LIFE 3.0: An Interesting Read on the Future of AI...
So I'm getting through WEB 3.0 by Max Tegmark which is a book about the possible futures of Artificial Intelligence, and once you get past the rather silly introduction it's actually quite a balanced read.
The general gist is that we are on the verge of major breakthroughs in Artificial Intelligence with the possibility of AIs being able to develop their own hardware within the next 100 years, at which point we will have self-improving AIs in the truest sense of the word.
It is this ability of technology to be able to develop its own hardware and thus self-improve that Tegmark defines as Life 3.0...
The Key Debates
The book covers a number of issues that are of concern to AI developers such as when we may have the first superintelligent AIs that basically do not NEED human input (within the next 100 years as a broad range), and then of course how broad or narrow we want those AI (probably broader than a chess AI, probably narrower than an AI programmed to think its' God).
Also there are issues of how we align the goals of the AI with ours as they are only as 'good' as the data we feed them, but of course a self-replicating AI may well be able to change its goals.
Basically there are lots of issues up for debate and Tegmark does a pretty good job of providing a balanced view somewhere between luddism and techno-utopianism.
Is AI a life form?
This was an interesting chapter. According to Tegmark if we define Life as anything with intelligence and intelligence is simply the ability to perform complex goals (self-admitted broad definitions) then yes because intelligence and thus life are just functions of:
- the ability to compute
- memory
- the ability to learn.
Which AIs can already do.
Hmmm, I somehow think this guy might be a humanist!
AI optimism or Pessimism...?
The intro chapter struck me as a bit silly - it's a fictional account of some AI researchers who initially programme an AI to do piece work on Amazon, then use that money to create another one which makes films, makes shed loads of money and then through various upgrades it takes over the stock market and eventually advises governments on policies resulting a highly unlikely AI-led libertarian world government in which everyone prospers.
Maybe that's one way AI can go, and maybe he wrote it to contrast against the Terminator style AI dystopia we are all more familiar with.
Either way it looks like we are not that far off self-replicating AIs which can produce their own hardware.
Although we probably won't see them having significant social influence in my life time, and frankly I think I'm glad of that!
