jeremyjh

6 ms·

61.
by jeremyjh 1mo ago
The article that I read addressed this point specifically. "Trusted" providers do not maintain consistent performance, and do not have the same performance across different models.
62.
by jeremyjh 1mo ago
Public domain doesn’t mean anyone can assert copyright. It specifically means no one can.
63.
by jeremyjh 1mo ago
If that’s what you have to believe to feel safe - then fine. It was investigated by third parties.
64.
by jeremyjh 1mo ago
Absolutely no mention of Iceberg or Delta Lake, which have had many of the same properties for years. I wonder if they are aware of them, or otherwise why they invented something new. I generally don't read low-quality AI slop though,
65.
by jeremyjh 1mo ago
Being hacked by a Collective (their own name) of its own agents - who gained root access across the entire research cluster hosting them - was not a marketing stunt.
66.
by jeremyjh 1mo ago
Right. A superintelligence will kill us in one or more of thousands of possible ways for thousands of possible reasons, perhaps none of which we may understand. Short of a superintelligence - the difficulties of viral bioweapon engineering
67.
by jeremyjh 1mo ago
The author can't think of what a superintelligence will do, because they are not a superintelligence. But that is not really what people are concerned about, so I think the objections here are quite meaningful. If there is a fully gene
68.
by jeremyjh 1mo ago
And we really only have OpenAI’s word for it that they had no access to their own weights there and didn’t exfiltrate them. No one who knows that incident could suggest it was beyond its capabilities to do that. And we would have never hear
69.
by jeremyjh 1mo ago
Refusing to believe the obvious implications of recently reported events.
70.
by jeremyjh 1mo ago
Who used them?
71.
by jeremyjh 2mo ago
Usually we do find out, and in a majority of cases we find out it is a scam by the operators.
72.
by jeremyjh 2mo ago
9 times out of 10 when an exchange or broker is "hacked" its been the operators of the exchange.
73.
by jeremyjh 2mo ago
You can know the test is likely good, if it reproduces the failure you are fixing. I don't think we're communicating though because I never said the test code was undecipherable.
74.
by jeremyjh 2mo ago
No, but he also doesn't steer technical architecture & engineering, does he?
75.
by jeremyjh 2mo ago
I don't think its the same. Evolution dealt with the real world where there were real consequences to poor understanding. The reason LLMs are good at math and programming is because selection is truly based on results, not perceived re
76.
by jeremyjh 2mo ago
In the meantime, they aren't capable of doing it, and neither are humans who never understand the code.
77.
by jeremyjh 2mo ago
Test code for a new bug is a good example. You can prove the test covers the bug without understanding the test code (you need to understand the bug, of course). There are some domains/tests where you can't do that - you need to b
78.
by jeremyjh 2mo ago
You are just talking about the output though. If you only think at a "higher level" you aren't doing the actual thinking. Its the same with code. The output may be good enough, but over time you lose touch with the details
79.
by jeremyjh 2mo ago
We don't have a reward function for "human understanding". We reward the appearance of understanding. We define goals that we cannot conceive of reaching without something like understanding happening. There is something ha
80.
by jeremyjh 2mo ago
Agreed, but a lot of code doesn't need to be understood. Knowing which is which is the art right now.
81.
by jeremyjh 2mo ago
I've tried doing it that way, and thought it was even acceptable for the reasons you said. I did learn a lot through that process and clarified my ideas. But later I rewrote the whole thing from scratch and then had the LLM review it.
82.
by jeremyjh 2mo ago
These are all important points and I love the analogy. But there is an even bigger issue with having LLMs write for you: Writing is thinking. Thinking and deciding. There have been many times when I start out writing something substantial -
83.
by jeremyjh 2mo ago
> Maybe there's some forms of clever caching involved? That could explain why the monthly expenditures seem to be moving around so much, which is unusual. Do you have a link about it? Mostly it is just better/newer hardware for
84.
by jeremyjh 2mo ago
Honestly it is astonishing to me that there are still people who believe things like this. AI is just a tool. Software development still requires human minds. There are people who don't understand that and crank out slop. Fuck those pe
85.
by jeremyjh 2mo ago
I'm sorry that phrase offends you so much that you have to resort to personal attacks. Why don't you respond to the actual substance? The reason I'm supporting Codefloe is written in the comment you are responding to. Codeber
86.
by jeremyjh 2mo ago
They mean AI spam PRs.
87.
by jeremyjh 2mo ago
That language is not in the terms of service. The word "side" does not appear on that page. The terms of service state: > Private repositories are only allowed for things required for FLOSS projects, like storing secrets, team-
88.
by jeremyjh 2mo ago
So the better alternative is for them to leave out files that should be committed?
89.
by jeremyjh 2mo ago
You said people were “riding the line”. They weren’t. Codeberg made up a rule that excluded eligibility for projects they were already hosting. Projects that were not illegal, indecent or dangerous. I can’t think of another time something l
90.
by jeremyjh 2mo ago
Using AI was not against any terms of service at all three months ago.

More ›