s. For example, a study by Jeffery Lin suggests that while about 8.2% of the US workforce shifted into new jobs during the 1980s which were associated with new technologies; during the 1990s this figured declined to 4.4%. Estimates by Thor Berger and Carl Benedikt Frey further suggest that less than 0.5% of the US workforce shifted into technology industries that emerged throughout the 2000s, including new industries such as online auctions, video and audio streaming, and web design.

The study suggests that new technologies are creating fewer and fewer jobs and it is likely that advances in automation and AI will destroy jobs at a much faster rate than it creates new roles.

Citi says "forecasts suggesting that there will be 9.5 million new job openings and 98 million replacement jobs in the EU from 2013 to 2025. However our analysis shows that roughly half of the jobs available in the EU would need highly skilled workers."

Yes, automation and robotics will bring advances and benefits to people — but only a select few. Shareholders, top earners, and the well-educated will enjoy most of the benefits that come from increased corporate productivity and a demand for technical, highly-skilled roles.

Meanwhile, the majority of society — middle classes and, in particular, the poor — will experience significant upheaval and little upside. They will be forced to retrain and relocate as their old jobs are replaced by smart machines.

All hail our new robot overlords!

Join the conversation about this story »

NOW WATCH: How Oprah Winfrey earns and spends her billions

Everyone has that one fact they tell people when they want to blow their minds. Some of the best mind-blowing facts come from scientists, and artificial-intelligence researchers have some amazing ones.

Tech Insider talked to 13 AI researchers, roboticists, and computer scientists and asked what were the most surprising facts that they learned during their careers.

Scroll down to see their lightly edited responses.

SEE ALSO: A Google computer program just destroyed a human champion in a game that's even harder than chess

CHECK OUT: Here's how well an AI scored on the math section of the SAT

Matthew Taylor is wowed by how many times AI has outperformed humans.

"For me, it's the fact that computers can outperform humans at a number of tasks — and there are more and more things they can beat us at. There will always be things that humans can do better, but it's increasingly impressive the number of things that computers can do better than us.

"There's 'Jeopardy,' where the machine was able to outperform the top 'Jeopardy' contestants. There's games like backgammon, poker, and chess. Now computers are [beating humans at] the game of Go.

"The game of Go is something that people for a while thought that computers would never be able to do well in. Now computers are able to play."

Commentary from Matthew Taylor, a computer scientist at Washington State University.

Yann LeCun says he's amazed at how the simplest ideas always work.

"I will repeat what Geoffrey Hinton told me, following a talk I gave shortly after I left his lab and joined Bell Labs: 'If you do all the sensible things, it actually works.'

"The mind-blowing fact is that the simplest ideas work, and they become totally obvious in hindsight, but convincing the research community at large of what you consider obvious is far from easy."

Commentary from Yann LeCun, Facebook's artificial-intelligence research director.

Hector Geffner is in awe of theories that define how the world works.

"I grew up at a time when relativity and quantum mechanics were the cream of the crop of science. With time, I've learned to appreciate 'simpler' theories that probably have a much more direct influence in our lives and our identities.

"In particular, Darwin's theory of evolution and Turing's theory of computation. These are simple but profound theories with high reaching consequences."

Commentary from Hector Geffner, an AI researcher at Universitat Pompeu Fabra.

Facebook CEO Mark Zuckerberg held a live fireside chat in Berlin on Thursday morning to discuss some of his company's most innovative projects.

At the end of the interview, the moderator asked Zuckerberg one final question:

"How do you see Facebook in 2025?"

Zuckerberg said by that time, he hopes to see progress in three key areas: Connectivity, AI, and virtual and augmented reality.

In terms of connectivity, Zuckerberg said 3 billion people are currently connected to the internet, but there are 7 billion people in the world.

"I think at the current trend, if there’s no real new breakthrough technologies, we’re on trajectory to connect to another billion people," Zuckerberg said. "So we’ll still be here in 2020 or a little bit afterwards and there’ll be 4 billion people connected in the world, but still almost half the people in the world would not have the opportunities the internet brings. So we hope to really bend that curve, through applying these technologies, working in partnerships with all these folks. I don’t know if we’ll get all seven billion, but maybe five, or closer to six."

In the realm of artificial intelligence, Zuckerberg said the biggest question in the industry relates to "unsupervised learning," where machines can infer what they don't know about and are given no positive or negative reinforcement for offering up any kind of solution. But that's a big problem that will be solved over time; right now, Zuckerberg and company will focus on more immediate problems like improving voice recognition software.

"I did a demo earlier [today] that didn't work, so that'll work in 10 years," Zuckerberg said. "Hopefully it'll work sooner."

He says AI will contribute to some breakthroughs that are currently possible, but have yet to be rolled out. Things like self-driving cars — either fully autonomous cars or just more cars that drive themselves more at a time or brake more often when they sense an issue — and systems that can help doctors diagnose diseases will be in global circulation, Zuckerberg hopes.

The 31-year-old Facebook CEO also talked about the potential for virtual and augmented reality.

"There’s real science questions that still need to get solved to create the experience we all want, which is kind of like the glasses we all wear on a daily basis, not this big set of goggles," he said. "But we’re in the beginning of that industry... it’s going to be really exciting to check in 10 years from now and see where that’s going."

