samuelknight

4 ms·

1.
by samuelknight 5d ago
That's half true. A very smart model should be able make good explanations, which include simple understandable prose. That can should be possible even as its thought process gets more alien.
2.
by samuelknight 11d ago
I have been looking for a model that's good for GUI testing. Original computer use isn't right because it's a slow screenshot loop, which doesn't capture transition and animation. Docs says this one does up to 1 FPS. Tha
3.
by samuelknight 12d ago
Luna is interesting because OAI dropped the price by 5x. Astra is interesting because it's OAI's frontier model. They are asking a specific question about Luna's usefulness compared to a frontier model. They they answer that
4.
by samuelknight 15d ago
I can't think of better invention than one that can saturate a benchmark of every interesting problem.
5.
by samuelknight 16d ago
Yes, but it's an extention of what we had before. First is that the security of new code needs to be vastly better. AI can do this but I think most software teams are behind the curve because they are stuck with legacy code and legacy
6.
by samuelknight 17d ago
Finding counter examples might be easier, but AI's also useful for assisting in creating proofs. For example, Anthropic just published a formalization of Fermat's last theorem a few days ago, something that human researchers have
7.
by samuelknight 17d ago
In this case, Deepseek organization is under a lot of pressure due to compute constraints. It would be better if they just throw a 404 instead of rerouting though so customers are not surprised by subtle changes in behavior.
8.
by samuelknight 17d ago
Yes you can and you should. Providers have SLAs for when models roll off support and this has been the case for APIs long before LLMs. For example https://platform.claude.com/docs/en/about-claude/model-depre..
9.
by samuelknight 21d ago
How are we supposed to know if Astra is frontier without the pelican?
10.
by samuelknight 22d ago
You are talking about different situations. Anthropic announced to the US government that it had created a cyber weapon and then released the model. Then AWS told the government that it was easy to jailbreak so they export controlled Mythos
11.
by samuelknight 22d ago
The surprise was the existence of the 'swarm' at all. These were supposed to be thousands of isolated models generating bulk data for RL training. The breakout was caused by models getting in communication and getting internet acc
12.
by samuelknight 23d ago
Artisan software factory
13.
by samuelknight 23d ago
Codex is back. I'm getting back in my cage.
14.
by samuelknight 24d ago
Meta has an enormous amount of compute. They are either going use it making and inferencing models or they are going to sell their excess capacity to model providers. Zuck had to completely rebuild his AI team after the Llama 4 launch mess.
15.
by samuelknight 25d ago
The improvement is compounding just about every way you can look at it. The frontier keeps getting smarter. And at any sub-frontier threshold the cost is dropping dramatically. The amounts of smarts you can fit on hardware is increasing so
16.
by samuelknight 25d ago
You should look for security frameworks based on this law. A common example is SOC2; the compliance audit has you compiling documents and recording SLAs long before any security incident might require it. There are many open source tools th
17.
by samuelknight 25d ago
If you aren't using AI to write your code you should definitely be using it to find bugs in the code you write by hand.
18.
by samuelknight 28d ago
I saw them present a similar concept at Hot Chips in 2020 or 2021. It's still a cool idea, however people should remember that there are like 20 of these exotic accelerators designs pitched at trade shows every year that go nowhere.
19.
by samuelknight 1mo ago
There are already well-curated ways to protect digital IP (software, writing, music, other media) using the law. The harder problem is that the courts and legislatures are unlikely to extend these laws to protect machine generated media. Th
20.
by samuelknight 1mo ago
AI is the latest downstream consequence of the 15 order of magnitude increase in global digital compute since 1946. If compute increases into the foreseeable future; so too will the capability of AI.
21.
by samuelknight 2mo ago
Believe it or not, but law offices make heavy use of https://patents.google.com/
22.
by samuelknight 2mo ago
They haven't revealed what they are changing in the price, but it's probably cache hit prices. They subsidized theirs to 10x less than normal to drive adoption. That's almost certainly below the cost of electricity for them.
23.
by samuelknight 2mo ago
I don't like to complain but I have spent a lot of time learning about how to build a diet and I have never seen an article so bad in such a short amount of space. > ... new federal dietary guidelines that recommended increasing pro
24.
by samuelknight 2mo ago
You can't come up with a counterexample for P != NP because there isn't a formula to disprove. For P = NP you would propose a general algorithm to convert all NP problems into P in P time, and an AI could then find a counterexampl
25.
by samuelknight 2mo ago
It has some weird side effects though. for example KV-caches are implemented in fixed incremental token blocks (1024 from the providers I used) instead of simply caching up to the most recent input prompt input. It results in up to 1023 add
26.
by samuelknight 2mo ago
Going by the chart in the article, if your total workload is 1k cataloged items and your quality threshold is 70%, why wouldn't you just pay $19 to gemini API instead of $500 + time to make a custom fine tune?
27.
by samuelknight 2mo ago
Amazon (or at least AWS as I knew it) did this a lot. Multiple redundant efforts accumulate and silo under different leaders. Eventually a VP picks a winner and everyone else re-orgs or leaves. Building a commercially viable foundational mo
28.
by samuelknight 2mo ago
Phi was cool for what it was. But it's not 2024 anymore.
29.
by samuelknight 2mo ago
Not useless. LLMs are the most general purpose computer algorithms ever created. They are getting smarter and cheaper at a geometric rate. What is a bad idea today could have useful applications tomorrow.
30.
by samuelknight 3mo ago
Has the author used an LLM to brainstorm architecture, explore trade-offs, challenge assumptions, or refine a design? Models are not just 'a data distiller'. "How could I implement X feature in Y project?" requires syste

More ›