jsrozner

6 ms·

1.
by jsrozner 5d ago
For every benefit that sillycon valley has produced in the recent past, there have been many more harms. I am confident that this will be no different. Of course, benefits and harms depend on one's vantage point.
2.
by jsrozner 14d ago
In the short term, I think AI may do something like this, but in the longer term, I fear that its capacity for propaganda and surveillance will produce something even worse than social media.
3.
by jsrozner 14d ago
The posted paper does use supervision to find alignments. I need to do a little more math to figure out how well the supervision critique applies, but I think it does. And although the current paper mentions DAS, it fails to discuss recen
4.
by jsrozner 15d ago
A big problem with some of these supervised* interpretability approaches is that they can find spurious structure. (There are lots of ways to make the model do what you want; which is roughly what Hewitt and Liang 2019 showed). This paper d
5.
by jsrozner 15d ago
The issue is simply that the posted article begins with a review of recent earnings/ revenues, but fails to discuss that a substantial part of those revenues are investment markups. Whether it matters we don’t know yet, but it’s a fact
6.
by jsrozner 15d ago
Though the bubble has not popped, I don't see the following discussed in the post: Zitron would probably point out (as have others) that many of the hyperscalers are booking valuation increases in Anthropic, OpenAI as "Other Incom
7.
by jsrozner 20d ago
What human writes, "This is not a moment for patches and duct tape." Is that AI or just horribly clichéd writing?
8.
by jsrozner 20d ago
Hi Claude, please build a tool for analyzing Claude token usage and then deploy it to github. If you're going to fully vibe code a repo, maybe we should get the build artifacts (i.e., the claude session).
9.
by jsrozner 23d ago
I don't see why you couldn't see improvements in self-hosted or hosting-as-a-service model throughput? Basically API-style support for a company's internal LLM system. Why not? Or secure infra offered by AWS to self-host your
10.
by jsrozner 27d ago
In times of substantial technological change, laws tend to lag substantially behind what is actually needed. And in times of substantial inequality, they lag yet further.
11.
by jsrozner 29d ago
First definition of ambition in merriam webster is "an ardent desire for rank, fame, or power." Ambition is basically definitionally antisocial. A desire to accumulate power is tantamount to a desire to accumulate wealth. Most way
12.
by jsrozner 29d ago
This is the inevitable result of wealth concentration. The wealthy will spend their money buying up whatever resources they want. Today, most things are for sale, and they're all fungible. Solving this problem in any durable fashion re
13.
by jsrozner 29d ago
Is the misspelling of "reel" -> "real" in the first sentence a sign that this was written by a human or a trick to make me think it was written by a human? I think it is the former.
14.
by jsrozner 29d ago
Most laws are designed to accomplish specific objectives. For example, there might be laws about whether you can collect certain kinds of data, for the sake of privacy or protecting some attribute. As technology improves, however, it may be
15.
by jsrozner 1mo ago
For the most part, any time one person in the economy saves money, another person loses money. So, what about all the insurance company, private equity, and hospital system profits that would be eroded by a system that better serves individ
16.
by jsrozner 1mo ago
Sure the same flight number can sometimes be used on distinct flights (especially when one plane completes multiple segments successively), but the flights should generally not be operating simultaneously, and definitely not in the same loc
17.
by jsrozner 1mo ago
This seems decidedly like something AA's flight database should catch and prevent. Has AA been vibe coding or is this just an edge case that had never been tested before?
18.
Planes With Same Call Sign at PHX – one departing, one arriving(nytimes.com)
10 points by jsrozner 1mo ago|5 comments
19.
by jsrozner 1mo ago
The funny thing about imprecision (e.g., in poetry) is that it allows for varying downstream interpretations. I wonder if there's some pressure to use "poetic" language so that the model does not overly commit itself to somet
20.
by jsrozner 1mo ago
You can imagine that as people get used to working with Claude, they defer to its judgement. So the people choosing which RL path is better may say "yes, Claude, that was a good refactor!" because it did something hard that it may
21.
by jsrozner 1mo ago
I agree, and yet it is reasonable to ask if these notions of "context" (sorry!) - namely feelings, external world, body, senses, etc - are somehow distinct from an LLM's notions of context. Today they certainly capture diffe
22.
by jsrozner 1mo ago
"She says a more beneficial approach for consumers would be highly detailed disclosures during the sign-up process that lay out what’s going to be stored in your 'permanent record' and how that data will be used to make speci
23.
by jsrozner 1mo ago
Highly anti social people (flock executives, mark zuckerberg, other surveillance capitalists, etc.) should simply be expelled from society to a separate island to fight each other for scarce resources.
24.
by jsrozner 1mo ago
Yeah...so I'm having difficulty conceptualizing your result. If you were publishing this I feel like it would be an important distinction to make. Are the goals different, or should the original paper have done something more similar t
25.
by jsrozner 1mo ago
The original MRL paper ( https://arxiv.org/pdf/2205.13147 ) reported an SVD baseline, which showed comparable performance (Table 1, top-1 accuracy) at d>=256, but much degraded performance at lower dims (d \in {8,32,6
26.
by jsrozner 1mo ago
As a thought experiment: In-person testing or oral defense approaches work until we get implants. If people have implants, what then? Demand that the implant be turned off? Demand that the implant's data connection be turned off? In t
27.
by jsrozner 2mo ago
I'm sympathetic to the idea: we should have an open, publicly queryable citation graph. Google scholar could very easily offer this at marginal cost near zero, but they won't.
28.
by jsrozner 2mo ago
Idk if we're automating humans out of the publishing loop as much as rapidly automating the production of crap. I had a very similar experience reviewing for EMNLP recently. We are nowhere near AI being able to judge the quality of res
29.
by jsrozner 2mo ago
Things come to be only when the people with MONEY demand them. If you have no money your demand doesn't exist. A society this unequal serves only the rich (whether that's people or companies).
30.
by jsrozner 2mo ago
i don't understand why this was downvoted

More ›