Right now, Facebook is heavily invested in virtual reality, where you're completely immersed in a virtual world — it acquired Oculus VR for $2 billion in March 2014, which will release its first VR headset on March 28. But Facebook is also working on augmented reality like Microsoft's HoloLens, where you can see digital elements in your field of view but you're still looking at the real world. "[Augmented reality is] a bit further out," Zuckerberg said in October.

Zuckerberg says these new reality-bending technologies will take at least 10 years to become a "mainstream big thing" like smartphones, which also took about a decade to reach one billion units sold. But he says Facebook is committed to VR and AR, and the company has the resources to invest around the world and improve research in these emerging technologies.

"I think virtual reality is going to make a big difference for giving everyone the power to share what they care about and helping everyone share in the opportunities of the internet,” Zuckerberg said.

Join the conversation about this story »

NOW WATCH: Facebook just made a massive change to its entire site

Silicon Valley giants like Google, Amazon, Facebook, and Apple are investing more of their time and resources into artificial intelligence (AI) research in the hope that it will improve their existing products and lead to new ones. 

But they're far from the only ones aiming to create machines that can learn and think for themselves.

A new generation of technology startups in Britain are focusing their efforts on developing products and services that are underpinned by AI, which has the potential to change the way we live our lives. Unfortunately, it also has the potential to end the human race if you listen to what billionaires like Elon Musk and intellects like Stephen Hawking have to say.

Here are 10 of the most interesting British AI startups to watch out for in 2016.

DeepMind — on a mission to solve general intelligence

Founded: 2011

Founders: Shane Legg, Mustafa Suleyman, Demis Hassabis

Headquartered: King's Cross, London 

Funding to date: acquired by Google for a reported £400 million

Google-owned startup DeepMind, a group of approximately 140 people aiming to "solve intelligence," is arguably one of the most interesting technology companies operating in the UK right now. 

Google bought DeepMind last year for a reported £400 million. Today it is using the company's technology across its organisation to make many of its best-known products and services smarter than they previously were. For example, it is starting to use DeepMind's algorithms to power its recommendation engines and improve image recognition on platforms like Google+. 

Magic Pony Technology — video compression platform

Founded: 2014

Founders: Dr. Zehan Wang, Rob Bishop

Headquartered: Kensington Olympia, London 

Funding to date: Undisclosed

Magic Pony Technology is pioneering a new approach to video compression by combining machine learning and computer vision research.

The company was founded by Rob Bishop who holds a masters degree in signal processing and Zehan Wang who holds a PhD in visual information processing. 

The company recently raised one of the largest seed rounds in Europe, at $2 million (£1.3 million).

Status Today — security startup that watches human behaviour

Founded: 2015

Founders: Ankur Modi and Mircea Dumitrescu

Headquartered: Liverpool Street, London 

Funding to date: Undisclosed

Status Today claims on its website that it employs a unique approach to security.

The company analyses behavior in the context of humans and their intended actions to protect individuals and the company they work for. 

Using machine learning techniques and "organisational human behaviour analysis," Status Today claims it can detect any possible malicious behavior, no matter how big or small. 

Fears of the oncoming artificial intelligence revolution are rampant and growing in Silicon Valley. 

If it comes down to man versus machine, more tech leaders are starting to worry that it will be the machine that will win. Industry leaders from Elon Musk to Peter Thiel to Reid Hoffman have poured money into a Y Combinator-led project to make sure it doesn't happen. 

In an interview with Axel Springer CEO Mathias Döpfner for the German newspaper "Die Welt am Sonntag", Facebook's CEO Mark Zuckerberg doesn't seem to worry much about it, calling Musk's reaction a little more on the hysterical side of things.

Rather, Zuckerberg argues that the machines will only overtake humans if we program them that way. Those chess-beating computers are designed to be that smart, they didn't just learn it, he argues. 

"I think that the default is that all the machines that we build serve humans so unless we really mess something up I think it should stay that way," Zuckberg told Döpfner in the interview.

Here's how Facebook's CEO sees the coming rise of robots and machines, and why it's not as scary as some make it out to be: 

Mathias Döpfner: How will Artificial Intelligence change society? 

Mark Zuckerberg: From my experience, there are really two ways that people learn. One is called supervised learning and the other is unsupervised. You can think of supervised learning as they way you read a children’s book to your son or daughter and point out everything.

Here’s a bird, here’s a dog, there’s another dog. By pointing things out, a child can eventually understand ‘oh that’s a dog’ because you told me 15 times that that was a dog.  So that’s supervised learning.  It’s really pattern recognition. And that’s all we know how to do today.

The other, the unsupervised learning, is the way most people will learn in the future. You have this model of how the world works in your head and you’re refining it to predict what you think is going to happen in the future.  Using that to inform what your actions are and you kind of have some model: Okay, I am going to take some actions and I expect this to happen in the world based on my action.  AI will help us with this. 

Döpfner: Can you understand the concerns that business magnate Elon Musk has expressed in that context? He seriously fears that artificial intelligence could one day dominate and take over the human brain, that the machine would be stronger than men. You think that is a valid fear or do you think it’s hysterical?

Zuckerberg: I think it is more hysterical. 

Döpfner: How can we make sure that computers and robots are serving people and not the other way around?

Zuckerberg: I think that the default is that all the machines that we build serve humans so unless we really mess something up I think it should stay that way. 

