janalsncm

6 ms·

61.
by janalsncm 1mo ago
I don’t need to know the value of their limit, I just need to know their bounds. Just like a prison doesn’t need to know the strength of each inmate, just that they can’t bend or bite through steel bars. Cryptography is real, physics is rea
62.
by janalsncm 1mo ago
To the extent this is tongue in cheek it is funny, but I could imagine if I was building a startup I’d like something like this if it was good . That said, if you look at what it’s actually doing (read the question carefully, then read wha
63.
by janalsncm 1mo ago
They were already pretty vocal about it, but this makes Nvidia pretty much the de facto face of open weight models. As battle lines get drawn over duopoly vs. open weight it’ll be interesting to see what Nvidia does. They definitely want a
64.
by janalsncm 1mo ago
In this case they wanted to test its cybersecurity capabilities and did not airgap it. The test itself did not require an internet connection.
65.
by janalsncm 1mo ago
Why do people think that omniscience is the same as omnipotence? There are limits to what smarts can accomplish.
66.
by janalsncm 1mo ago
Until the AI finds a zero day exploit in physics, a faraday cage works pretty well to block WiFi.
67.
by janalsncm 1mo ago
The real problem with alignment is that if someone ever “solves” it the party will be over and no one will get funding to “research” it anymore.
68.
by janalsncm 1mo ago
> There is no amount of care that will be able to fully protect you. I disagree. A properly engineered sandbox would have prevented the escape. Monitoring the agents’ plans would have prevented it. Interrupting one stage in a multi-stage
69.
by janalsncm 1mo ago
Or worse. Look at what happened to Figma.
70.
by janalsncm 1mo ago
Are their TOS significantly more vague or restrictive than OpenAI or Anthropic’s? In any case what matters is what is enforced in practice. It will be a mild inconvenience to switch providers on Openrouter. If Anthropic or OpenAI decide to
71.
by janalsncm 2mo ago
Counterpoint: what we call AI is really just more sophisticated statistical algorithms, and so all of the stories about using AI, making AI more efficient, and all of the vibe coded apps are really just the world coming to terms with a new
72.
by janalsncm 2mo ago
The fair comparison would be to also do the reverse: start with Sol then have Fable clean up. Then compare the Fable-Sol and Sol-Fable outputs side by side.
73.
by janalsncm 2mo ago
You are right that the field doesn’t have a theoretically sound explanation for the architectural choices aside from “A works better than B”. However, I would argue this is an ideal opportunity for the “gentleman scientist” or eager 17 year
74.
by janalsncm 2mo ago
I think ChatGPT is a great way to get depth on a topic. You can Q&A for a while until you get it, and it won’t get impatient with you. That said, I think traditional resources are a better way to get breadth and to frame the topic. Just
75.
by janalsncm 2mo ago
I think it’s reasonable to have a shallow understanding of most parts and a deep understanding of a small number of parts. That’s how most engineers are. Most software engineers do not have a deep understanding of CPU architectures. In fact
76.
by janalsncm 2mo ago
This is essentially the anti- Bitter Lesson lesson which I feel has become a bit of a thought terminating cliche lately. The Bitter Lesson says that eventually general approaches which leverage more data and more compute will outperform th
77.
by janalsncm 2mo ago
> Data retention rules imposed by the Trump administration have also hampered Fable’s adoption, according to Ara Kharazian, chief economist at Ramp. This is in the article.
78.
by janalsncm 2mo ago
https://news.ycombinator.com/item?id=3656581 I really wonder what he would say today. In a way, being devoured by LLMs achieved SO’s ultimate goal of helping people solve programming problems quickly, but at the cost of SO.
79.
by janalsncm 2mo ago
We don’t need to keep chasing this moving goalpost. The Patriot Act is not an “iffy example”. It is the canonical one, and was subsequently reauthorized in bipartisan votes.
80.
by janalsncm 2mo ago
If the model and its organization are focused on strength at agentic coding tasks, they are not so concerned with the prose in the middle. They might even have a version that writes less annoying prose, but they are being squeezed hard by O
81.
by janalsncm 2mo ago
https://en.wikipedia.org/wiki/Patriot_Act The senate vote for the patriot act was 98 to 1.
82.
by janalsncm 2mo ago
I would recommend “Means of Control” for a more clear picture of how the surveillance state works today. Here is a review: https://reason.com/2024/02/27/the-spy-in-your-pocket/ The current situation is a
83.
by janalsncm 2mo ago
The US doesn’t protect positive rights (requiring the government to give you something) in the same way as Europeans. But this case is pretty clearly about negative rights (preventing the government from doing something) which based on the
84.
by janalsncm 2mo ago
I don’t see how that responds to the point in the parent comment. Censorship or not, chatbots are unreliable for serious history questions. Just today Gemma told me that for a long time the Iliad and Odyssey were considered mediocre literat
85.
by janalsncm 2mo ago
You assume your highly charged political query is hitting the main LLM at all and not some external short circuit.
86.
by janalsncm 2mo ago
Call me old school but I feel like the solution to humans clogging up the slop cannon will be for AI generated code to look more human. That is, the code should be correct but also idiomatic.
87.
by janalsncm 2mo ago
I would imagine the number of people who choose Claude code or Codex because it gives a political opinion they like rather than producing quality code is pretty close to zero.
88.
by janalsncm 2mo ago
If you think my comment is bleak you should read Jane Eyre. Then come back to my comment if you’d like to get cheered up. It is truly a miserable book.
89.
by janalsncm 2mo ago
I disagree with many of the points. First, as a point of style it makes no sense for the author to have chosen “thick” rather than the more common adjective “deep”. They describe the same thing, and “deep” even works much better for the cav
90.
by janalsncm 2mo ago
Really a shame because understanding why your results were incorrect is much more important than stumbling on the right answer by accident.

More ›