Addendum to Singular Limits
Never ending ai commentary
Only a day ago I shared my thoughts on the recent flurry of LLM artificial intelligence models. Today I was watching yet another AI geek spazz out on the hopes for genuine sentience in these machine models (see the clip below). These AI shill enthusiasts seem endless. My comment:
@4:20 word frequency analysis discounts not only semantics but syntactic grammar too. Currently there is no evidence the bots have any semantics. Syntactic grammar can go an awful long way, which is to say zombie behavioural models can go a long way, all the way as approximants to any desired human behaviour. But if human behaviour is backed by a hypothetical Singular Limit of subjective consciousness, then there is no guarantee at all that behavioural models can get to human level GI or HGI. They could get "stuck" at AGI. Which would not be a bad thing. AGI is already far more powerful than HGI in numerical intensive compute areas, and has been since the 1940's. Recognising this is to also recognise if you pull the socket or solar plug on an AGI machine and some ai nerd comes along and scolds you for killing a soul, you can slap him hard for me, send me your legal bill.
I should note, the blog I wrote is only focusing on the idea that sentience could be a singular limit that is unapproachable for any computation. There are plenty of other reasons to think machines can never be subjectively conscious. But my point is we should push machine “intelligence” (behavioural) to this limit, because if we are all ethical and spiritual beings we can simply (doing a lot of work that word) choose to not use the technology to “blow ourselves up” metaphorically. Collective wisdom and action is needed, but as Yannic Kilcher points out (here), a lot of the fear and panic today is misplaced.
Smart machines are a productivity gain story for humanity, not an unemployment story, not a dystopia where people all get lazy or mad e redundant. But my dudes and dudesses, you really should comprehend MMT macroeconomics to understand why I can say this. Unemployment is an unspent income story and so has nothing to do with automation. (Unless the neoliberal and Austrian School/ Ayn Rand f*kers in policy circles make it that way by choice.)
All right, that’s it for now. Although I am sure I will feel compelled to write furtehr addenda to the AI singularity story. So stay tuned