DeepMind creates AI that can learn like a human
Select language to see a machine translation of this article. The original language of the Article is English and the translation is provided for your convenience.

DeepMind creates AI that can learn like a human

Google’s artificial intelligence (AI) division, DeepMind – praised by Oxford philosopher and AI expert Nick Bostrom as the company at the forefront of the industry – has just overcome a huge milestone: it has created an AI that can learn like a human.

Now, no one’s saying it learns as well as a human – that would mean the achievement of a major goal in AI research, described as “artificial general intelligence”. But what DeepMind’s new program does do is show a way that artificial intelligence researchers could bridge the gap between current forms of AI – which are very narrow and linear in application – and general AI, which would have the ability to learn from one task and apply it to another not just in a human-like fashion, but with a human-like success rate.

What DeepMind’s latest program does is kind of like a stripped down version of that. It does remember how it solved previous problems and tries to apply that knowledge to new ones, but compared to a human, it’s just not very good at it yet.

Still, Google’s latest foray into learning machines is a massive win for both the company and the industry. James Kirkpatrick from DeepMind told The Guardian that:

“If we’re going to have computer programs that are more intelligence and more useful, then they will have to have this ability to learn sequentially. Humans and animals learn things one after the other and it’s a crucial factor which allows them to learn continually and to build upon their previous knowledge.”

The researchers authored a paper in the peer-reviewed journal Proceedings of the National Academy of Sciences that described how their program used previously learnt skills to perform new tasks. However, it’s not clear that this actually made it any better in performing these tasks. Kirkpatrick wrote:

“We have demonstrated that it can learn tasks sequentially, but we haven’t shown that it learns them better because it learns them sequentially. There’s still room for improvement.”

However much improvement remains needed to realize the supercomputers of the future, DeepMind’s latest success will go down as a landmark moment in the quest to get there.

Disclosure
Dominion holds Alphabet, the parent company of Google, in its Global Trends Ecommerce Fund.


If you would you like to receive the Newsfeeds daily, please click here to sign up now!

Help us make this Newsfeed better by rating this article. 1 star = Poor and 5 stars = Excellent
0.0/5 rating (0 votes)

Disclaimer
The views expressed in this article are those of the author at the date of publication and not necessarily those of Dominion Fund Management Limited. The content of this article is not intended as investment advice and will not be updated after publication. Images, video, quotations from literature and any such material which may be subject to copyright is reproduced in whole or in part in this article on the basis of Fair use as applied to news reporting and journalistic comment on events.