It seems to be de rigeur for most news stories in 2019 to have some sort of angle on Artificial Intelligence ( AI ), neural nets or algorithms. Usually for the technical press this is a good thing and finds exoplanets or learns to drive better than us. For the the non-technical press the outlook isn’t as rosy with, for example, Facebook’s AI failing to detect uploads of the New Zealand mosque shooting video despite having having been flagged several times or totally confusing your Tesla.
Up until now I’d largely ignored the subject but when I spotted the book “The Creativity Code” in my local library I decided to bring myself more up to date.
One of the problems of terms like “Artificial Intelligence” is that they are suitably vague and can mean anything from a simple database lookup all the way up to Cyberdyne Systems and the T-800 ( right ). Personally my definition comes more from the Hitch Hikers Guide to the Galaxy and its description of Deep Thought – ” … a stupendous super computer which was so amazingly intelligent that even before the data banks had been connected up it had started from I think therefore I am and got as far as the existence of rice pudding and income tax before anyone managed to turn it off.”
Marcus Du Sautoy takes a more practical approach and, early on, uses Margaret Boden’s definitions of human creativity:
- Exploratory creativity – taking what is already there and exploring its outer edges, extending the limits of what is possible while remaining bound by the rules.
- Combination creativity – take two completely different constructs and seek to combine them.
- Transformational creativity – complete game changers … like phase changes, when water suddenly goes from a liquid to a gas.
To me the first two lean more towards automation of existing processes rather than actual real intelligence which only happens at the third level.
The book starts with a description of game playing software. Deep Blue is well known for beating human chess grand masters but the game of Go was considered more of a challenge for computers. The description of Demis Hassibis’ AlphaGo software and its predecessors that learned how to play video games like Breakout and Space Invaders is absolutely fascinating. In beating the Go world champion AlphaGo managed to achieve the transformational level of creativity above.
Needless to say DeepMind, the company that was originally behind the software, is now part of the Google empire 🙂
The remainder of the book is arranged around discussions of possible AI applications to various fields such as mathematics, music, painting and literature. Personally I feel that the artistic examples provided lean more towards the automation side rather than true creativity. It should be noted that Mozart created a set of musical phrases that could be arranged in different orders by the roll of a dice so there’s at least some pattern repetition going on even in the mind of genius. However the descriptions of machines writing music in the style of Bach or painting in the style of Jackson Pollock are fascinating.
Being a mathematician then Du Sautoy spends some time discussing the application of AI and computers in general to mathematics. As far as I can see there are three basic levels:
- Pure number crunching – the Four Colour Map Problem was eventually solved with the aid of computer analysis
- Proof checking – mathematical theorem proofs can now stretch to hundreds or thousands of pages so some form of formal notation would be advantageous
- New theorems – where the AI either solves existing problems or creates completely new work
Fortunately, for Marcus Du Sautoy’s continuing employment, whereas the first two seem to be well accepted, the third has some way to go yet!
Probably because I read too much science fiction I have wondered whether it would be easier to approach the AI problem from the other direction. In other words, rather than teaching computers to think, we should try to enhance human cognitive capabilities ( a real life Dune mentat? ). There’s a small section of the book that touches on this subject when Oxford University was trying to recruit the mathematician Vladimir Voevodsky. One of his conditions was that the university obtain some Class B drugs to prove its credentials as Voevodsky felt that they helped his creativity! Sadly we never learn the full outcome of that particular interview but I don’t think that Voevodsky ever worked at Oxford. The short novel “Diamond Dogs” by Alistair Reynolds is based around this concept and is a very good read.
Overall this was a very interesting book and I thoroughly enjoyed it. Marcus Du Sautoy writes very well and any technical explanations are done very clearly and concisely. Although I’m not completely convinced that everything in the book is about my definition of AI I think that it’s a great introduction to the subject and one which provides many pointers and avenues for further investigation.
The Retired Engineer 