Döpfner: But in chess, Garry Kasparov was beaten by the computer Big Blue in the end. So there may be more and more situations where a computer is simply smarter than a human brain.

Zuckerberg: Yes, but in that case people built that machine to do something better than a human can.  There are many machines throughout history that were built to do something better than a human can. I think this is an area where people overestimate what is possible with AI. 

Just because you can build a machine that is better than a person at something doesn’t mean that it is going to have the ability to learn new domains or connect different types of information or context to do superhuman things. This is critically important to appreciate.

Döpfner: So this is science fiction fantasy and is not going to happen in real life and we don’t need to worry about the safety of human intelligence?

Zuckerberg: I think that along the way, we will also figure out how to make it safe. The dialogue today kind of reminds me of someone in the 1800s sitting around and saying: one day we might have planes and they may crash. Nonetheless, people developed planes first and then took care of flight safety. If people were focused on safety first, no one would ever have built a plane.

This fearful thinking might be standing in the way of real progress.  Because if you recognize that self-driving cars are going to prevent car accidents, AI will be responsible for reducing one of the leading causes of death in the world.  Similarly, AI systems will enable doctors to diagnose diseases and treat people better, so blocking that progress is probably one of the worst things you can do for making the world better.  

SEE ALSO: Mark Zuckerberg talks about the future of Facebook, virtual reality and artificial intelligence

Join the conversation about this story »

In the future the global employment market will rely heavily on robots, artificial intelligence, and all sorts of automation.

In fact, technology is so crucial going forward, that in January, the World Economic Forum predicted that in less than five years more than five million human jobs will be replaced by automation, AI, and robots. 

Just this week, a new report showed nearly a third of retails jobs in the UK could disappear by 2025, with many workers replaced by technology in some way or another.

But which jobs in London are at the most risk of being done by robots in the future? As part of its "Global cities, global talent" report, Big Four accountancy firm Deloitte took a look into the jobs in Britain's capital at the highest risk of automation moving forward.

Deloitte ranked job sectors using data from Oxford University academics Carl Frey and Michael Osborne, and the ONS, before creating a ranking from what it calls the "proportion of jobs at high risk of automation." Most of the jobs at risk of being taken by robots generally involved low skilled, manual work, but some of them might surprise you.

Check out the list below.

13. Information and communication — 3%: Information and communication workers don't need to lose much sleep over their incomes. A massive 97% of jobs in the sector are considered to be at "low risk of automation."

12. Education — 8%: Teachers don't need to worry too much about having their jobs stolen by robots in the coming years, with just 8% of jobs in the education sector at high risk of automation.

11. Arts, entertainment and recreation — 23%: Actors, musicians, and anyone else working in the entertainment industry have a 23% chance of having their job automated. However, nearly two thirds of jobs in the sector are safe for the time being.

A high-flying London tech entrepreneur turned up to Prime Minister David Cameron's house on Thursday evening wearing shorts and trainers after being invited to help choose the next "Founders of the Future."

The Founders of the Future initiative — being led by Lastminute.com cofounder Brent Hoberman through Founders Forum, a network of entrepreneurs and business leaders — is designed to uncover and nurture future technology founders across Europe that are aged 16-35.

Matt Miller, or "Mills" as he likes to be known, told Business Insider at Downing Street that he didn't want to pretend to be something he wasn't by wearing a suit.

"It’s about being real," said Mills. "I’m not going to turn up in something I don’t usually wear. If I was going to meet 100 of the future founders, I want them to know who I am."

He added: "These are my smart shorts so I have made an effort, I bought new trainers."

Mills is the cofounder of a digital product studio called ustwo, which was launched in 2004. The company employs approximately 300 people and has created games like Monument Valley and apps like Dice, which has been backed by the cofounders of Google DeepMind.

Digital economy minister Ed Vaizey applauded Mills for his "extremely impressive tech dress down" while giving a speech about tech entrepreneurship at the Founders of the Future reception.

Heading to the @Number10gov crib ~ gonna be smooth pic.twitter.com/SJn0POFj9H

— RODEO ROBOT RANGER (@millsustwo) March 3, 2016

Founders Forum claims the 100 Founders of the Future were selected using an artificial intelligence (AI) algorithm and peer recommendations from Mills and the likes of Niklas Zennstrom of Skype, Jimmy Wales of Wikipedia, and Dame Natalie Massenet.

Mills selected 24-year-old Rikke Koblauch as one of the Founders of the Future. She is part of the UX/UI team at Ustwo and was also selected by the AI algorithm.

Hoberman said that he hoped the event would encourage the 100 Founders of the Future to start their own businesses. "We're trying to build a community of incredible entrepreneurs that aren't yet entrepreneurs," he said.

The cofounder of one successful startup whose employee attended the event told Business Insider: "It was more like look at the most promising hyped companies, invite the management."

Other Founders of the Future include George Burgess, who has created an education app called Gojimo, Dylan Baker, who used to work as a journalist for Tech City News and is about to join ecommerce startup Yieldify, and Josephine Goube, who is co-directing manager at Girls in Tech UK, which is an organisation aiming to get more young female entrepreneurs into technology. One other notable inclusion was Mustafa Al-Bassam, who at 16 was jailed for 20 months for hacking the computer systems of Sony and the CIA.

