Raising Money for a Hard Tech Startup

From Ben's Personal Wiki
Revision as of 19:50, 6 June 2024 by Ben (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

I recently read a story[1] about some blatant plagiarism in the machine learning world. Here's a few bullet points explaining what happened:

  1. Three Stanford undergrads released a model on Huggingface called "Llama3-V". Two of them were non-technical and were brought on essentially just to aggressively promote the model's capabilities through blog posts
  2. The model essentially just took the weights from OpenBMB's MiniCPM-V model[2] and added random noise
  3. This was discovered by some people in the open-source community who smelled something fishy
  4. So far nothing has happened, but hopefully Stanford expels the students, or at least applies harsh disciplines

This triggered me a bit and left me feeling more than a little disgusted with the incentives at play for hard tech fundraising.

References