1. More Leopold on alignment. Leopold continues to be good on this subject.
2. I feel gated, however right here is Scott Alexander on my AI argument. I’m a giant fan of Scott’s, however this can be a gross misrepresentation of what I wrote. Scott ignores my important level that that is all taking place anyway (he ought to discuss extra to folks in DC), doesn’t have interaction with the notion of historic reasoning (there may be solely a slim conception of rationalism in his put up), doesn’t contemplate Hayek and the class of Knightian uncertainty, and doesn’t contemplate the all-critical China argument, amongst different factors. Or how in regards to the notion that we will’t repair for extra security till we see extra of the progress? Or the unfavorable bias in long-worded, rationalist therapies of this subject? Plus his restatement of my argument is just not what I wrote. Sorry Scott! There are many arguments you simply can’t put into the classes outlined in LessWrong posts.
This may occasionally sound a bit of harsh, however the rationality group, EA motion, and the AGI arguers all have to radically increase the sorts of arguments they can course of and take care of. By rather a lot. Some of the placing options of the “six-month Pause” plea was how intellectually restricted and non-diverse — throughout fields — the signers had been. The place had been the clergy, the politicians, the historians, and so forth? This ought to be a wake-up name, however to this point it has not been. As an alternative, this is the kind of arrogance we see. Precisely like the general public well being authorities through the pandemic who thought that they had “the experience,” however they had been weak of their artificial skills for understanding social processes and the way the entire image suits collectively.
Virtually as a rule, you’ll discover the best weak point (and least actual engaged curiosity) within the Doomer arguments when factual issues are up for grabs, resembling whether or not there’s a solution to flip again (GPTs are super-popular client merchandise with low marginal prices, and plenty of worthwhile enterprise and navy functions, and in contrast to cloning regular folks don’t discover them gross…they exist already, throughout a number of establishments, and sure the regulatory state is obstructive however no the CPSC isn’t going to ban them, sorry! For higher or worse, there may be remarkably little panic about AGI in DC, and that will not change if all of them learn all these LessWrong weblog posts. That’s merely not how our world works, and moreover I feel it’s superb if I toss that time out in a single remark slightly than going by a type of prolonged circumlocutions.). One other instance of Doomer hand-waving is on the China query. The Chips Act is one method, however it’s unlikely to alter the medium-term trajectory of what China can do, and in some methods it might speed up it. If something, it raises tensions and boosts the case for America extending its AI lead. To not point out there are different nations and establishments in addition to China, and the scale-up prices aren’t clearly so massive any extra. How about Open Source, for that matter? Horse, barn door — stay with it! There simply aren’t any many-step summary arguments which are going to undo that actuality. In order that ought to be the start line for all the remainder of the dialogue.
3. The following step: the call for violence, imprisonment, etc. Not shocking, however you may not less than say that Eliezer is consistent. You actually do have to take initially a historic perspective on this name for violence on the premise of a fairly summary, not typically accepted (to say the least) argument. Are airstrikes on “rogue information facilities” actually going to decrease existential danger? I take it these are throughout borders as properly and would cowl rogue information facilities in Beijing too? How about Tel Aviv? I’m joyful to see this placed on the desk, nonetheless, and I hope it snaps some onlookers again to their senses. If that is the place your argument constantly leads, maybe the tactic and premises require a slightly radical reexamination. Or not less than that’s what historic modes of reasoning would are inclined to counsel.
Leave a Reply