After the event, Baker said: "Founders of the Future is great because it is taking a very proactive approach to creating the next generation of high growth businesses in the UK.

"I think there were lots of people in the room last night who previously wouldn't necessarily have considered entrepreneurship as part of their career trajectory but after having felt the buzz that there was at the launch event and having engaged with their peers and established entrepreneurs in the room, will definitely consider it going forward, and that's incredibly important."

Join the conversation about this story »

NOW WATCH: We tried Lenovo’s new $500 tablet that features a built-in projector

Donald Trump has mastered the art of the one-liner.

The US presidential candidate now has a Twitter bot twin that tries to mimic his signature soundbites.

The bot tweets Trump-like statements using an artificial intelligence algorithm based on hours of the candidate's debate speech transcripts.

"Trump’s language tends to be more simplistic, so I figured that, as a modeling problem, he would be the most manageable candidate to study," says Bradley Hayes, the bot's creator and a researcher at MIT's Computer Science and and Artificial Intelligence Lab, in a statement. 

Hayes calls the bot DeepDrumpf, which refers to John Oliver's recent segment about Trump's ancestral name. 

Take a look at a few of DeepTrumpf's hilarious, bizarre, and oddly poetic tweets.

The bot creates tweets one letter at a time. For example, if the bot randomly begins with "M," it will likely follow with an "A,""K," and "E," until it produces Trump's campaign slogan, "Make America Great Again." It then starts over for the next sentence until it reaches the 140-character limit.

Tweet Embed:
https://twitter.com/mims/statuses/705598421508481024
...whather it's not a great guns -- nominating security in america. We can make America great again. Thank you very much

Just like the real Trump, DeepTrumpf thinks many things are "great,""not great," or could be "great again."

Tweet Embed:
https://twitter.com/mims/statuses/705604644114927616
Great manufacturing, bring back our jobs, bring back our manufacturing, because my file, you know, I don’t need anybody’s money. ...

 Tweet Embed:
https://twitter.com/mims/statuses/705552709911445505
[Tonight's debate will ]make America great.

 Tweet Embed:
https://twitter.com/mims/statuses/705519209418268672
The military, make it great again. And we’re 26th in the world on President Obama’s illegal executive order.

It doesn't seem to be too far off from Trump's speech patterns IRL. Compared to the last five US presidents, "Trump's language is darker, more violent and more prone to insults and aggrandizing," according to a recent analysis by The New York Times.

Tweet Embed:
https://twitter.com/mims/statuses/705476932742156288
I’ll bring back our jobs. They all have everything, they don’t have a clue.

Tweet Embed:
https://twitter.com/mims/statuses/705480367239659520
I’m what ISIS doesn’t need.

Source: The New York Times

Eric Schmidt, the chairman of Alphabet, Google's parent company, believes that breakthroughs in artificial intelligence (AI) will make every human on the planet better in some way.

Speaking in South Korea on Tuesday, Schmidt said: "Advances in artificial intelligence and machine learning will make each and every other human being in the entire world, smarter, more capable, better human beings."

Other powerful tech leaders and scientists including PayPal founder Elon Musk and renowned physicist Stephen Hawking have warned that advances in artificial intelligence should be treated with caution, with Hawking going as far as to say that AI could spell the end of mankind.

Schmidt's comments were made in Seoul, where Google subsidiary company DeepMind, a London AI startup, is about to pit its "AlphaGo" algorithm against Lee Sedol— the best player "Go" player in the world. Go is a Chinese boardgame with billions of possible moves that has been impossible for machines to master.

"I’ve concluded that the winner here, no matter what happens, is humanity," Schmidt said, referring to DeepMind as a "brilliant company" that has been created by three of his "friends."

He added: "This is a great day for humanity and this week we do not know who will win but we know at the end humans will be smarter and the world will be a much better place because of all the things all of us are doing to make this technology viable."

Developments in the field of artificial intelligence (AI) have been slow during Schmidt's lifetime, the billionaire and former Google CEO continued.

"I have been a computer scientist for my whole life," said Schmidt. "In the 1960s there was great hope that we would be able to do the things that we’re going to do this week. But for 30 years, a time that is called the artificial intelligence winter, people made all of these claims and nothing happened. People tried and they tried and they tried.

"But in the last decade, there has been a huge increase in steps forward for a number of reasons: new algorithms, faster computers, many more people, more money."

Schmidt highlighted how people use AI in many ways today, pointing to a number of Google services that are underpinned by AI, including Google Image Search and Google Translate.

Join the conversation about this story »

NOW WATCH: This bicycle generator could bring electricity to millions of people living without it

Demis Hassabis, the CEO and cofounder of the artificial-intelligence company DeepMind, is in South Korea this week hoping that an AI created in a discreet office in London's King's Cross will outsmart the best player in the world at a complex Chinese board game known as Go.

It’s a huge deal in the field of AI and potentially a landmark moment for Hassabis and DeepMind. Unlike chess, where computers have already beaten the best human players, Go has billions of possible moves (more moves than there are atoms in the universe, according to Hassabis), meaning it can’t be mastered by brute calculation.

Some of the most impressive minds at some of the biggest technology companies in the world — including Facebook CEO and cofounder Mark Zuckerberg— have turned their attention to building AIs that can beat humans at Go. But it’s the 39-year-old from Finchley, North London, who is leading the charge, albeit with the help of some of DeepMind’s 200 or so employees.

