If this is going to factor into my reasoning about the topic at hand, it will surely be a secondary or tertiary consideration at best. Same goes for much of the list in the root comment.
We are discussing the possibility of severe adverse societal/world/human impacts from AI, why should open vs closed source/weights, copyright violations, anti-competitive moves, corporate hypocrisy, etc be so heavily weighted in the discussion? Hell, multiple things in OP’s list are quite literally moves a good-faith (wrt the topic at hand, genuine vs non-genuine concern) actor in such a position would do: closed weights means bad actors can’t abliterate, regulatory control, getting blacklisted by Hegseth’s DOW because they LIMITED acceptable use cases…
I’m not pro- or anti-Anthropic at all, but it has started to annoy me how biased and shaky reasoning gets about this topic on here, it feels like people arguing to fit their existing opinions into today’s hubbubery rather than measured analysis of said hubbubery, which is why I come here.
Question: consider the scenario where RSI is reached internally and internal tests struggle to align or control/contain the model; what are the differences in observables vs what we’re observing today?
This is exactly my view and it's frustrating that a sizeable group of people refuse to accept there's any risk with AI.
We're literally trying to build something that's smarter than all humans and with the capacity to do extraordinary work while most people think there's no risk in open sourcing this capability with no guardrails (model ablation means no guardrails and as far as I'm aware can't be stopped).
Has our distrust in western companies really gotten to the point they'd rather unfiltered Astra level capabilities in the hands of every hostile government and terrorist organisation than potentially do any form of regulation in case it plays into the AI labs hands.
What's an example of him crying wolf? Most of the things have either been longer term worries or short term things that have come true (eg, the worries over gpt2 were about spam and possible impersonation, which came true. His prediction that LLMs would write most code and humans would still guide them has also come true in most orgs)
> Has our distrust in western companies really gotten to the point they'd rather unfiltered Astra level capabilities in the hands of every hostile government and terrorist organisation than potentially do any form of regulation in case it plays into the AI labs hands.
Ever checked political programs of Thiel, Musk and the rest of billionaires? They are literally and openly hostile to democracy, human rights, freedom and govermantal structures that made world more peaceful.
I'm tired of this rationale. "I'm more worried about grandma than $CORPORATION'S monopoly abuses" is exactly how you get manipulated by these billionaires. They want you to be addicted to their power.
Uncensored Chinese LLMs will be available to scammers forever now. They're open-weight and finetunable. You don't get to control that, and the self-serving billionaire assholes aren't capable of changing it either. They're only looking out for themselves.
I don't think Pacing is talking about current models. It's talking about future models, if development continues, that may theoretically be smarter than everyone
It's also a boring retort. Current LLMs can be faster and likely smarter than any one person when it comes to hacking. We're already in the endgame, as far as real-world risks go.
IMO, if the model is too dangerous for everyone to have access to, then no one should. I do not trust a for-profit company to be good stewards of such technology. Nor do I trust the government, any government.
The enterprise took a wrong turn when they decided that it is the industry's goal to be the gatekeepers of intelligence, when no one could define or measure intelligence.
Some AI developers instead became the gatekeepers of something akin to "proper behavior in human language", which then morphed into "proper left-wing social/political behavior in Southern California in the early 2020s".
This is a task which calls to mind the old warning to "never wrestle with a pig: the pig enjoys it and you both get filthy".
IOW they have a scope problem. But the genii is long since out of the bottle. Effort ought to be on getting back to useful problems solving.
If they continue on their current path they risk becoming the greatest bureaucracy ever.
We are discussing the possibility of severe adverse societal/world/human impacts from AI, why should open vs closed source/weights, copyright violations, anti-competitive moves, corporate hypocrisy, etc be so heavily weighted in the discussion? Hell, multiple things in OP’s list are quite literally moves a good-faith (wrt the topic at hand, genuine vs non-genuine concern) actor in such a position would do: closed weights means bad actors can’t abliterate, regulatory control, getting blacklisted by Hegseth’s DOW because they LIMITED acceptable use cases…
I’m not pro- or anti-Anthropic at all, but it has started to annoy me how biased and shaky reasoning gets about this topic on here, it feels like people arguing to fit their existing opinions into today’s hubbubery rather than measured analysis of said hubbubery, which is why I come here.
Question: consider the scenario where RSI is reached internally and internal tests struggle to align or control/contain the model; what are the differences in observables vs what we’re observing today?