agucova

5 ms·

1.
SPAR – Fall 2026 AI Safety Research Projects(sparai.org)
1 points by agucova 1mo ago|0 comments
2.
by agucova 10mo ago
FWIW I work on AI and I also trust Pangram quite a lot (though exclusively on long-form text spanning at least 4 or more paragraphs). I'm pretty sure the book is heavily AI written.
3.
by agucova 10mo ago
How long were the extracts you gave to Pangram? Pangram only has the stated very high accuracy for long-form text covering at least a handful of paragraphs. When I ran this book, I used an entire chapter.
4.
by agucova 10mo ago
I ran the introduction chapter through Pangram [1], which is one of the most reliable AI-generated text classifiers out there [2] (with a benchmarked accuracy of 99.85% over long-form text), and it gives high confidence for it having been A
5.
by agucova 2y ago
This benchmark’s questions and answers will be kept fully private, and the benchmark will only be run by Epoch. Short of the companies fishing out the questions from API logs (which seems quite unlikely), this shouldn’t be a problem.
6.
by agucova 2y ago
For some context on why this is important: this benchmark was designed to be extremely challenging for LLMs, with problems requiring several hours or days of work by expert mathematicians. Currently, LLMs solve 2% of problems in the set (wh
7.
by agucova 2y ago
I’m guessing he’s probably talking about LessWrong, which nowadays also hosts a ton of serious safety research (and is often dismissed offhandedly because of its reputation as an insular internet community).
8.
by agucova 2y ago
I mean, this is how the Reflection model works. It's just hiding that from you in an interface.
9.
by agucova 2y ago
You can use Daggity.jl: https://docs.juliahub.com/Dagitty/kxRMH/0.0.1/
10.
by agucova 2y ago
I agree. I’m really more concerned about bioweapons, for which it’s generally understood (in security studies) that access to technical expertise is the limiting factor for terrorists. See Al Qaeda’s attempts to develop bio weapons in 2001.
11.
by agucova 2y ago
I imagine you meant societal harms? I think this was mostly my fault. I edited the areas of work a bit to better reflect what the UK AISI is actually working on right now.
12.
by agucova 2y ago
I recommend checking out the UK AISI's work on this: - https://www.gov.uk/government/publications/ai-safety-institu... - https://www.aisi.gov.uk/work/advanced-ai-evaluations-may-upd...
13.
by agucova 2y ago
> A government agency determining limits on, say, heavy metals in drinking water is materially different than the government making declarations of what ideas are safe and which are not Access to evaluate the models basically means the U
14.
by agucova 2y ago
> Because lobbying exists in this country, and because legislators receive financial support from corporations like OpenAI, any so-called concession by a major US-based company to the US Government is likely a deal that will only benefit
15.
by agucova 2y ago
> My issue with AI safety is that it's an overloaded term. It could mean anything from an llm giving you instructions on how to make an atomic bomb to writing spicy jokes if you prompt it to do so. it's not clear which safety t
16.
by agucova 2y ago
> What exactly does the evaluation entail? I believe the US AISI has published less on their specific approach, but they’re largely expected to follow the general approach implemented by the UK AISI [1] and METR [2]. This is mostly focus
17.
by agucova 2y ago
> If training data contains multiple conflicting perspectives on a topic, the LLM has a limited ability to recognize that a disagreement is present and what types of entities are more likely to adopt which side. That is what those studie
18.
Asterisk Magazine Issue 6 – California(asteriskmag.com)
1 points by agucova 2y ago|0 comments
19.
by agucova 2y ago
I messed up the second reference, it should be https://arxiv.org/abs/2212.03827
20.
by agucova 2y ago
This isn't really true. LLMs are discriminating actual truth (though perhaps not perfectly). Other similar studies suggest that they can differentiate, say, between commonly held misconceptions and scientific facts, even when they&#x27
21.
by agucova 2y ago
> Now, we can see from this description that nothing about the modeling ensures that the outputs accurately depict anything in the world. There is not much reason to think that the outputs are connected to any sort of internal representa
22.
by agucova 2y ago
Is your hypothesis that what, Jan Leike resigned as part of an elaborate conspiracy to boost OpenAI's prospect by... criticizing it? I find these theories to be extremely convoluted and implausible, and they often lack awareness of the
23.
by agucova 2y ago
For context, the point of the Superalignment team was to work on a problem known as scalable oversight: the problem of aligning models in a way that holds up as models become more capable [1]. The reason behind this is that current alignmen
24.
by agucova 2y ago
I find this kind of dismissive attitude annoying. There are good arguments in the literature for why you might want to care about these risks [1, 2], and I think there's lots of room for reasonable disagreement about whether these are
25.
by agucova 2y ago
Agreeing with circuit10's comments, I don't think many proponents of AI Safety are doing so through Pascal wagers. People differ a lot in their assessment of how likely certain risks are, but people I know working on AI Safety ten
26.
by agucova 2y ago
This is true, but even engineers see the advantages of Julia. My engineering school has went from almost pure Matlab usage to many key engineering courses switching to Julia due to its simplicity and friendliness. It's also SOTA for ma
27.
by agucova 2y ago
Which pathways exist for a Chilean citizen who didn't complete their bachelor's degree on CS, but nonetheless wants to take a US job for a 501(c)(3) in a technical capacity? I worked several years as a software engineer, but this
28.
by agucova 2y ago
Now I'm curious about what's your take on them
29.
by agucova 3y ago
From the article: The COSMIC system is a cluster of FPGA, CPU, and GPU hardware, which at all times receives a copy of digitized data streams from each of the VLA’s antennas. The data received by COSMIC necessarily reflects antenna-poin
30.
by agucova 3y ago
Note that: > ML cannot conceptualize of things in the abstract like people can And: > They cannot offer reasons, a train of thought like a person Are very different claims! The first one just seems wrong: LLMs require abstraction to w

More ›