Hassabis has risen to fame in the AI world since selling DeepMind, which he founded with Mustafa Suleyman and Shane Legg in 2012, to Google for a reported £400 million ($568 million) in 2014.

This week, the entrepreneur — who holds a degree in computer science from Cambridge and in neuroscience from University College London — faces what is arguably his biggest challenge to date. The self-learning AlphaGo algorithm he has led the development on is being pitted against Lee Sedol in a five-game Go match at the Four Seasons Hotel in Seoul. Prize money for the match stands at $1 million (£700,000).

AlphaGo has already beaten Fan Hui, the European Go champion, but the outcome of the match against Sedol is anybody's guess.

Sedol is confident that he has what it takes to beat Hassabis, who was described as one of the smartest human beings on the planet by internet creator Tim Berners-Lee, according to this profile published by The Guardian.

"I'm confident about this match," Sedol said Tuesday at a news conference in Seoul. But after hearing Hassabis tell a room full of media members how AlphaGo works, Sedol said, "I'm quite nervous so I might not have a 5-0 victory now."

The general public will most likely question the importance of developing a machine that can take on humans at a board game but Hassabis says this is all part of a bigger vision. Ultimately, he wants to apply DeepMind's AI technology to some of the world's biggest problems, including healthcare, climate change, and macroeconomics.

Join the conversation about this story »

NOW WATCH: Avoid these 5 mistakes when texting someone you want to date

If you're familiar with DeepMind — the research-intensive London startup that's taken on the best player in the world at notoriously complex Chinese board game "Go"— then you've probably heard of Demis Hassabis by now. He's the CEO and cofounder of DeepMind. You may have also heard of Mustafa Suleyman and Shane Legg, the other cofounders. One person you may not have heard of, however, is David Silver.

According to The Guardian, Silver is the main programmer on the Go team at DeepMind, which was bought by Google for £400 million in 2014.

Despite contributing to more research papers (16 now) than any other DeepMind employee, Silver has largely stayed out of the limelight. He consulted for DeepMind from its inception and joined full-time in 2013.

He's got an impressive academic record, having achieved top marks in his computer science class at Cambridge University, which is where he met and befriended Hassabis, who reportedly taught Silver how to play board games, including Go. "Dave and I have got a long history together," Hassabis told The Guardian in February. "We used to dream about doing this [creating powerful AIs] in our lifetimes, so our 19-year-old selves would probably have been very relieved that we got here."

After Cambridge, Silver cofounded the videogames company Elixir Studios, where he was CTO and lead programmer, winning a number of awards for technology and innovation.

Silver returned to academia in 2004 to study for a PhD on reinforcement learning in computer Go, making him an ideal recruit for DeepMind. During his PhD, he cointroduced the algorithms used in the first "master-level" Go programs. However, they could only beat humans on 9x9 boards, not the standard 19x19 boards, which allow for more moves, thereby making them harder for computers to grasp.

DeepMind has been relatively quiet about who builds the AIs that are taking on the best humans at computer games like "Space Invaders" and Go. In the last month, however, Google DeepMind has started to open up, possibly in a bid to capitalise on the growing interest in its AlphaGo algorithm, which is taking on Go world champion Lee Sedol in Seoul this week. 

Join the conversation about this story »

NOW WATCH: John McAfee explains why an iPhone backdoor is a terrible idea

Google just made artificial-intelligence history.

A program called AlphaGo, designed by Google's DeepMind artificial-intelligence team, just won a game against Lee Sedol, one of the world's greatest Go players.

AlphaGo AI won the game in Seoul, South Korea, by resignation after 186 moves. The win marks the first time that a computer program has defeated a top-ranked human Go player on a full 19-by-19 board with no handicap.

DeepMind founder Demis Hassabis took to Twitter to celebrate, saying his team had "landed on the moon."

Lee and AlphaGo are playing a series of matches over the course of five days. Go looks to be the latest in artificial intelligence's mastery of games. Checkers fell in 1994, and it was followed by chess in 1997 and Jeopardy in 2011. In October, AlphaGo became to first program to beat a professional Go player; now it's taking on one of the best players alive.

"It's definitely an important milestone," Brown University computer scientist Michael L. Littman tells Tech Insider.

What makes Go — a game whose dominance by humans seemed secure as recently as 2014— such a beguiling target for artificial intelligence is the nature of the game.

Created in China 2,500 years ago, Go appears simple. A game begins with an empty board. Two players (one using black stones, the other white), alternate placing stones in squares, trying to grab territory without getting their pieces captured.

As Alan Levinovitz noted in Wired, the game quickly gets complex. There are 400 possible board positions after the first round of moves in Chess and 129,960 in Go. There are 35 possible moves on any turn in a Chess game, compared with 250 for Go.

In a blog post in January, DeepMind's David Silver and Hassabis noted that the search space (the number of possible board configurations) in Go is larger than the number of atoms in the universe.

Given that level of complexity, DeepMind couldn't rely on what's called brute-force AI, in which a program maps out the breadth of possible game states in a decision tree.

