kalkin

11 ms·

151.
by kalkin 2y ago
By reputation -- I can't vouch for this personally, and I don't know if it'll still be true with this update -- Opus is still often better for things like creative writing and conversations about emotional or political topics
152.
by kalkin 2y ago
In context the thing about a purple-haired person is about a hypothetical Twitch trust-and-safety employee, not about Netanyahu.
153.
by kalkin 2y ago
The Twitter account posting this seems a bit unhinged, with the snitch tagging of everyone from th ADL to the ACLU and the weird speculation about a "purple haired" person. I'm not inclined to take their description of the sc
154.
by kalkin 2y ago
I think this is because logged-out Twitter now shows top Tweets of all time from a user, rather than most recent Tweets. Good reason why companies shouldn't be using Twitter/X for status updates anymore!
155.
by kalkin 2y ago
Zed ( https://zed.dev/ ) does pretty well at IDE LLM integration. Haven't done anything extremely ambitious with it but the editing experience is much smoother than copy pasting from a browser.
156.
by kalkin 2y ago
There's definitely something interesting to be learned from the examples here - it's valuable work in that sense - but "VLMs are blind" isn't it. That's just clickbait.
157.
by kalkin 2y ago
The comic at the beginning paints the "stack more layers" LLM people as clowns, not neurosymbolic people or by proxy Chollet. Yes, it suggests the "stack more layers" approach works anyway, but in a self-deprecating way.
158.
by kalkin 2y ago
The question the paper claims to address isn't whether current LLMs are AGI, the question is whether LLMs can reason. And solving a problem that seems to require reasoning some of the time is still evidence of reasoning - especially if
159.
by kalkin 2y ago
> People are being told it can do far more than it really is. Meanwhile these HN comments are split between: * Lots of people confirming what the paper itself notes (but doesn't highlight), that the most advanced models actually can
160.
by kalkin 2y ago
I wouldn't be surprised if anti-gender-bias training (which, to be clear, is a good thing to do IMO) does some cognitive damage to these particular formulations because questions about "sister" and "brother" get ass
161.
by kalkin 2y ago
> Paper finds a failure case for current LLMs and draws broad conclusions about the fundamental limits of LLM technology > Evaluates the largest LLMs and finds evidence that actually scale overcomes the problem: "Notable exceptio
162.
by kalkin 2y ago
This is an illustrative comment for meta reasons, I think. Karpathy's lecture almost certainly doesn't cover the superposition hypothesis (which hadn't been invented for ANNs 8 years ago), or sparse dictionary learning (whose
163.
by kalkin 2y ago
This line of argument indicates a basic refusal to take the threat model seriously, I think. Should Google worry about Chinese state-backed attackers using attacking its systems to target dissidents or for corporate or military espionage? &
164.
by kalkin 3y ago
This talks about higher interest rates due to investors being reluctant to purchase US government debt, which is not something we've observed recently. It doesn't talk about higher interest rates due to deliberate action by the Fe
165.
by kalkin 3y ago
Interesting that the article doesn't mention what seems to be an obvious takeaway if you believe that interest rates explain the gap between CPI and sentiment: raising interest rates to fight inflation will make people feel inflation i
166.
by kalkin 3y ago
This isn't correct. What you may be remembering is that some (not all) NP complete problems have limits on how accurately they can be approximated (unless P = NP). But approximation algorithms for NP complete problems form a whole subf
167.
by kalkin 3y ago
Just ran across this useful comparison with another very recent paper that effectively corroborates some of the core findings, I believe by an author of the other paper: https://www.lesswrong.com/posts/F4iogK5xdNd7jDNyw
168.
by kalkin 3y ago
"Money comes way cheaper outside of the big cloud providers" Say more about it this? It's not obvious to me why it should be true or what evidence there is for it.
169.
by kalkin 3y ago
I think dismissing problem-solving as "smoke and mirrors" based on regurgitating training data will give you a poor predictive model for what else models can do. For example, do you think that if you change the variable names to s
170.
by kalkin 3y ago
"Surfaces patterns in the training data" seems not to pin things down very much. You could describe "doing math" as a pattern in the training data, or really anything a human might learn from reading the same text. I sus
171.
by kalkin 3y ago
What do you mean by the "stochastic parrots" (null) hypothesis in this case? Cards on the table, I think by any reasonable interpretation it's either uninformative or pretty conclusively refuted, but I'm curious what you
172.
by kalkin 3y ago
Can you make a novel sound? One that's not part of any human language? Perhaps you can, using a tool. However, if we're allowing tools, I bet GPT4 could also write a program that would produce a novel token, by whatever definition
173.
by kalkin 3y ago
I was talking about competitive policy debate, which I think I was pretty explicit about. I am unaware of a case of anyone showing up to one of those with a baseball bat, and I don't know where you're getting the "suppression
174.
by kalkin 3y ago
The article says a lot of things, some of which are even true. Debate judges are volunteers and there are tens or hundreds of thousands of people who've done it. I don't doubt there are judges who make bad decisions, but that ideo
175.
by kalkin 3y ago
It's funny, I have mixed feelings about my policy debate experience but the sort of tendentious criticism in the article makes me want to defend it. (I want to call the article ill-informed but I suspect the author is actually at least
176.
by kalkin 3y ago
The distinction between derailing and "understanding the values and otherwise" is contestable and in fact routinely contested by debaters - and there is of course no de jure or de facto rule that the person who goes further in try
177.
by kalkin 3y ago
Good thing nobody gets an instant-win button! (That would actually kill the activity which is why even this rather poorly-argued article doesn't claim it's happening.)
178.
by kalkin 3y ago
There is no ground rule that "forces you" to implicate yourself as person as committed to the positions you take as debater. But there's also no ground rule that prevents your opponent from trying to implicate you in that way
179.
by kalkin 3y ago
> I think this an attempt to champion the idea of rhetoric as a virtue, in the face of arguments made in bad faith. I'm legitimately not sure whether by "this is an attempt" you're referring to critiques in debate or
180.
by kalkin 3y ago
The way policy debate works (traditionally) is that the affirmative side gets to choose a particular policy to advocate within a broad space. A negative side that sticks to refuting the particular details of that policy is putting themselve

More ›