gdiamos

5 ms·

31.
by gdiamos 3mo ago
I wonder if Amazon eventually gets cut out by 3D printing/replicators for imitable objects.
32.
by gdiamos 3mo ago
Scaling laws assume the error metric and data distribution. There is a lot of follow on work that explains what happens as you change them, e.g. Scaling Laws for Transfer - https://arxiv.org/pdf/2102.01293 I think it’s
33.
by gdiamos 3mo ago
When I first saw scaling laws in that deep speech experiment notebook, I didn’t believe it could be real. I was worried for months that we made a mistake, or that it only worked for that one dataset. I started to believe it after we (Joel
34.
by gdiamos 4mo ago
He started with tinkercad and thingiverse. I tried basic elegoo and bambu printers. He can’t read very well but he likes dragging shapes around on a tablet. He would ask me to find shapes using the search engines then he mixes them together
35.
by gdiamos 4mo ago
My kindergartner has a 3D printer. I got a call from the school principal. She said “another parent called and said your son 3D printed a gun and brought it to school”. I looked at the print history. It was a tiny toy mandalorian figurine h
36.
by gdiamos 4mo ago
Usually breakthroughs in computing lead to more usage of computing, not less.
37.
by gdiamos 4mo ago
This is why I use a router to send my own IP to my own models, and general information to Claude. https://split-brain-ui.scalarxlm.com/docs/clients I expect Claude to train on my general tokens. I train my own model on
38.
by gdiamos 4mo ago
This weekend I was reading this paper on programming the Cerebras wafer scale engine, https://arxiv.org/html/2405.07898v1 . Data movement is the expensive part of computing, and some algorithms like stencils only requi
39.
CUDA-like programming of Cerebras WSE(github.com)
2 points by gdiamos 4mo ago|1 comments
40.
by gdiamos 4mo ago
What I do is route general data to Mythos, and my own IP to a local model. I expect them to train on their traffic, and I train on mine.
41.
by gdiamos 4mo ago
I don’t get it. That’s what I am using.
42.
by gdiamos 4mo ago
Think about the worst enterprise SaaS apps you have used…
43.
by gdiamos 4mo ago
What I tell my team to do is to drop using so many cloud saas apps, and build more themselves using LLMs. I’m not planning on firing people, but I am planning on building more, using more tokens, and less app subscriptions. One aspect of bu
44.
by gdiamos 4mo ago
It’s about enterprises who care about supply chain risk and having a throat to choke if they have a problem. Here’s a real example. I’m in a design meeting talking about a model use case. We have a question about the data pipeline or the pr
45.
by gdiamos 4mo ago
There is demand for US open models.
46.
I aint gonna work on Maggie's Datacenter no more [video](youtube.com)
3 points by gdiamos 4mo ago|0 comments
47.
Demo: Fold your coding sessions into LLM weights(app.scalarlmforge.com)
2 points by gdiamos 5mo ago|0 comments
48.
by gdiamos 5mo ago
Instead of move to duck duck go I just stopped using search
49.
by gdiamos 5mo ago
How far can a pure mercenary culture get?
50.
by gdiamos 5mo ago
Don’t put it past Dario to buy spaceX
51.
by gdiamos 5mo ago
I’m seeing founders being encouraged to run their business with AI and cut out the etc etc
52.
by gdiamos 5mo ago
Smart move
53.
by gdiamos 5mo ago
“We hold a meeting to talk about the meetings, and another to plan the meetings about the meetings.“ I dropped scrum last year.
54.
by gdiamos 5mo ago
I used to think evil killer robot discussions among AI researchers was an idea based in Hollywood, not science. Then I realized how effective the fear was at fundraising...
55.
by gdiamos 5mo ago
I originally thought evil killer robots discussions in AI labs was an idea out of Hollywood. Then I saw how effective it was at raising money.
56.
by gdiamos 7mo ago
I personally don't mind letting Claude write about work. You could spend 80% doing the work and 20% writing about it, or 99% doing the work and 1% copy-pasting Claude's writeup about it into a blog. There is nothing wrong with wri
57.
by gdiamos 7mo ago
One of my lessons in using different accelerators, whether they be different NVIDIA versions, or GPU->TPU, etc is that someone needs to do this work of indexing, partitioning, mapping, scheduling, and benchmarking. That work is labor int
58.
by gdiamos 7mo ago
Results as good as Qwen has been posting would seem to trigger a power struggle. I think companies that don’t navigate these correctly eventually lose.
59.
by gdiamos 7mo ago
It was inevitable.
60.
by gdiamos 8mo ago
This is why I like Dario as a CEO - he has a system of ethics that is not jus about who writes the largest check. You may not agree with it, but I appreciate that it exists.

More ›