Yoshua Bengio

---

The Last ten years, was a great day for Artificial Intelligence , but researchers believe that the industry is entering a new phase.

The Hype around AI has aroused and troughed over The Years , as the capabilities of the technology overestimated, And Then re-evaluated.

The peaks are known as AI summer and the troughs of the AI winters.

The 10 were probably the hottest AI summer on record with tech giants again and again, the KI-capabilities advertises.

KI-pioneer Yoshua Bengio , sometimes as one of the "founding fathers "AI", told the BBC that the AI skills were a little overhyped in the 10s of particular companies with an interest in the process.

however, There are signs that The Hype start can cool off.

Katja Hofmann, a senior researcher at Microsoft Research for the game Intelligence group

"I have the feeling that AI says the transition to a new phase", Katja Hofmann, a senior researcher at Microsoft Research in Cambridge.

Given the billions invested in AI, and the fact that it is likely to be more innovation, some researchers believe that it would be wrong to this new phase of an AI winter.

Robot Wars judge Noel Sharkey , who also said a professor of AI and robotics at the University of Sheffield, the BBC that he likes the term "KI-Autumn " and some other voices.

'feeling of the plateau,'

at the beginning of the 2010s, one of The World 's leading companies in the AI deep mind, which often develops to the so-called AGI-or "artificial general intelligence", which at some point in The Future .

machines, the thought about AGI - as far as the Holy Grail in AI would be as smart as people on the board, it promised.

deep mind the high AGI ambitions to the attention of Google caught, paid around £400m for the London-based AI lab in the year 2014, if they had injected The Following mission statement on its website: "Solve intelligence, And Then use it to anything else to solve. "

Several of the others began to talk to AGI a reality, including Elon Musk 's $1-billion-AI lab, OpenAI, and scientists such as mit professor Max Tegmark .

In the year 2014, Nick Bostrom , philosopher at Oxford University, went One Step further with his book, super intelligence. It is a world predicts, where the machines are firmly in control.

But these conversations were taken less and less seriously as the decade went on. By the end of 2019, the smartest Computer only excel in a "narrow" selection of tasks.

Gary Marcus , an AI researcher at New York University , said: "By the end of the decade there was a growing perception that the current techniques can only carry us so far. "

He believes the industry needs to go a couple of "real innovation" to continue.

"There is a General feeling of the plateau," said Verena Rieser, a professor in the plain text AI, in Edinburgh, at Heriot Watt University.

AI researchers, who said wishes to remain anonymous, we are in a period where we are particularly skeptical of the AGI.

"The Public 's perception of AI is becoming increasingly darker: The Public believes AI is a sinister technology," she said.

For its part, the deep mind has a rather optimistic view of AI the potential, suggesting that there are still "we are just scratching The Surface of what could be possible".

"is dissolving As a community and discovered more and more problems opening up," says Koray Kavukcuoglu, the Vice President for research.

"That's the reason why AI is a long-term scientific research trip.

"We believe that AI is one of The Most powerful enabling technologies ever created - a single invention that could unlock solutions to thousands of problems. The next decade will show, to generalize for a renewed effort, the capabilities of AI systems to reach that potential - both based on methods that were already successful, and the exploration of How To build general-purpose AI that can tackle issues in a variety of tasks. "

'Far away'

While AGI is not at any time be created soon, machines have learned complex tasks, such as:

The importance of this progress was overhyped sometimes, says ex-deep less Edward Grefenstette, who is now working in the Facebook AI Research group as a research Scientist .

Edward Grefenstette is a research Scientist at Facebook in London

"The Field is very far in The Last ten years, but we are very aware that we're still far to go in the scientific and technological progress, the machines are really smart," he said.

"One of the biggest challenges is to develop methods, which required significantly more efficient in terms of data and computing power, to learn, to solve a problem well. In The Last ten years, we have seen impressive progress made by the increase in the scaling of the data and the calculation, but that is not reasonable or scalable to every problem.

"the more complex If we want to scale-out behavior, we need to generalize better with less data, and we need more. "

Neil Lawrence, the recently, the Links to Amazon and entered the University of Cambridge, as The First deep mind-financed professor of Machine Learning , thinks that the AI industry is very much still in the "wonder years".

Reality check

So what will look like an AI as at the end of the 20s, and how the researcher will go about developing it?

"In the next ten years, I hope we'll see a more measured, realistic view of AI 's ability, rather than The Hype that we have seen so far," said Catherine Breslin, ex-Amazon AI researchers.

The term "AI" has a snap in a real buzzword in The Last ten years, with businesses of all shapes and sizes on the term, often for marketing purposes.

"The variety of things that were thrown in a pot, in the concept of "KI" to be recognized and discussed separately," says Samim Winiger, former AI researcher at Google in Berlin.

"What will we be as 'AI' or 'Machine Learning ' in The Last 10-20 years, as just one more form of 'calculation'".

google, artificial intelligence

Source of news: bbc.com