Lafargue

A lot of the anti-AI discourse is making me realise that people truly don't want a post-scarcity society. There's a lot of posts out there saying things like: "A meaningful life is all about human striving for excellence - what will become of humanity if we simply hand such challenges over to AI?" And like, humanity can find meaning in all kinds of things. This is not a problem I take seriously at all.

Even with the best will in the world there are huge practical problems implementing communism with any actually scarce resources - but many forms of "information communism" are extremely easy to implement, because information is a pure public good - it's not diminished when shared. There are still major practical problems - information enclosure is one of the major ways in which information production is materially supported - but *many* of the problems are straightforwardly political.

parrots all the way down

I don't think it's at all obvious that Hume's basic model of cognition is wrong, though!

sometimes I wish I had time to pursue all these ideas!

@ajswritesthings@plasmatrap.com I would expect, just on general "innovation cycle" grounds, that as the technology consolidates you're going to get more focus on cost savings relative to 'frontier' innovation, and we'll see what that spits out. But, again, who knows where it's all headed (not me)

@ajswritesthings@plasmatrap.com yeah - you may well be right about personal computing - wdik lol. On how much they cost to run, I feel like it must be the case that quite a bit of the money these companies are spending is going into the race for AGI, rather than the business-as-usual operations of existing models?

Of course I would say this (given my politics and research), but I think the key issue in the 'AI moment' is open access: we (I) want open (non-proprietary, public weight) models accessing an open (not paywalled, not blocked to crawlers) web.

Relevant bit from the abstract: "Strikingly, left-leaning stances cause all five models to converge on Japan’s Communist Party, despite other parties holding similar positions on the relevant issues. We trace this to the information environment: JCP operates a fully open website with a party newspaper that models freely access and frequently misclassify as independent journalism, while major Japanese news outlets block AI crawlers."

another win for open data 👍👍 https://andrewbenjaminhall.com/pdfs/Miyazaki_Hall_Japan_AI.pdf

@ajswritesthings@plasmatrap.com we've basically got two "actually existing" institutions that can channel this kind of capital concentration - firms run by wildly wealthy capitalists, or states. Obviously both of them have significant likely problems! I myself don't really have a better solution that "let's try to make sure the states are run by nice principled leftists, with some institutional constraints on despotism" tbh, but I do think it's a general problem rather than one specific to this tech, iykwim

@ajswritesthings@plasmatrap.com 2) I sort of feel like, although the technologies that are at the centre of attention change over time, the need for massive capital aggregation to achieve *some* collective goal (railways, or dams, or power, or whatever) is sort of a constant feature of technological modernity - so the same governance problem keeps showing up in different forms.

like - the Humean model of cognition as "association of ideas" is basically what Bender et al are critically characterising as the reason why LLMs can't really be exhibiting cognitive faculties

@ajswritesthings@plasmatrap.com Well you need material resources to run any of these things - the internet too, which has always obviously required a vast physical infrastructure. Capitalists will always aim to rent seek - I see that as something close to axiomatic lol - but I don't know that it's more of a problem with this tech specifically than with previous technologies?

@ajswritesthings@plasmatrap.com this is obviously all pretty half-baked on my part, but I guess my vague view is: 1) the trend across my lifetime has been steady and dramatic increase in personal computing power, so it wouldn't be astonishing if stuff that is currently only feasible for organisations with access to eye-watering levels of capital becomes more broadly accessible (though of course there'll presumably be new innovations that require similar massive capital aggregation!); and