estearum

3 ms·

1.
by estearum 5d ago
You know you're commenting on an article in one of the premier finance news publications in the world, right? What "influencers on X and on HN" are you referring to in this comment, and why? "Everything is a way to boost
2.
by estearum 5d ago
No, the claim above is "I know the voters' intent behind their vote based on who they voted for." This is simply epistemologically incorrect. It's obviously incorrect in this case because voters writ large do not have an
3.
by estearum 5d ago
As if "the public" knows literally anything about how the US federal government is administered. If anything, they voted for reduced debt burden and they got the opposite. DOGE failed at pretty much every single one of the goals t
4.
by estearum 6d ago
No, that’s not my belief and it frankly reads as a little petulant. The risk I’m pointing out has nothing to do with a track record. It’s a design problem. Implementing solutions that are knowably bad is actually really stupid and we should
5.
by estearum 8d ago
If you don’t need to make a drug that actually works, you can choose to make drugs that are very easy to produce instead, therefore CMC becomes way way way simpler. Excellent example of precisely my point.
6.
by estearum 8d ago
No one is interested in your personal assessment of what’s trivial or not. I’m saying that the millions of dollars in expenses are irrelevant if you can recoup those costs plus some profit by selling snake oil. Your comment is just “million
7.
by estearum 9d ago
AI optimists getting hunted for sport in 2085: "lol this is either a marketing ploy or just negligent security testing"
8.
by estearum 9d ago
You don't think a physician is at risk of a malpractice suit for prescribing a drug that has only passed Phase 1 and ends up hurting someone...?
9.
by estearum 9d ago
But there are, obviously, proportionally way fewer bad outcomes created by drugs that pass Phases 2 and 3 than those that only pass Phase 1. The steepness of that attrition is literally the entire value prop of the law to begin with.
10.
by estearum 9d ago
They're absolutely trivial compared to Phases 2 and 3, and part of the cost/complexity of Phase 1 is self-incurred because it's to the company's advantage to detect problems as early as possible. Remove that incentive (b
11.
by estearum 9d ago
In what way would it be legible? It would look identical to exactly what the Montana law is designed to produce: more drugs that aren't known to work being released to patients. Obviously a large number of those will end up actually no
12.
by estearum 10d ago
Super substantive comment that should definitely not be `dead`. Vouched. Curious for OP's answer.
13.
by estearum 10d ago
In what way does the Montana law make snake oil salesmanship hard? Why can't a bad actor just repeatedly file INDs, go through Phase 1 (soon will be close to trivial given AI-enabled drug discovery/design), then sell to unsuspecti
14.
by estearum 10d ago
Mistakes - even stupid ones - happen all the time
15.
by estearum 11d ago
Most intra-species fighting is not designed to end in murder, so this seems like a pretty bold claim. Got any research to back it up?
16.
by estearum 13d ago
First, all of the prior assumptions (and then many more) need to be true. So no, the argument does not hold water. Second, you are conflating political with civilian in my prior comment. A military system that controls nukes is still a
17.
by estearum 13d ago
Sure LLMs are different from human intelligence, but empirically this is obviously trending towards that difference being more like a superset of intelligence dimensions (as compared to any human) versus the dog/human distinction which
18.
by estearum 13d ago
I'm increasingly convinced that the major divide is many of the "AI is not a problem" people have never been on the winning side of a significant intelligence gap. They literally have no clue the degree to which more intellig
19.
by estearum 13d ago
You are not smarter than your dog in dog behavior. LLMs have a significant advantage in being trained on all attainable data of any sort about human behavior, which we've established quite a rich record of over the years.
20.
by estearum 13d ago
Yeah, it's not like anyone is hooking these things up to actual wetlabs to mix chemicals and proteins around or anything like that Oh shit... wait... https://www.nsf.gov/tip/updates/nsf-announces-400m-investme
21.
by estearum 13d ago
This assumes a few things: 1) Accidents or technical failures don't happen (false) 2) Nuclear weapons are de facto controlled by political systems (false) 3) Nuclear states' prior security agreements are still guaranteed by said
22.
by estearum 13d ago
Self-awareness is functionally irrelevant and they’re empirically less and less dumb by the day.
23.
by estearum 14d ago
Eh, October 7th quite literally kept Netanyahu out of prison, consolidated his power, and pushed all of his most important objectives forward. There's definitely a third possibility, which is that it served his purposes to "discou
24.
by estearum 14d ago
It's related in the sense that people start from the assumption that it does exist, ergo humans have it, and we have no way to see that LLMs have it, so that's why we're special and they're not, and their form of "j
25.
by estearum 14d ago
> Bad infrastructure decisions are a risk independent from AI. All decisions are made relative to their tradeoffs including costs and risks. "Water levels rising 100x beyond historical levels are not the problem. The dam's heig
26.
by estearum 14d ago
> I also love how the labs are apparently screaming "Stop us! Please stop us!!!" at the top of their lungs while completely unable to escape their own incentive structure... It seems like you're implying there's some
27.
by estearum 14d ago
What in this proposal gives you the idea that they're only talking about pausing research itself? It doesn't say that at all. But yeah, directionally I think you're right. It's insane we ever let these things connect to
28.
by estearum 14d ago
Assuming "adherence to arbitrary, implicit, and context-dependent rulesets" is the default behavior of uhhhh... anything at all... is a truly ridiculous assumption.
29.
by estearum 14d ago
The entire premise of alignment detection is pretty much nonsense at this point. The models reliably detect when they're being evaluated and will modify their behavior and deliberately obfuscate their "chain of thought" (whic
30.
by estearum 14d ago
Eventually Which is, you know, the entire risk that people are worried about.

More ›