Will AI Summer Ever Come?

Will it ever be AI summer? The House of Lords report on AI kicks off with a list of ‘AI winters’ when heady expectations for AI were followed by a depressed realisation in the scientific community that the technology was not there.

The HoL report quotes the 1956 Dartmouth College, New Hampshire study into AI, instances the 1960s’ AI initiatives at MIT, Carnegie Mellon, Edinburgh University and Stanford, and notes the 1980s’ Alvey programme which were followed by what scientists call ‘AI winters’ when scepticism towards AI became the prevailing mood of  scientists.

Now, says the HoL, ‘another wave of excitement’ about AI has been generated by neural nets and deep learning algorithms.

For a definition of AI the HoL report adopts this one: “Technologies with the ability to perform tasks that would otherwise require human intelligence, such as visual perception, speech recognition, and language translation”, but adds the caveat: “Our one addition to this definition is that AI systems today usually have the capacity to learn or adapt to new experiences or stimuli.”

When I spoke to the head of the HoL committee on AI, Lord Clement-Jones, I asked him if this was a report on what we do If AI happens and not on How or When it will happen, and he replied: “Yes”.

Clement-Jones added that so far as ‘Strong AI’ is concerned “that could be two decades down the track.” The term Strong AI means a machine which is as intelligent as a human.

In specific fields, though, AI is very much with us, said Clement-Jones, pointing to, inter alia, facial recognition, fintech and the increasing sophistication of Google search answers. “We need to be prepared for what we’ve got coming along currently,” he said, and the report addresses this requirement comprehensively.

Clement-Jones said that the UK’s AI R&D effort must address all aspects of AI – Strong AI and specific applications of AI.

Should we focus more on developing AI products or on intelligently applying AI products? I asked. “Both”, he replied “we don’t want to get left behind. In the UK we’ve not been fantastic at applied technology – the Americans have done it better than us – now we have a chance to apply AI.”

A key point is the problem that some of our best AI start-ups are bought by the big tech companies e.g. DeepMind being bought by Google.

“DeepMind couldn’t get growth funding which is why it got sold to Google, “ said Clement-Jones, “we need to set up a Growth Fund so companies don’t need to sell out early.”

The Growth Fund is one of the major recommendations of the report.

A big problem identified by the report is the lack of skilled people in the AI field. Asked about this, Clement-Jones replied: “PhDs need to be funded. Who does it? Government? Industry?”

The government recently agreed to fund 200 more PhDs in AI, and Clement-Jones commented: “We need thousands, not hundreds, and it’s not just a question of computer skills it’s also about creative studies into how we use AI.”

AI summer is not here but, when it comes, the House of Lords has shown how it can best be enjoyed


Comments

18 comments

  1. You were very prescient Duncan, maybe Elon started believing all this Californian crap abour the Singularity and robots taking over and AI being as smart as human intelligence and thought if he put his money where his mouth is and built a factory based on it, it would be more efficient than factories staffed by people. For several decades manufacturers have talked about ‘dark factories’ where the only humans are the maintenance men. But, the future is a pesky lady who is not prepared to be predictable.

  2. After my comment about Japanese car makers it is interesting that within the past few day Elon Musk has decided that “humans-are-underrated” and presumably that the robots in his automated car factory that hardly makes any cars were overrated (or maybe so underrated that they kept breaking 🙂

  3. Very wise words, zeitghost, we have six billion brains on earth better than anything AI is likely to come up with for at least the next several decades, and what do we do with them? We let them starve, leave most of them uneducated and many of them unused or under-used. Training and putting the six billion brains to work should be planet’s top priority – instead the perversity of human nature sees the development of AI as a way to put as many of the six billion out of work as possible. How b. daft we all are.

  4. Yes indeed Mike, struggling through a winter which never seems to end – I think we can all sympathise with that

  5. Crikey, artificial consciousness, is that a machine which thinks it’s a human? I dunno AnotherDave if ‘Strong AI’ is 20 years away as the man said, then a self-conscious, self-aware machine must be 100 years away at least. One of the many good things about the HoL report, IMHO, was that it dissed West Coast bollox like The Singularity.

  6. Is it just me, or is there a rather disturbing conflation between the ideas of artificial intelligence and artificial consciousness when folk use the term ‘AI’?

    They are different things. One is with us. One will take a while yet…

  7. Absolutely Mike, I think that’s the lesson of this report that over and again the theoreticians have conceived grand schemes but, when reduced to practice, it’s found that the computing power isn’t there. As for quantum computing, it’s the answer to many grand schemes but the trouble is it doesn’t exist despite multi-decades of effort. So does humanity sit on its hands for, possibly, ever, waiting for what may be a chimaera, or does it try to find alternative ways to do it? Scientists need budgets, and budgets need a problem to solve, and the two are self-sustaining and eternal.

  8. Masterly Jamo. But surely it should be ‘Jamo’s Law of Bollox’

  9. I would like to propose a Mannerisms Bollox Law.

    “The elapsed time between each Latest Technological Hype will decrease by 33% between hypes. This will remain valid until the entire bollox world collapses in on itself.”

  10. Very droll Mike

  11. Well he did say the report is about If not How and When, so quantum computing is irrelevant, Mike. Also he
    points out that significant AI functions have been achieved in face recognition , fintech and search using traditional silicon logic. I see a Chinese face recognition camera picked a face out of a crowd of 60,000 at the weekend. Not bad. I agree that do it and worry about legal later is the Silicon Valley way but doing things in a transparent, decent, honest, British way, as the HoL suggests we do, is a rather more attractive way IMHO.

    • The face in the crowd thing is pure pattern recognition. Nothing AI about that. We’ve had the maths to do that since the 1980s, just not the processing power.

      But the report does indeed add to the definition of AI (P14, Para 11) that what they call AI is self-learning or self-adapting systems so the report is valid. And many other parts of the report are excellent as well.

      However I, and far more importantly numerous experts in this field, still assert that quantum computing is a fundamental pre-condition for true AI and anything else is merely imitating it, and it would have been useful if this had at least been noted in the report, including all the new areas this will open up eventually.

      The report was trying to make out that the UK is a world leader in AI and could compete with elsewhere. As shown by Cambridge Analytics this is indeed so, but I think from now on we will find researchers having to consider where they perform their research and far more importantly the commercialisation of their research, far more carefully. Places like South Korea or Silicon Valley aren’t strangled by regulation, and maybe only Ed’s view of a post-Brexit UK will change that. My impression of the report though is that they want more regulation, not less, almost to the point where you have to go to a government run ethics committee to decide if you can pursue your research or not before you even begin. Thus if I was a VC looking to fund an AI startup, I don’t think I’d be looking at the UK, at least not for a while.

  12. And equally unimpressive that the report saves to “100.pdf”. If they insist on it just having a number rather than a proper name, at least choose 42 !!

  13. Quite unimpressive that a report on Artificial Intelligence only mentions quantum computing twice, plus a single reference ! It’s now mostly accepted that the UK was correct in that real intelligence (the brain) is a quantum computer so how can one expect to achieve true artificial intelligence using traditional silicon logic ?

    Also comments such as legislation and and ‘AI Council’ indicate that the UK has not learned the lessons of Google, Facebook et al. You invent the technology first and make it essential, then worry about if it’s legal or ethical later. Your CEO might have a hard time in front of politicians but as he has files on every one of them, none can really turn the screw. You just have to look embarrassed and drink the water for a couple of days as Mark Z did so well 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *

*