As Business Insider's Tanya Lewis has noted, AlphaGo combines two AI methodologies:

• Monte Carlo tree search: This involves choosing moves at random and then simulating the game to the very end to find a winning strategy.
• Deep neural networks: A 12-layer network of neuron-like connections that consists of a "policy network" that selects the next move and a "value network" that predicts the winner of the game.

DeepMind didn't "program" AlphaGo with evaluations of "good" and "bad" moves. Instead, AlphaGo's algorithms studied a database of online Go matches, giving it the equivalent experience of doing nothing but playing Go for 80 years straight.

"This deep neural net is able to train and train and run forever on these thousands or millions of moves, to extract these patterns that leads to selection of good actions," says Carnegie Mellon computer scientist Manuela Veloso, who studies agency in artificial-intelligence systems.

"Deep learning has been limited to descriptions, putting captions on images, saying, 'This is a cat or a laptop,'" she tells Tech Insider. But with AlphaGo, "It's the ability, given the description, and the value of the game state, which action should I take."

Google acquired DeepMind in 2014. Founded in 2010 by chess prodigy turned artificial-intelligence researcher Hassabis, the company's mission is to "solve intelligence," and it says its algorithms "are capable of learning for themselves directly from raw experience or data."In February 2015, DeepMind revealed in Nature that the program learned to play vintage arcade games like Pong or Space Invaders as well as human players. 

Littman, the Brown computer scientist, says he could see AlphaGo's technology applied toward Google's self-driving cars, where the AI has to make lots of little decisions continuously, similar to a game of Go. It could also be used in a problem-solving search capacity, if, for example, you wanted to ask Google to give you a recipe for baking a cake for your gluten-free cousin.

"It's inevitable that we have Go programs that beat the best people," Littman says. "What we're finding is that any kind of computational challenge that is sufficiently well defined, we can build a machine that can do better. We can build machines that are optimized to that one task, and people are not optimized to one task. Once you narrow the task to playing Go, the machine is going to be better, ultimately."

You can watch AlphaGo's matches against Lee here.

Rob Price and Sam Shead contributed reporting.

Join the conversation about this story »

NOW WATCH: Mark Zuckerberg explains why the FBI is wrong on encryption

Just as ATMs changed banking and computers took over the home and workplace, robots and artificial intelligence are going to transform a bunch of industries over the next decade.

By 2025, a machine may be putting together your driverless car in a factory with no human oversight. A robot maid could be cleaning up after you at home, and your financial advisor might be a computer investing for you automatically. 

And with at least 90 countries operating unmanned aerial vehicles, the wars of the future may increasingly be fought with "drone" aircraft.

These are just some of the interesting — and sometimes scary — predictions to come from a 300-page report released by Merrill Lynch in November, which estimates the global market for robots and AI will grow from $28 billion to more than $150 billion just five years from now.

There's plenty of disruption bound to happen across the world as drones and much-smarter-than-you AI take over. But we're likely to see the biggest changes across eight industries in China, Japan, the US, and Korea — the countries currently investing the most in these technologies.

Here are the big predictions from Merrill Lynch:

The auto industry is going to change big-time, especially when fully autonomous — aka driverless — cars officially go mainstream.

Over the next five years, the report says most new cars will be smarter "connected" cars, and in 2025, that'll mean about 10% of them are fully autonomous.

While the initial price will be about $10,000 more than regular cars, it will inevitably come down as more people and companies adopt them.

The dystopian war between robots and humans of the "Terminator" films is probably not going to happen, but there is still reason to worry about so-called "killer robots."

A new report by Paul Scharre of the Center for a New American Security argues that, while militaries develop semi- and fully-autonomous weapons systems such as missiles and drone aircraft, they are facing "potentially catastrophic consequences" if human controllers are taken out of the loop.

"Anyone who has ever been frustrated with an automated telephone call support helpline, an alarm clock mistakenly set to ‘p.m.’ instead of ‘a.m.,’ or any of the countless frustrations that come with interacting with computers, has experienced the problem of ‘brittleness’ that plagues automated systems," Scharre, a former Army Ranger who helped draft policies related to autonomous weapons systems for the Pentagon, writes in the report.

His main point: Automated systems can be really useful, but they are limited by their programming, and lack the use of "common sense" that a human may employ in certain cases.

Such was the case in 1983, when the human skepticism of Stanislav Petrov, a Russian military officer, was the biggest safety in stopping the Soviet Union from launching its missiles, after an early-warning system reported five incoming missiles from the United States.

It was a computer error. A fully-automated system had the data and would have launched. But Petrov rightly believed the system was malfunctioning.

"What would an autonomous system have done if it was in the same situation as Stanislav Petrov found himself on September 26, 1983? Whatever it was programmed to do."

Software error, cannot compute. Launch missile?

Scharre evaluates a number of past failures — involving humans and automated systems — to illustrate his point. While humans were in the loop during the disasters at Three Mile Island or Fukushima, these rare accidents expose the problem with tightly-controlled systems.

In the case of Fukushima for instance, many of the safety features activated by loss of power, flooding, and earthquakes worked as designed, but the engineers did not account for the possibility that all three of these things could happen at the same time. 

Engineers may be able to hypothesize and program the machine's response to nightmare scenarios but on a "long enough time horizon," Scharre writes, "Unanticipated system interactions are inevitable."

When something unanticipated happens to a computer that isn't programmed to deal with it, plenty of bad stuff can happen. Most computer users know the famous "blue screen of death" error and constantly update their software to fix bugs, and security problems are often found in systems after they have been exploited by hackers.

