Loading video...

Video Failed to Load

Go Home

On the pod: "Constrained Adaptive Rejection Sampling" with UCSD CSE professor Loris D'Antoni. Hear how symbolic AI experts have navigated the LLM era and why the future of AI code generation depends on program synthesis and formal methods.

18,106 views • 3 months ago •via X (Twitter)

6 Comments

Ndea's profile picture
Ndea3 months ago

@lorisdanto Watch the full interview:

Ndea's profile picture
Ndea3 months ago

@lorisdanto Read the paper:

Loris D'Antoni's profile picture
Loris D'Antoni3 months ago

@ucsd_cse Thanks for hosting me! This was a lot of fun

Mikhail Rogov's profile picture
Mikhail Rogov3 months ago

@ucsd_cse @lorisdanto interesting. do you think formal methods move into the generator, or stay as review layer after LLM output?

nonceense's profile picture
nonceense3 months ago

@ucsd_cse @lorisdanto Saw this last week: a team at a fintech startup spent 6 months fine-tuning GPT for contract generation, then switched to constraint-based synthesis in 3 weeks. The constraints were worth more than the parameters.

Joel Kreager's profile picture
Joel Kreager3 months ago

@ucsd_cse @lorisdanto Or just tell it to read the processor spec and dump out the most optimized op codes it can.

Related Videos