keeda

63 ms·

901.
by keeda 1y ago
What Jensen giveth, Altman taketh away.
902.
by keeda 1y ago
I've observed the same in NYC. I experienced the nightlife in NYC shortly before the pandemic, and then again a couple of years after, and the difference was stark. The streets used to be buzzing at 3am on any given weekend. Nowadays,
903.
by keeda 1y ago
I've seen countless examples where it does do a lot of logical reasoning from a very high-level requirement. I only gave that example because that is the more common form of programming logic required, but there is a huge range of logi
904.
by keeda 1y ago
> It's very good at some kinds of reasoning and very bad at others. There's not much it's mediocre at. Yes, one of the terms for this phenomenon is "jagged frontier" from here: https://www.oneusefult
905.
by keeda 1y ago
That may be the case, but how often do you have to provide these kinds of puzzles as programming tasks? And even then, by your own example, it did product correct code, so it did "reason" through the problem statement!
906.
by keeda 1y ago
People themselves don't know when they are wrong, and that is why high-functioning organizations have all sorts of guardrails in place. Trivial example, code reviews. Now, code reviews are multi-purpose, and their primary benefit is no
907.
by keeda 1y ago
Maybe a related point is that hangovers, of which headaches are likely the most common symptom, are caused in a large part by dehydration as well as electrolyte imbalance.
908.
by keeda 1y ago
> ... LLMs aren't very good at logical reasonsing. I'm curious about what experiences led you to that conclusion. IME, LLMs are very good at the type of logical reasoning required for most programming tasks. E.g. I only have
909.
by keeda 1y ago
There are more studies out there, but here are a couple I know of offhand, showing a 25% to 55% boost. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4945566 https://arxiv.org/abs/2302.06590
910.
by keeda 1y ago
That whole bit about color blending and transparency and LLMs "not knowing colors" is hard to believe. I am literally using LLMs every day to write image-processing and computer vision code using OpenCV. It seamlessly reasons acro
911.
by keeda 1y ago
In my experience, it's not just prompting that needs to be figured out, it's a whole new workstyle that works for you, your technologies and even your current project. As an example, I write almost all my code functional-programmi
912.
by keeda 1y ago
You're missing a third option, which is actually closer to the role of managing coding agents: being a "senior engineer / architect / what-have-you". IME the more senior engineering roles (staff, principal, fellow,
913.
by keeda 1y ago
Maybe the dissonance could be explained by being in different states? I can totally imagine social workers in some localities imposing their personal or religious beliefs on people needing help...
914.
by keeda 1y ago
This seems to be the report and the specific section of interest: https://www.bankrate.com/banking/savings/emergency-savings-r... You're right that the question they used is a bit vague, but there is a ton of
915.
by keeda 1y ago
Yes, self-reporting has biases and estimating tasks is still a fool's errand, which is why I noted that the estimates from these surveys matched the findings from other RTC studies. However, what doesn't get discussed enough about
916.
by keeda 1y ago
Good question, I don't believe they break out their workloads into training versus inference, in fact they don't even break out any mumbers in any useful detail. But anecdotally the public clouds did seem to be most GPU-constraine
917.
by keeda 1y ago
The quote is from the paper linked in the article: https://s3.amazonaws.com/real.stlouisfed.org/wp/2024/2024-02... It's actually based on the results of three surveys conducted by two different parties.
918.
by keeda 1y ago
That study gets mentioned all the time, somehow this one and many of the others it cites don't get much airtime: https://www.stlouisfed.org/on-the-economy/2024/sep/rapid-ado... >This implies that eac
919.
by keeda 1y ago
I posted a comment yesterday regarding this with links to a couple relevant studies: https://news.ycombinator.com/item?id=44793392 -- briefly: * Even with all this infra buildout all the hyperscalers are constantly capacity
920.
by keeda 1y ago
From your first link: > Those donation totals may be explained, at least in part, by the former president’s policies related to those agencies. Trump has repeatedly vowed to eliminate Education if he is elected. He maintained an adversa
921.
by keeda 1y ago
If you look at all the hyperscalers' earnings reports, they all report their data centers being capacity constrained, especially for GPUs. Then couple this with survey-based reports [1, 2] showing that so far a minority of people used
922.
by keeda 1y ago
Yes, the simplistic "immigrants are competion" narratives always disregard the fact that immigrants also contribute to the economy and uplift it. For instance, there is a credible theory that the "immigration crisis" was
923.
by keeda 1y ago
That's where the timeliness versus accuracy trade-off comes in. There is significant value in having an early signal to act on, especially as long as there is awareness of its limitations. And as I mentioned, this data does come with v
924.
by keeda 1y ago
India's statistical institutions have been facing similar "challenges" for a while now. Here's a long article that covers multiple issues, but the relevant bit is that multiple people have "resigned" after repo
925.
by keeda 1y ago
Such an organization cannot exist. These agencies are always balancing two opposing forces, timeliness and accuracy. Data collection is inherently delayed (e.g. a lot of it is from surveys that businesses complete at their own speed, or fro
926.
by keeda 1y ago
You should look into the work of Nick Bloom, an Economics prof at Standford (active on various social media platforms) who has been researching remote work for quite a while: https://economics.stanford.edu/people/nichol
927.
by keeda 1y ago
I watched a hummingbird decimate a cloud of mosquitoes just a few weeks ago. It was fascinating; it hovered around them like an attack helicopter, and would dart in and out, then find a different angle, then dart in and out again. With ever
928.
by keeda 1y ago
+1 I would like to know too. Especially the experimental demo of infinite paths -- I'm a complete noob in quantum physics, and the video made sense of so many topics I "learned" in college but never managed to grok. It'd
929.
by keeda 1y ago
As someone only tangentially familiar with this domain, I have questions about this: > Look at their CPU density and do the math on power. It's fairly low density. Look at the interconnects (100gb per system). Also fairly conservat
930.
by keeda 1y ago
This is reminiscent of that 2024 Apple paper about how adding red herrings drastically reduced LLM accuracy. However, back then I had run a quick experiment of my own ( https://news.ycombinator.com/item?id=42150769 ) by simpl

More ›