theptip

33 ms·

571.
by theptip 1y ago
> ewaste is out of control What’s the specific harm that you are worried about?
572.
by theptip 1y ago
It’s all relative. While there is skill, pretty much anyone can learn it in short order. Vs “skilled labor” like trades which require many years of apprenticing. And no, it’s not a proxy for “physical labor”. Trades are highly skilled, pick
573.
by theptip 1y ago
I worry about this stuff a fair bit, of all the net-negative AI outcomes I think perhaps the most likely group is some variant of this combined with dystopian techno-feudalism that drives society into an attractor state that’s hard to get o
574.
by theptip 1y ago
You’re right that you don’t have guaranteed recourse for foreign fraud. But at least the banks play whack-a-mole and make it somewhat expensive for criminals to set up accounts; crypto undoubtedly makes money laundering substantially easier
575.
by theptip 1y ago
> Let's pick one of the best possible examples of AI-generated code changes. An example so good, that the Cursor Editor uses it to advertise on their front page. I can see where the author is coming from, but “marketing image is dum
576.
by theptip 1y ago
“Who cares” is one way to frame it, perhaps a bit blame-y. I view this as more epistemically significant, it’s the Post-Truth era. Politics and national discourse has been heading this way for some time, but now AI may mean that nobody can
577.
by theptip 1y ago
A band saw is more dangerous than a spoon.
578.
by theptip 1y ago
Evolution in action. This is what the free market is good for.
579.
by theptip 1y ago
I wonder if we need some new affordances to help with these sorts of issue. While folks want a single uber-agent, can we make things better with partitioned sub-agents? Eg "hire" a DevRel agent to handle all your public facing int
580.
by theptip 1y ago
This is a great case study. I wonder how hard o3 would find it to build a minimal repro for these vulns? This would of course make it easier to identify true positives and discard false positives. This is I suppose an area where the enginee
581.
by theptip 1y ago
It’s a good point. For epistemic hygiene I think it’s critical to actually have models of the growth rate and what is implied. Eg we are seeing exponential growth on many capability metrics (some with doubling-times of 7 months), but haven’
582.
by theptip 1y ago
Thanks! So, how does this impact Deliberative Alignment[1], where IIUC the intermediate tokens are assessed (eg for referencing the appropriate policy fragment)? Does you see your result as putting that paradigm in question, or does the exp
583.
by theptip 1y ago
So is the interpretation here something like “CoT tokens are actually neuraleese”? They do boost performance, so the model must be stashing some intermediate reasoning outputs there. But perhaps not using the literal human meaning of those
584.
by theptip 1y ago
It’s a maturity/seniority/team cohesion thing. Junior devs might stay blocked for a week without a nudge in some pathological cases. A new team that hasn’t gelled might do the same. A seasoned team full of staff engineers might no
585.
by theptip 1y ago
I think “going full generalist” is perhaps not a generalizable solution, though directionally it’s good advice if you have more than two dev roles (or maybe two plus Q) in a 10-person team. No mention of task composition, but I wonder if fr
586.
by theptip 1y ago
“Human can’t fly, much like other humans. Therefore it’s bad” Spot the problem now? AI capabilities are highly jagged, they are clearly superhuman in many dimensions, and laughably bad compared to humans in others.
587.
by theptip 1y ago
No, it’s never going to be precise but it’s important to have a good rough definition. Can we just use Morris et al and move on with our lives? Position: Levels of AGI for Operationalizing Progress on the Path to AGI: https://arx
588.
by theptip 1y ago
Great example of this general class of reasoning failure. “AI does badly on my test therefore it’s bad”. The correct question to ask is, of course, what is it good at? (For bonus points, think in terms of $/task rather than simply bein
589.
by theptip 1y ago
Can someone help with the intuition here? My understanding from vision transformers is you start with noise and use a series of hierarchical models to iteratively refine the noise into the target. Each layer is trained to produce images at
590.
by theptip 1y ago
I somewhat doubt that YT cares much about AI content being uploaded, as long as it’s clearly marked as such. What they do care about is their training set getting tainted, so I imagine they will push quite hard to have some mechanism to det
591.
by theptip 1y ago
Remarkable how quickly they were able to break that historical invariant. Meanwhile Fox News posted that akshually Trump had a cunning plan to lower interest rates by tanking the stock market temporarily so that the bond refinance this year
592.
by theptip 1y ago
Thing is, Musk actually did find some savings. It’s just the admin plans to pass massive unfunded tax cuts that will make those savings a drop in the ocean. It’s almost like destroying state capacity for “woke” causes like foreign aid and s
593.
by theptip 1y ago
Meta seems to place a pretty high premium on controlling its open source projects, especially dev tooling. I guess dating back at least to the git maintainers telling them they were doing things wrong with their monorepo and refusing to ups
594.
by theptip 1y ago
Exactly. Whatever your opinions on crypto, it should not be controversial that black market transactions are a perfect fit.
595.
The only way to go fast, is to go well(factorio.com)
2 points by theptip 1y ago|1 comments
596.
by theptip 1y ago
This was the joke when I did Physics in the ‘00s too. The hilarious part is that the kids these days misquote it as “fusion is 20 years away and always has been”, or even 10 years. I read that as at least 0.5 years per year of progress agai
597.
by theptip 1y ago
> In the best case, killing these will force consumers to move to Apple. I think there are better options than you suggest. The whole doc suite can be had from Microsoft. Maybe folks need to ten bucks a month for something that used to b
598.
by theptip 1y ago
You’re getting a lot of advice here, I’d just echo the sentiment of trying to give yourself permission to have fun. In my experience the best way to learn is by doing; that “uncomfortable having gaps” is there for most folks to some degree.
599.
Testing AI's GeoGuessr Genius(astralcodexten.com)
1 points by theptip 1y ago|0 comments
600.
by theptip 1y ago
> A fundamental problem with this entire class of machine learning is that it is based on a model / simulation of reality… all such simulations that attempt to model any non-trivial system are imperfect Depends on what your goal is.

More ›