"Without a human in the loop to act as a fail-safe, the consequences of failure with an autonomous weapon could be far more severe than an equivalent semi-autonomous weapon," he writes.

Scharre advocates a similar framework for humans and machines to work together, called "centaur warfighting." It's based on Gary Kasparov's model of "centaur," or advanced chess — in which an artificially-intelligent machine helps the Chessmaster think smarter about his next move.

"The best chess players in the world are human-machine teams," he writes.

Join the conversation about this story »

NOW WATCH: Psychologists discovered how to make people like you

Artificial intelligence (AI) became a scientific field almost 60 years ago. Ever since then, researchers have tried to achieve human-level smarts or better.

Yet even with recent feats of computational genius — for example, Google DeepMind beating a human player in the game Go — AI scientists say they still have a long road ahead.

Tech Insider spoke with AI researchers, computer scientists, and roboticists around the world about what it is going to take to build a machine that's able to think, work, and feel like a human.

Scroll down to see their lightly edited responses.

Bart Selman said computers need to learn how to understand the world like a human.

"The big obstacle, though it's not an obstacle because I think it will just take time, is the computer has to learn more about the way we see the world.

"It's very hard to understand the world from a human perspective. Intelligence relies on the way we view the world as humans, and the way we think about the world.

"Computers are just starting to be able to hear and starting to being able to see images. Those are tremendous improvements in the field in the last five years.

"We're doing that by having computers read millions of texts and pages from the web, by hooking them up to cameras and moving them around human environments."

Commentary from Bart Selman, a computer scientist at Cornell University.

This experience of the world will foster more intelligent AI, Peter Norvig says.

"AI needs to experience living in the world.

"We are very good at gathering data and developing algorithms to reason with that data. But that reasoning is only as good as the data, which, for the AI we have now, is one step removed from reality.

"Reasoning will be improved as we develop systems that continuously sense and interact with the world, as opposed to learning systems that passively observe information that others have chosen."

Commentary from Peter Norvig, director of research at Google.

To do that, Yoshua Bengio says computers should be trained to learn like children.

"Right now, all of the impressive progress we've made is mostly due to supervised learning, where we take advantage of large quantities of data that have already been annotated by humans.

"This supervised learning thing is not how humans learn.

"Before two years of age, a child understands the visual world through experiencing it, moving their head and looking around.

"There's no teacher that tells the child, 'in the image that's currently in your retina, there's a cat, and furthermore it's at this location' and for each pixel of the image say 'this is background and this is cat.' Humans are able to learn just by observation and experience with the world.

"In comparison to human learning, AI researchers are not doing that great."

Commentary from Yoshua Bengio, a computer scientist at University of Montreal.

Google has just reached a milestone for artificial intelligence (AI). AlphaGo, a program designed by the company's DeepMind AI division, just beat Lee Sedol, a world champion Go player.

It's another entry in a growing list of computers that can beat us at our own games.

While we're (unfairly) quick to cede our human shortcomings to the unblinking precision of computers, however, our brains are really good at something current AI can't quite crack: Thinking.

We can process our thoughts at a level unparalleled by anything else we've come across, living or artificial. And we can do it with remarkable efficiency.

One of the best characterizations of just how ill-equipped modern artificial intelligence is comes from the creator of the "Jeopardy"-dominating IBM Watson computer, Dave Ferrucci.

As The Atlantic wrote:

"He likes to tell crowds that whereas Watson played using a room’s worth of processors and 20 tons of air-conditioning equipment, its opponents relied on a machine that fits in a shoebox and can run for hours on a tuna sandwich. A machine, no less, that would allow them to get up when the match was over, have a conversation, enjoy a bagel, argue, dance, think—while Watson would be left humming, hot and dumb and un-alive, answering questions about presidents and potent potables."

That's not to say that current AI is necessarily weak, though. It depends if you take more of a generalist or specialist perspective.

On one hand, machines are nowhere near competing with the wide range of humanity's achievements — which only required some food, water, and oxygen to pull off.

On the other hand, AI definitely has us beat when it comes to things like precision computing and processing power. This makes machines really good at specialized tasks.

AI systems are also getting better at learning. A program from Stanford University, for example, figured out how to fly a model helicopter at world champion level in mere hours just by watching. Others are processing boxes in Amazon warehouses and translating our speech almost instantly.

But something more nuanced, artistic, and human, like writing, is still hilariously bad when done by AI, so you probably won't have to worry about robots stealing our jobs anytime soon — or at least all of them, anyway.

SEE ALSO: 'WE MADE HISTORY': Google's DeepMind AI just beat a human world champion at Go

Join the conversation about this story »

NOW WATCH: 7 iPhone texting tips only power users know about

Google's DeepMind AI just beat a human world champion at the ancient game of Go for a second time, cementing its historic achievement.

The AlphaGo software played Lee Sedol in South Korea on Thursday, in the second of a series of five planned matches.

It was a closely fought game, with both players going into overtime.

The games are a significant milestone for AI research. Go is a simple game but has been notoriously difficult for computers to master because of the sheer number of potential moves. While AI programs began being able to beat human champions at chess decades ago, the best Go players in the world have always been able to outsmart Go-playing software — until now.

