LPisGood

10 ms·

61.
by LPisGood 2mo ago
You go to AWS or Azure, etc. These may as well be called OpenLinuxVM
62.
by LPisGood 2mo ago
Here are some things that are abundantly clear to me. 1.) LLMs are useful for programming 2.) Open models are excellent and will continue to improve 3.) Economies of scale and ease of access mean self hosting is out of the question for a la
63.
by LPisGood 2mo ago
They address the following concern of Zhang (2005), which contained prior work involving similar techniques: > […] Of course, this conclusion is not proved mathematically. It is possible that the programs have some bugs, or the user (mys
64.
by LPisGood 2mo ago
Truly, you could say that about many conjectures, especially more “fun” classical ones.
65.
by LPisGood 2mo ago
Anthropic’s watermark is just using a known collection of random seeds during output token generation. That’s it. The models always relied on some randomness, here the randomness is fixed and remembered for querying later.
66.
by LPisGood 2mo ago
The system prompt is (and cannot be) the only guardrail against things like that, because any system prompt is little more than a good suggestion.
67.
by LPisGood 2mo ago
I don’t think it will be easy. Making a virus involves far more than reading some instructions.
68.
by LPisGood 2mo ago
I agree. If he predicted it using GPT-2 or GPT-3 I’d be quick to yell genius.
69.
by LPisGood 2mo ago
Elaborate on the proposed modification — are you referring to the fact that this can be done if distances obey a metric function?
70.
by LPisGood 2mo ago
This is still a theoretical problem. Whether or not a particular problem class admits and approximation or an arbitrarily good approximation is often of theoretical interest. One interesting example is metric TSP versus general TSP. We are
71.
by LPisGood 2mo ago
Ever since I first saw a binary integer program with millions of variables solved in less time than it took me to hit enter I realized that the fact that I had made it through graduate school for computer science, and never encountered the
72.
by LPisGood 2mo ago
“Controls” for familial wealth have mostly been debunked by any kind of basic statistical analysis.
73.
by LPisGood 2mo ago
That’s just not true. There are bypasses that happen all the time.
74.
by LPisGood 2mo ago
Could it also be called wealth-mellishence or perhaps familialValuationOfWorkEthic-melloshence?
75.
by LPisGood 2mo ago
> It's been shown to be strongly correlated with success in general Has it? I was under the impression we could hardly define, much less measure that.
76.
by LPisGood 2mo ago
System prompts are more like suggestions than hard constraints.
77.
by LPisGood 2mo ago
You don’t need to do that, actually. The legal system in the United States is highly capable of sussing out descriptions like “undue or unreasonable”.
78.
by LPisGood 2mo ago
A difference in degree is a difference in kind. Computer scientists struggle with a lot of legal concepts because they love to reduce situations in attempt to distill some particular facet and argue from analogy. A lot of poor legal analysi
79.
by LPisGood 2mo ago
I’m unfamiliar with systemd’s haters or why this would give them a valid point. Could you please elaborate?
80.
by LPisGood 2mo ago
What does that mean, practically? The person who installed linux on that particular device is liable? >no algorithmic feeds for minors by default Any choice of what content to display is an algorithm. Maybe they want a simple or easily e
81.
by LPisGood 2mo ago
It’s actually pretty good at this if you have some very specific constraints and objective.
82.
by LPisGood 2mo ago
I can’t totally tell if this is satire but we are all subject to incentives, even if (especially if?) we are aware of it.
83.
by LPisGood 2mo ago
Usually I would not point out a typo, but this one makes it difficult to grasp the magnitude of potential improvements: > the number of nodes A* has to explore decreases from 12693 to 12693
84.
by LPisGood 2mo ago
PKL files and the entire multiprocessing paradigm is one of the worst experiences I’ve ever had with a programming language feature. Surely there has to be a better way.
85.
by LPisGood 2mo ago
The AI slop readme is a real turn off. Makes me question the quality/accuracy of the rest of the project.
86.
by LPisGood 2mo ago
Setting aside AI, it’s a little outrageous they had her come in to communicate with it rather than write down how to communicate with it.
87.
by LPisGood 2mo ago
Does auto generating tests even do anything helpful? Don’t they just sort of tautologically say the code does what it does at best or do something completely ridiculous like test and implementation that only exists in the test file at worst
88.
by LPisGood 2mo ago
I’m surprised Nvidia hasn’t partnered to make a Claude chip yet. It’s a win/win you can license them out, sell them when they become obsolete, etc.
89.
by LPisGood 2mo ago
Almost surely. Apple is extremely well positioned to take advantage of this over the next decade.
90.
by LPisGood 2mo ago
And people seek out the highest quality product they can pay the lease for

More ›