SMI = Superhuman Machine Intelligence
RESEARCH
-
Programming Language Journey: Assembly to Python and Compiler Development
By
–
1. Assembly (8080, later 68000, x86, …)
2. Basic
3. Pascal
4. Modula-2
5. C
6. Lisp/Scheme
7. Ada
8. Perl
9. Miranda
10. C++
11. Self
12. Cecil
13. Modula-3
14. Java
15. Javascript
16. Python
17. Go
Fun! (Written/worked on compilers or interpreters for 6, 10, 11, 12, 13, 14) -
Loss Function Finiteness in Langevin Dynamics Analysis
By
–
(2)The Loss function defined in @UCIrvine paper also appears to become infinite while we ensure the loss is finite and results in valid Langevin dynamics.
-
Diffusion Models in Function Spaces: Key Differences from Prior Work
By
–
Yes, we are aware of paper on diffusion models in function spaces from @UCIrvine There are two main differences from our work: (1) We assume data in function space vs. other work: data discretized with IID Gaussian noise. Latter not well-defined as resolution increases. See Fig
-
Imperfection Filter Makes AI-Generated Images Indistinguishable From Reality
By
–
But what if we added an 'imperfection filter' to these images, making them indistinguishable from real art or reality? 2/3
-
OpenAI Recruits Engineers for Unprecedented AI Impact
By
–
great engineers can have more impact on the trajectory of AI than ever before. if you like making computers do amazing things at unprecedented scale and with unprecedented difficulty, please consider joining OpenAI!
-
ChatGPT’s Attention Problem: Irony of Focus in AI Era
By
–
#ChatGPT and #GPT3 are awesome. But the are distracting. How cosmically ironic that the paper that started all this transformer stuff was titled "Attention Is All You Need". Yes it is true, after #GPT has robbed us of our ability to focus, attention is all we need. Funny, right?
-
Language Models Can Be Conditioned to Avoid Controversy Through RLHF
By
–
And yet they can be conditioned to be boring and non controversial through RLHF.
-
Genome-Scale Language Models Learn SARS-CoV-2 Evolutionary Landscape
By
–
Tomorrow @argonne
's Arvind Ramanathan shares with our community how his team built genome-scale language models that can learn the evolutionary landscape of SARS-CoV-2 genomes. This research won the 2022 ACM Gordon Bell Special Prize. Join us! https://
hubs.li/Q01CV9rF0 -
GPT Implementation in 60 Lines of NumPy and JAX
By
–
Late to the party but "GPT in 60 Lines of NumPy" / picoGPT is nicely done: https://
jaykmody.com/blog/gpt-from-
scratch/
…
– good supporting links/pointers
– flexes some of the benefits of JAX: 1) trivial to port numpy -> jax.numpy, 2) get gradients, 3) batch with jax.vmap
– inferences gpt-2 checkpoints