Go expert Yoo Changhyuk said Lee tried to make "difficult moves to agitate AlphaGo" in the first game." Today, he tried the opposite — he played safe and entered the endgame," he said. "He made some mistakes, which I think caused the defeat."

Google DeepMind CEO and cofounder Demis Hassabis said in a press statement: "That was dramatic to say the least! Lee Sedol put up an incredible performance and had our expert commentators divided over the result until the end. AlphaGo played some really surprising and beautiful moves in this game."

Lee added: "Yesterday, I was surprised, but today I am quite speechless. If you look at the way the game was played, it was a very clear loss on my part. Yesterday I felt like AlphaGo played certain problematic positions, but today I felt that AlphaGo played a near perfect game. There was not a moment I felt like its moves were unreasonable."

If you’re not familiar with Go, it’s a two-player turn-based strategy game. Each player puts down either black or white stones in an attempt to outmaneuver and surround the other player. It's easy to pick up but takes years to master.

AlphaGo won the first game On Wednesday after Lee Sedol resigned after 186 moves.

Before the first game, Lee Sedol had said he was "confident" about his prospects, telling reporters he was aiming for a 5-0 victory across all the games. But there's now a real possibility that AlphaGo could win every game.

The contenders need to win at least three games to win the Challenge Match. Victory for AlphaGo is now in sight.

You can watch the entire game here:

Join the conversation about this story »

NOW WATCH: THE 'EYEBORG': This filmmaker lost an eye — and replaced it with a tiny camera

Google's DeepMind artificial intelligence team is making history.

Its AlphaGo program is up 2-0 on Lee Sedol, one of the top Go players alive. This is the first time a computer has beat a human champion without a handicap. 

While there will be three more games between Sedol and AlphaGo, the victory suggests that Go is the latest game that computers are outwitting people in. Checkers fell in 1994, Chess in 1997, and Jeopardy in 2011.

"It there's a social component with players playing together, it's not clear," Littman says. 

"If it's a fixed set of rules, with players acting according to preexisting behaviors, then it becomes a sufficiently well-defined problems that you can optimize against and the machine will do better than people," he says. "If it's fluid and changing and contextual, it's much harder to say what it means to do well." 

He gave the example of crossword puzzles, which are an interaction between the crossword puzzle maker. The maker can bend the rules of the game, like by letting the player put two letters in a square, and the crossword player.

A team of Littman's students helped make Proverb, a crossword playing AI, that played better than the average player but not as well as a champion. Littman recalls one puzzle that was bedeviling for the program — where all the clues were spoonerized. For instance, one clue was "home is near," meaning "Nome is here," so the answer was Alaska. (For more on Proverb, check this analysis.)

"In puzzles like Sudoku, where the rules are crisp, there's no issue. Computers crush human beings," he says. "But games like crossword puzzles, where the rules are fluid, then it seems like we find something that beats people."

The lines become even more clear when you extend the "games" to include sports. It's not hard to imagine a machine that could beat a human in shotput, Littman says; consider the nearest cannon. 

But the more creative the sport is, the harder it is to mechanically replicate. 

"With figure skating, you're trying to do something that creates an impact on the judges, surprises them with the artistry," he says. "The idea that we'd have a computer figure skater that beats the best humans is hard to imagine, but building a machine that beats a human in the hundred meter dash, that's obvious."

Join the conversation about this story »

NOW WATCH: Neil deGrasse Tyson explains why killer robots don't scare him

Google's DeepMind just won its second of five games of the classic board game Go against a top-ranked world champion — a feat that impressed even Elon Musk.

Now that Google has proven that the DeepMind artificial intelligence is capable of playing the notoriously complicated Go at a world-class level, what's next?

"'StarCraft,' I think, is our likely next target," Google Senior Fellow Jeff Dean said at today's Structure Data event in San Francisco. 

Dean is referring to the 1998 smash-hit PC game "StarCraft," which casts players as supreme commanders in an interstellar conflict between humans, the bug-like Zerg, and psychic warrior Protoss. Players compete to gather resources, build their forces, and outmanuever their opponents. 

Board games like Go and chess are what researchers call a "perfect information" game, where both players have a total awareness of everything that happens on the board at all times. 

"The thing about Go is obviously you can see everything on the board, so that makes it slightly easier for computers," Google DeepMind founder Demis Hassabis told The Verge recently.

Meanwhile, games like "StarCraft" and its sequel keep your opponents' moves largely secret, at least until you come in to conflict — skilled players watch closely for clues as to their opponents' strategy and try to anticipate their next move. 

Of course, when Google is ready to test DeepMind's "StarCraft" skills, there are no shortage of skilled players who might be willing to step up: StarCraft birthed an insanely competitive professional gaming scene, including a status as a major spectator sport in South Korea.

Here's a professional StarCraft 2 match in action:

"You have to keep track of things happening off the screen," Dean says.

It means that Google's DeepMind would have a brand-new challenge of trying to outguess their opponent, and react if and when they come up with something totally crazy. It would test a new set of skills for artificial intelligence. 

 "Ultimately we want to apply this to big real-world problems," Hassabis told The Verge.

SEE ALSO: This Microsoft exec's hilarious presentation fail shows why computer translation is so difficult

Join the conversation about this story »

NOW WATCH: This heart surgeon saved a baby's life with Google Cardboard