back
user profile

porridgeraisin

1,976karma·1,374submissions·March 26, 2023
recent activity (1,374 total)
comment
Yes, you can even have it tpm-backed. echo -n "sk-proj-12345..." | systemd-creds encrypt --with-key=tpm2 --name=openai_key - openai.cred And then at runtime export OPENAI_API_KEY=$(sudo syst…
10h ago·view thread
comment
I wouldn't call reuters opinion free. They are bad enough to be bbc-tier I'd say. Better for sure though.
19h ago·view thread
comment
It's a straightforward "prompting" + single evolutionary algorithm technique, The paper looks like well, a paper, but the actual thing is simple.
1d ago·view thread
comment
In this setting, they are essentially using it as a feature extractor. As for moving to this versus your bespoke feature extractor, _given_ that your existing features and their compositions are not a…
1d ago·view thread
comment
They are easily verifiable and hill-climbable. Because pre-LLMs humans partially "autogenerated" kernels through hyperparameter search and in some sense eating the code complexity in return …
1d ago·view thread
comment
This is the way. Checkout the technique mentioned in the alphaevolve paper and see if it works well for your setting.
1d ago·view thread
comment
Oh yeah, forgot about those.
2d ago·view thread
comment
No doubt.
2d ago·view thread
comment
> NP-hard problems are solvable in theory but it's hopelessly expensive in practice. It's basically proven that no good algorithms exist. At least that's what I took away. You took a…
2d ago·view thread
comment
Cerebras is a large plate sized chip. It has 50GB of SRAM, and few hundred K simple cores that can access that SRAM really fast. I don't know semiconductors well, but I understand that the same m…
2d ago·view thread
comment
> Powered by cerebras > Launching first in the OpenAI API to a select group of customers with expanded access to more businesses as capacity grows.
3d ago·view thread
comment
You get much better supervised data in bio/chem though. These data companies have people working on exactly that. While it's not going to give you an "alphago" effect, it is still …
3d ago·view thread
comment
Yes. But there is also no other choice for people in these professions. The underlying job has been automated already. What's left is automating the last leg. If you consider a 5-year outlook, it…
3d ago·view thread
comment
Correct. You can look at the AI tutor jobs in the job board of any of these companies.
3d ago·view thread
comment
Supposedly grok 4.7 is a 5T model. But that's in musk units, so im not sure.
3d ago·view thread
comment
It's a different type of model. In my admittedly judgemental observation, people that aren't the type to configure fully automated harnesses with good tools and skills and verifiers for thei…
3d ago·view thread
comment
https://x.com/rao2z/status/2077383211659710848 https://xcancel.com/rao2z/status/2077383211659710848 > Summary of "Stop Anthropomorphizing …
3d ago·view thread
comment
Yeah, that got stopped I think.
4d ago·view thread
comment
I believe it is because of the CEO and his recent forays into politics. The model itself is great though, especially in grok build, which is a really nice harness I find myself preferring these days.…
4d ago·view thread
comment
Yes, it doesn't transfer. This is because the type of work required to perform in school is completely different from the type of work required to perform in college. This means first sem perform…
4d ago·view thread
comment
I share the same thinking. What do you think is a good way to try to define this "elegance"? If we try to use the mental framework of Step 1. LLM "brute forces" a search Step 2. We…
4d ago·view thread
comment
https://archive.is/rEBfL
4d ago·view thread
comment
Makes sense
4d ago·view thread
comment
I'm all for not diluting anything but dropping first semester/year grades is unlikely to cause any major issue IMO. You can encourage lighter electives maybe and remove some elective require…
4d ago·view thread