feanaro

5 ms·

1.
by feanaro 13d ago
If you consider the data on which an LLM was trained on to be points on a very highly multidimensional object, the claim is that the LLM can interpolate a convex hull spanned by those points, therefore recovering a subset of consequences at
2.
by feanaro 23d ago
Full text: https://hal.science/hal-05643897
3.
by feanaro 2mo ago
Surveillance is decidedly and completely unlike locks.
4.
by feanaro 2mo ago
> This is the million dollar we'll see answered in our lifetime. Software engineering exists to automate work, are we arrogant to think we are not destined to the same fate? Is this truly a job befitting of a human over an agent? Th
5.
by feanaro 2mo ago
What is also lower is your understanding of the change. So yes, if you are now essentially only doing the final mile of paper pushing for the LLM, then the mental burden is lower but so is the assurance of what has just transpired. Whether
6.
by feanaro 3mo ago
> How do you type an em-dash? Is it OS specific? On Linux X11 at least, you can enable the Compose key and then press `<Compose>---` which results in — and `<Compose>--.` which gives you –
7.
by feanaro 4mo ago
> The almond thing is false, but I'd argue that "misleading" might be defensible if you were to accompany it with "the majority of almonds are grown in California, but not all of them". The "majority" i
8.
by feanaro 8mo ago
> I mean, "the pieces were already there" is true of everything? Einstein was synthesizing existing math and existing data is your point right? If it's true of everything, then surely having an LLM work iteratively on the
9.
by feanaro 8mo ago
No, by saying this, I am not downplaying Einstein's sizeable achievements nor trying to imply everyone was wrong about him. His was an impressive breadth of knowledge and mathematical prowess and there's no denying this. However
10.
by feanaro 8mo ago
This alone still wouldn't be a clear demonstration that AGI is around the corner. It's quite possible a LLM could've done Einstein's job, if Einstein's job was truly just synthesising already available information i
11.
by feanaro 8mo ago
No, that's a completely different concept, because we have faultless machines which perfectly and deterministically translate high-level code into byte-level machine code. This is another case of (nearly) perfect abstraction. On the ot
12.
by feanaro 11mo ago
This is essentially a group theoretical result about permutation groups. Would be nice to see a treatment from this angle.
13.
by feanaro 1y ago
And yet the categorical concepts in Hask are undoubtedly practically useful, more so than an arbitrary sample of concepts, and compose extraordinarily well. Does that have nothing to do with those concepts deriving from (even more general c
14.
by feanaro 2y ago
I don't think anything good ever came from Ylva Johansson. Mentions of her name on something should make one automatically treat that thing with suspicion.
15.
by feanaro 2y ago
How do you discover the principles in the first place? You can discover them once and then apply them in all applicable places precisely because you generalised them. You have a point that the result may very well be more easily explained i
16.
by feanaro 2y ago
What do you mean by the Wireguard option for mitmproxy? EDIT: Oh, look at this https://mitmproxy.org/posts/wireguard-mode/ . TIL.
17.
by feanaro 2y ago
A server is someone else's device. Your phone is your own device. So no, doing the scan on your own device and making your device your potential adversary is not better than doing it on the server. You can always choose not to use the
18.
by feanaro 2y ago
So there is an upper limit, which is the real price?
19.
by feanaro 2y ago
What are you even on about, mate? A hacker's multi tool with infinite potential for exploration is an idea "too malicious" to consider?
20.
by feanaro 2y ago
lnav is amazing and I use it often. I do have a list of gripes where I think it could be improved, so I'm just going to dump them here in case you're interested: - regex101 support for quickly defining custom formats is just awes
21.
by feanaro 2y ago
What does it mean that Peano Arithmetic isn't categorical?
22.
by feanaro 3y ago
Yes, it's called rr. https://rr-project.org/
23.
by feanaro 3y ago
It's not a "lost" key if it's found hardcoded in an easily available place (e.g. an application). It's a negligently placed key leading to a vulnerable place that is going to get into the hands of a malicious person
24.
by feanaro 3y ago
Your machine regularly uses that key to access what's behind the door on your behalf; there is no reason you shouldn't be able to access it yourself. If you don't find the key and realise it's actually a lost one, leadin
25.
by feanaro 3y ago
"Viewing" private data for purposes of verification of the issue. How about you just don't ship passwords in the application like some negligent troglodyte?
26.
by feanaro 3y ago
More likely the model has a system prompt authoritatively saying what "now" is, and it can reason about other times specified in other resources in the training set because those resources specified their own time reference. So ev
27.
by feanaro 3y ago
Kagi should hire the Marginalia author.
28.
by feanaro 3y ago
I think they are expressing a view that we ought to offer less protection / more scrutiny to larger commercial entities, which concentrate disproportionate amounts of wealth and power, compared to smaller entities. I tend to agree.
29.
by feanaro 3y ago
> Even if I'm not using bisect (I rarely do), having the 'git log' be polluted by merges makes it harder for me to fit everything into my head. How and why? Done properly, a merge commit of a PR breach is semantically equi
30.
by feanaro 3y ago
How does merging have the same problem? By definition, merging only requires you to resolve the conflicts once, at the point of the merge commit.

More ›