thatjoeoverthr

6 ms·

31.
by thatjoeoverthr 2mo ago
"ChatGPT says" can be annoying but it's unequivocally better than passing ChatGPT off as yourself. Quoting it for convenience ("the relevant part") --- ok, great. Does it look lazy? Well, if you respond _as its sock
32.
by thatjoeoverthr 2mo ago
I'll push back and say it's the former. Good business writing is all about the editing, and autoregressive LLMs are terrible editors. They roll dice on every word and put the same tics on every text. They meander. Suppose you chec
33.
by thatjoeoverthr 2mo ago
"Because it wasn't ..." Not so. Posting links is not the same. If you were to copy+paste someone's blog post into the forum and post is as your own, without attribution, it would be seen as plagiarism. Everyone understan
34.
by thatjoeoverthr 2mo ago
Example from work. I described a problem to a new hire. Symptoms, how to reproduce, etc. and asked him to go investigate it, pin down the problem and solve it. It was a non-trivial and hard to reproduce bug. Instead of doing any of that, he
35.
by thatjoeoverthr 2mo ago
“Not exposed” = has no knobs. Of course all autoregressive LLMs can be operated this way but Claude, the product, employs LLMs but isn’t one.
36.
by thatjoeoverthr 2mo ago
They’re a way to mitigate it. It still writes like an LLM and everyone can see it.
37.
by thatjoeoverthr 2mo ago
1) we’re not discussing those systems. We’re discussing a chat AI product called Claude, which does not offer those knobs. 2) Claude’s PRNG having a P is immaterial
38.
by thatjoeoverthr 2mo ago
Smart! I've done the same trick for resolving extracted intents to selection. But if accuracy matters, you can't rely on embedding sort to get a closet match. With a real test set they usually don't hold up under scrutiny. Ev
39.
by thatjoeoverthr 2mo ago
@andsoitis Thanks for the advice, I'll think about how I can be more diplomatic like that going forward. The SER is extremely interesting now that I dug into it so I will probe around if there is something there.
40.
by thatjoeoverthr 2mo ago
It looked like really interesting, but when I said I'd prefer to not send anything until having a chat, they ghosted.
41.
by thatjoeoverthr 2mo ago
No. It seems they found my profile on the Y Combinator job board and contacted me through that, referencing similar work on my CV. https://www.ycombinator.com/launches/Oeo-autoace-ai-co-pilot... This seems to be them.
42.
by thatjoeoverthr 2mo ago
It looked interesting, so I studied a bit into it, but I asked to present something in a meeting (no source code) so they ghosted me.
43.
Ask HN: Is this take-home a scam?(drive.google.com)
5 points by thatjoeoverthr 2mo ago|9 comments
44.
by thatjoeoverthr 2mo ago
A Y-Combinator backed company, AutoAce AI, sent me this "take home" unsolicited, in which they ask for what appears to be a complete, "production ready", deployed internal tool. They want completed source for the service
45.
by thatjoeoverthr 2mo ago
This. Who cares about TV?
46.
by thatjoeoverthr 2mo ago
It has been for years. The actual audience for a great deal of text you see is Google’s ranking system. Look up any recipe and ask yourself who reads the ten paragraph story time about grandma’s cookies. It literally isn’t intended to be re
47.
by thatjoeoverthr 2mo ago
It’s natural to the fact it’s the same model. Everyone has tics, and when a given model is asked to write millions of texts, they become visible. But! Some portion of the audience and user base can’t see it, so there is no benefit to fixing
48.
by thatjoeoverthr 3mo ago
I'm a lifetime computer guy with broad, cross-domain experience. Now going hard into deep learning and language modelling since late 2021. I'm looking to make myself useful in this field. Bring me your worst problems. English nati
49.
by thatjoeoverthr 3mo ago
If you decided at the outset they're of equal value to a $20/mo AI, lay them off.
50.
by thatjoeoverthr 3mo ago
I’m having a great time with an NVIDIA 3090. 24 GB RAM will run a lot of neat models. But at zero you can for sure just do CPU until you build a project ambitious enough.
51.
by thatjoeoverthr 3mo ago
Thanks for sharing the real one!
52.
by thatjoeoverthr 3mo ago
Thanks!
53.
by thatjoeoverthr 3mo ago
Correct. Ideal. The story deserves zero time. In this way, I am in agreement with its publisher.
54.
by thatjoeoverthr 3mo ago
It’s not. It’s very simple, but you like to say “circular reasoning”. Maybe you heard it in debate class. There is unlimited slop. I decide what to read _without reading it_. So do you. I have deemed this unworthy. It’s more fun to insult i
55.
by thatjoeoverthr 3mo ago
It doesn’t matter which model it was sampled from. In any case, time isn’t fungible. I consider censuring it substantially more valuable use of time than reading it. And it’s definitely more pleasurable than reading the same verbal tics ove
56.
by thatjoeoverthr 3mo ago
Oh, you think it’s just this story? The cost of reading slop is infinity. No thanks.
57.
by thatjoeoverthr 3mo ago
Not all arguments can be evaluated rigorously. Life is too short. They will be filtered on heuristics. No choice. This text was obviously randomly sampled from a popular, off the shelf model. You will have to accept that wise audiences will
58.
by thatjoeoverthr 3mo ago
Sufficiently advanced AI use is probably fine. The slop everyone complains about has certain tells specifically due to some combination of the following: - The author is conducting some kind of hustle. - The author doesn't bother editi
59.
by thatjoeoverthr 3mo ago
So are all the other shady websites.
60.
by thatjoeoverthr 3mo ago
Seems absurd to run a COTS agent at token rates to "play" a game. Like watching the computer beat its meat. Building your own models for it would be an eye-opener though. Learn a lot.

More ›