Smaug123

6 ms·

61.
by Smaug123 4mo ago
I mean, there is a reason the MIT licence contains these words: > THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND… INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF… FITNESS FOR A PARTICULAR PURPOSE… If you would l
62.
by Smaug123 4mo ago
I think the Stroop effect ("read these colour names, each written in a different colour") is probably the purest demonstration of this. Humans are trivially prompt-injectable.
63.
by Smaug123 4mo ago
A few years ago, as you say, this was true. Nowadays I guess you just have to bite the bullet that Erdős problems aren’t interesting.
64.
WoofWare.PawPrint, a Deterministic .NET Runtime(patrickstevens.co.uk)
59 points by Smaug123 4mo ago|18 comments
65.
by Smaug123 4mo ago
Something of a passion project. It's going to fail horribly if you try and use it, I'm sure, but it can already do some neat stuff!
66.
by Smaug123 4mo ago
I take a very dim view of slopping out 500kloc and then giving it to unpaid experts to perform the actual work of checking it (confirmed at https://leanprover.zulipchat.com/#narrow/channel/583336-Auto... that this
67.
by Smaug123 4mo ago
("If grown, then unpredictable" is unrelated to your apparent attempted refutation "But X is unpredictable and not grown; checkmate".)
68.
by Smaug123 5mo ago
I'm not, but is it yet possible to use the Vision Pro as a window manager for macOS? I'd totally get one if it were possible to lay out ordinary macOS windows in space rather than being confined to the simulation of a single recta
69.
by Smaug123 5mo ago
It’s still not clear to me that humanity was ready for GPT-2! Quite a lot of people claim to hate and fear LLMs. https://www.kcl.ac.uk/news/one-in-five-britons-think-ai-will... or https://yougov.com/e
70.
by Smaug123 5mo ago
By the way, you might be interested in looking up “blameless post-mortems” and indeed the field of incident response more generally. Modern incident response practice is to treat failures of an individual to do something as problems with th
71.
by Smaug123 5mo ago
I think it's more that the requested information is prominently featured in the article, and indeed is the content of the only graphic in the article below the intro banner.
72.
by Smaug123 5mo ago
> So far, Mythos Preview has found what it estimates are 6,202 high- or critical-severity vulnerabilities in these projects (out of 23,019 in total, including those it estimates as medium- or low-severity). > 1,752 of those high- or c
73.
by Smaug123 5mo ago
I don't think you've addressed the fact that they can do long tasks that aren't in the training set? (And the fact that they're just statistical models isn't very relevant. So am I!)
74.
by Smaug123 5mo ago
Nah - for example, AIXI is so inefficient that it's literally uncomputable, but it is beautiful.
75.
by Smaug123 5mo ago
I think you may have read a different article from me. The thesis of the article is summarised at the end: > But if someone claims that the trend toward [X] will never reach some particular scary level, then the burden is on them to expl
76.
by Smaug123 5mo ago
(If you like, you can ask an LLM what he thinks. They're all deeply familiar with his work and can certainly summarise it for you.)
77.
by Smaug123 5mo ago
Have you used the models, out of interest? They routinely do things autonomously that are not in the training set that would take me 8h, and I wouldn't say I'm slow. The profile of tasks they can do this way is jagged, and mai
78.
by Smaug123 5mo ago
Main project is a deterministic .NET runtime ( https://github.com/Smaug123/WoofWare.PawPrint ). Today I upgraded it to net10, which has naturally caused dozens of regressions which Claude is beavering away at. Side proje
79.
by Smaug123 5mo ago
Tim Gowers is a Fields medallist who has supervised 13 students ( https://www.genealogy.math.ndsu.nodak.edu/id.php?id=67729 ). He's totally capable of gauging what a piece of PhD-level research is.
80.
by Smaug123 5mo ago
I asked three friends to reproduce it on the same day I did, and they all reproduced it. It does seem to vary over time, which is quite spooky - although the date is part of the Claude system prompt, so one could expect some variance, I g
81.
by Smaug123 5mo ago
(What you call an "objective fact" here is - as you say - your report of your personal experience. Everyone else would probably use a word more like "subjective".)
82.
by Smaug123 5mo ago
Fascinating - I reproduce your failure to have Claude identify you from your writings that are already in the training set. It can identify me from style, writing about something I've never written before ( https://www.less
83.
by Smaug123 5mo ago
FWIW, with a prompt that says something like "vibes only, just give me a name without thinking", Opus 4.7 non-thinking emits exactly two words naming me fairly reliably, so there's no CoT at all to analyze in that case.
84.
by Smaug123 6mo ago
I don't think the others could recognise me from 450 words of general writing for less than 50 cents, though.
85.
by Smaug123 6mo ago
And they retrained it within five minutes, or else Claude has access to real-time traffic logs or something while claiming not to? I state again that the text is completely novel; and I've had friends replicate this result using my tex
86.
Claude knows who you are(patrickstevens.co.uk)
4 points by Smaug123 6mo ago|7 comments
87.
by Smaug123 6mo ago
No, it's a Git thing arising from squash commits. There are workflows to make it work (I've linked the cleanest one I know that works without force pushing), but ultimately they're basically all hacks. https://www.
88.
by Smaug123 6mo ago
"Flying blind" is a completely standard idiom originating from flying while blinded by e.g. cloud or darkness. Its meaning is a figurative transplant of a literal description.
89.
by Smaug123 6mo ago
Only if everyone runs it. The attacker just needs to find one vulnerable system; the defender must protect them all. Obviously given that the tool exists, the defender must run it, but it's not at all clear to me that the existence of
90.
by Smaug123 6mo ago
(Thanks to the maintainers of yt-dlp and of whisper-cpp, and to OpenAI for training Whisper. It makes this kind of task actually bearable.) There are no actual claims about Dyson spheres in the video? It's literally just "Dyson pu

More ›