I guess I'm trying to understand. I'm hearing this paper has been around for a year -- I would think that many companies would have already implemented and measured its performance in production by now... is that not the case?
12,479 karma · joined April 18, 2014
I think the concept that you have a process for peer-review of architectural decisions open to anybody is important -- whether it's in writing, whatever you call it, where it goes I'm not sure those things matter.
After a certain number of years I begin to wonder if architect type roles love ADRs so much because their impact is hard to measure/validate, so they overcompensate or attempt to put a more verbose process on others to try to produce documents for optics reasons.
Just for illustration -- when I worked at Zynga they'd sell in-game purchases for over $10k USD for virtual items because there were people who just couldn't help themselves, and those "whales" were actually the bulk of the profit of the company.
That's why these platforms should have mandatory sane limits that can only be exempted with special circumstances.
It'd actually be quite easy to set certain sane limits on gambling like you can't gamble more than 1% of your annual income per year, but I bet gambling platforms would fight that like the plague because those are their whales, the true addicts.
Sometimes I feel that way too, but I'd love to hear more on your theory of this.
Cocaine is a party drug afaik, I'll give you that one.
Hallucinogens have the ability change people's perspectives on the world, often for the positive. Now the current psychiatry lens is you can only have "medicine" if you have a "disorder" but that doesn't really seem to make sense. Why can't you take medicine to move something from average to above-average? So I agree it's not prescribed for treating many disorders, but that doesn't mean it isn't therapeutic.
Really? I think the opposite is true.
Given 1/6 adults admit to using it, I think it's totally underrepresented in media -- in theory 1/6 characters would be using it.
It's only very recently that I see characters who just casually say something like "I take a gummy for a long flight" or whatever, rather than be a stereotypical comedic stoner character.
I feel like really it's alcohol that's glamorized in the media, and before that it was smoking.
No you aren't
You said "The paper says there's no evidence of effectiveness in treatment, and evidence of harm." which is a completely inaccurate summary.
- Expressed hesitation on whether the human should survive without being moved into computers [1]
- That Greta Thunburg could be the antichrist and cause the end of humanity [2]
- (Leaked) Apparently he has also called Pope Leo the antichrist [3]
1. https://www.youtube.com/watch?v=YSp07P8jvYs
2. https://www.youtube.com/shorts/Ao_umPlSV6o
3. https://www.thelettersfromleo.com/p/new-jd-vances-top-donor-...
That's not really a reasonable argument, because Thiel hasn't had the power of the Vatican (especially the power the vatican used to have), but what he's done with his power so far is much more concerning to me that what the vatican has done in the last 4 years, yes.
I think we both agree that the catholic church has received an unwarranted elevation and presumption of beneficence in media, but the distinction I'm drawing is that a billionaire who's toiling in American politics and claiming Greta Thunburg could be the antichrist is actively concerning.
The paper says "insufficient data" for helpfulness for most positive categories (but leans more positive than negative just doesn't reach 95% confidence), but also insufficient data on most negative categories. It finds 5 conditions it's helpful for, and 3 it hurts for.
Just to clarify, the study is not saying that.
The study is saying "there isn't conclusive evidence at this point, but it leans more toward helping than hurting on many categories". Please read the paper if in doubt.
Uhhh... no? Did you even read it? This research actually found more benefits than harms. I see it only identified two harms both graded very low.
Let's just quote here the researcher's own conclusions:
"Interpretation There was some evidence that cannabinoids can reduce symptoms of cannabis use disorder, insomnia, tic or Tourette’s syndrome, and autism spectrum disorder, but the quality of this evidence was generally low. Cannabinoids were associated with a greater risk of any adverse events but not of serious adverse events. Overall, there is a crucial need for more high-quality research. Given the scarcity of evidence, the routine use of cannabinoids for the treatment of mental disorders and SUDs is currently rarely justified"
>>> "The error bars are too big to say it works, so we shouldn't tell people it works"
I can see you didn't really understand my comment. There's a huge difference between not saying something is proven to work, and saying it's proven not to work. This study falls in the former category, by the authors own words.
Okay, I've read the meta-study now and I think the summary article isn't representing the picture very well. In particular they found for anxiety there actually seems to be evidence in this exact data set that does help.
What they are doing is saying "there isn't 95% evidence it reduces anxiety" therefore "no evidence" even though they mean "some evidence, just not at the statistical significance level" -- it's one of the biggest confusions (and sometimes it feels deliberate) you'll see people do.
Also when you have a confidence interval that big it's a red flag. They themselves admit the data is all over the place.
In summary, don't assume much from the title of the summary article.
Like for writing a best-seller, these AIs have so many advantages in that they've read every notable work ever, so if they can't craft something impressive and creative after all that then it's really indicative that they are actually quite below human on the creativity/writing side but just masking it on the massive-data-side.
Or put another way -- it's not really AGI until there is a model can learn at human speeds, no amount of being pre-trained on specific problem sets (e.g. human emotions, coding, math theorems, etc) will close that gap.
What does "making a framework" even mean, it feels like a nothing post.
When I think of what real AGI would be I think:
- Passes the turing test
- Writes a New York Times Bestseller without revealing it was written by AI
- Writes journal articles that pass peer review
- Wins a Nobel Prize
- Writes a successful comedy routine
- Creates a new invention
And no, nobody is going to make an automated kaggle benchmark to verify these. Which is fine, because an LLM will never be AGI. An LLM can't even learn mid-conversation.
There was a time when software was closer to a freer market and there was room for anyone to be the next google. But this moment is deeply political, tactical, and the existing incumbents have so much power that learning the intricacies of their systems may be make or break for any of your goals (e.g. how to advertise on google, or get featured on the app store, how to secure a government contract, etc). It isn't a time for abstract generalities (like chasing "golden ages")
At least I don't think anybody is or should be asking how to build "great" stuff anymore, great hasn't won in many categories. Most things that were great were so only to gain a moat and then enshittify.
I'd be really curious if he's happy where the tech-world is now, if he sees it as an oligopoly, if he sees it as meritocratic, and if he sees any hope of it getting better.
So currently we're at 428 with 3.5 increase per year, yeah, that's scary if it doesn't slow down soon. Makes you wonder about what indirect health side-effects that could have on us.
I just don’t want them to succeed anymore, and I don’t think there’s really any world where they regain my trust.
And to be clear I don’t expect my actions do make a difference, I just would feel dirty now that they have gotten into bed with this administration. Plus I should probably assume I’d have zero privacy now too…
Because as far as I know, Anthropic is taking the most moral stance of any AI company.