lou1306

5 ms·

31.
by lou1306 4mo ago
Yeah except that the agent cannot be held accountable if its fix is crap, or if it went off-track mid-fix.
32.
by lou1306 4mo ago
You actually hire a developer to work on that issue and not something else.
33.
by lou1306 4mo ago
Perhaps relevant context: The EU commission just ignored the "Tech Sovereignty Package" it launched ~3 weeks ago, and explicitly referred open-source as a core element of their strategy, and endorsed W, another ATproto-based socia
34.
by lou1306 4mo ago
Not just all BSky traffic: all traffic from every ATProto service
35.
by lou1306 4mo ago
Are we really discussing a plan that has never been attempted before (training a frontier model on federated hardware) and that would require coordination on a continental scale, sketched on a five-page PDF on GitHub with no discernible aut
36.
by lou1306 4mo ago
The "good lawyer" is/was a major archetype in modern Western culture. See "The Devil and Daniel Webster", "To Kill a Mockingbird", "Paths of Glory", just to name a few examples.
37.
by lou1306 4mo ago
Good. This will force kids to be creative, hack around, and question authority.
38.
by lou1306 4mo ago
Well the size of IJ is itself a statement: 1. there's an awful lot to unpack once you start exploring the intersection of capitalism, addiction, and entertainment; 2. can you even _manage_ to go through it all without switching to mind
39.
by lou1306 4mo ago
If the backend function was so poorly coded to allow such a gargantuan security hole, then it is an even worse problem. Basically Meta is throwing its own engineers under the bus so that its AI chatbot can save face. Scary stuff. Unless the
40.
by lou1306 4mo ago
You joke but this is almost literally what Chain-of-Thought does, at least in the early days. They basically just added "Wait," to the model's output and fed it back to the model iirc
41.
by lou1306 4mo ago
Well people _used_ to pay for it, in fact. Except it was 3 EUR/year rather than 3 EUR/month...
42.
by lou1306 4mo ago
To be honest, Zork at times makes precious little sense: you are supposed to die over and over before you figure stuff out. For instance, you have to grab the endless-light-source treasure very early on, or you mathematically cannot win. An
43.
by lou1306 4mo ago
I think the order in which the different elements of the book are introduced is crucial, as it leads to a lot of "aha!" moments. > What put me off it is it just kind of reads like a rambling stoner conversation. Yeah well, that
44.
by lou1306 5mo ago
If things get as much as 10% as chaotic as you predict, there will be massive turmoil and the "Destroy all Clankers" party will actually be the one in charge.
45.
by lou1306 5mo ago
This has really little to do with embrace and extend. They are not taking over an open standard or anything like that. If anything, it's forced dogfooding, i.e., forcing their own workforce to beta-test their product.
46.
by lou1306 5mo ago
I don't think they mention it in the CoC, but anyone can upload an ArXiv preprint listing literally anybody else as a coauthor, so it is only logical that only "confirmed" coauthors should be affected.
47.
by lou1306 5mo ago
As I mentioned in another thread: To be a coauthor on a preprint that you have not submitted, you have to actively "claim" it (using a password given to the author who submitted). It's on you to double-check before claiming.
48.
by lou1306 5mo ago
To be a coauthor on a preprint that you have not submitted, you have to actively "claim" it (using a password given to the author who submitted). It's on you to double-check before claiming.
49.
by lou1306 5mo ago
> some people don't like LLMs It's not even that they "don't like LLMs". They just don't like academic fraud! If references were fabricated with a Markov chain it would be just as bad!
50.
by lou1306 6mo ago
EXIF data enabled some pretty cool things, like PhotoSynth [1]. This stuff looks futuristic even now, ~20 years later, basically what the Metaverse could/should have been. [1] https://www.ted.com/talks/blaise_aguer
51.
by lou1306 6mo ago
But most "normal" neural networks are feed-forward, so they are guaranteed to terminate in a bounded amount of time. This rules Turing completeness right out. And even recurrent NNs can be "unfolded" into feed-forward eq
52.
by lou1306 6mo ago
> Now imagine trying to land on that airship from orbit or get back into orbit (and beyond) from that airship. None of this is easy here on Earth. Also, if anything goes wrong on Earth and you're in the atmosphere, there's sti
53.
by lou1306 6mo ago
The alternative is have hooks _forcibly_ run on people's machines, which is fantastic as an attack vector and CVE generator but probably not a good choice in other respects.
54.
by lou1306 6mo ago
Well you can tell them to please enable hooks in the PR guidelines, but you cannot really police what they do or don't run on their own machines.
55.
by lou1306 6mo ago
> On top of that, there are a lot of personal and subjective aspects to code. You might have certain preferences about formatting, style, structure, dependencies, and approach, and I have mine. 95% of this is covered by a warning that sa
56.
by lou1306 6mo ago
Yes, this is well-known (eg. in abstract interpretation). As you said, usually you can set a "cap" to the size of these objects, and start merging intervals when you hit the cap. But at least in abstract interpretation it seems th
57.
by lou1306 6mo ago
Sorry to be a party pooper, the Web app is neat, but I have some reservations about the paper. Namely, the "powerset of intervals" domain has been known since the '70s [1], and powerset domains have been generalised to arbitr
58.
by lou1306 6mo ago
This guy named Ludwig Wittgenstein would like to have a word with you.
59.
by lou1306 6mo ago
> they're identically-named things that both repackage OpenAI's LLMs Haven't tried it yet but the GitHub Copilot extension for VSCode also seems to integrate Claude, Gemini and other non-OAI stuff
60.
by lou1306 7mo ago
> Every conversation is sent to an llm at least a thousand times the size of gpt2 which could one shot this nearly a decade ago. Yes, but that is _what the product does_. What we are talking about is _telemetry_.

More ›