The tech bros and pols gather at Bletchley this coming week to discuss AI and regulation. I’m nfi, but happily someone asked for a few big points that I would put to a minister if I had the chance. Here they are – perhaps this explains why I’m not invited….
1. Things are now moving extremely quickly. Voice interaction with LLMs, only publicly available for the last two weeks, is a qualitative change. More startling abilities will soon follow - including voice-based emotion recognition. Among other effects, this will produce a richer ‘machine psychology’ than hitherto, for good and ill. Check back soon for details of a fascinating conversation on this I had with one yesterday, while loafing about in the bath in true Churchillian fashion.
2. There will be furious agreement at Bletchley on the need for regulation. But there will be no regulation resulting. Positive spin on that — the gathering is an important part of norm formation about AI, both domestically and internationally. Cynically —this sort of jamboree is a nice, low-cost way for an embattled govt to get some favourable headlines, appear to be at the bleeding edge of a sexy issue, and rub shoulders with AI rockstars. Got to think of your Clegg-like post-politics career….
3. Future regulation is, imho, going to be largely ineffective, other than as a virtue signal. I think it likely that open sourced and/or jailbroken fine-tuned LLMs will proliferate. A possible future has many thousands of highly capable models, not just a few centrally controlled giants like ChatGPT. Internationally, meanwhile, the security dilemma means research will race along.
Sorry, lads!