Viewing profile — gdiamos

How big is the open model? 30B?

Christopher Nolan beat you to it

vLLM is originally marketed as paged attention, but in hindsight, separating the web server and GPU process, continuous batching, kv caching / chunking, and a huge model library in…

I wish I could get a model to state its assumptions.

How do you ban melted sand?

I think we should shut it off. It would force US companies to build open models.

I want a hosted paper to be archived. That means that 10 years from now I don’t want think about making sure the hosting server is up. I also want it to have a standard format for …

thank god, these parameters are so confusing

as soon as you release a way of measuring it, you give LLMs a signal to optimize

Being on the review board comes with a promise to not be evil right?

No, I want arxiv to host the paper, not to review the paper. I wouldn't want my google drive to start telling me my paper was too sloppy. I just want a link.

I wonder if Amazon eventually gets cut out by 3D printing/replicators for imitable objects.

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:/…

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 fo…

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…

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…

Usually breakthroughs in computing lead to more usage of computing, not less.

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 g…

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 so…

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.

I don’t get it. That’s what I am using.

Think about the worst enterprise SaaS apps you have used…

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, usi…

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 us…