I like to think of the first talk on a work as the "world premiere." The fact that this one premiered simultaneously at multiple locations is even better. (For some context: a breakthrough in combinatorics regarding bounds on Ramsey numbers https://
x.com/Methemedantics
/status/1636562848364077057
…)
@thegautamkamath
-
Breakthrough in Combinatorics: Ramsey Numbers Bounds Advance
By
–
-
AI struggles with humor comprehension and understanding jokes
By
–
Bing Chat/GPT-4/Sydney gets the humour totally wrong. Too bad. I'm an AI optimist. I like to believe that, some day, there will be a tool powerful enough that can help me actually laugh at @ccanonne_'s jokes. https://
x.com/jdcmedlock/sta
/jdcmedlock/status/1636480393007489024
… -
Tim Roughgarden’s Algorithms Illuminated Book Review
By
–
Excited to find @Tim_Roughgarden
's Algorithms Illuminated in my mailbox! Took a quick look and seems like an excellent book for learning from (as opposed to a reference manual…). I like the intuitions, exercises, and conversational style. Will try it next time I'm teaching! -
GPT-3 Release Strategy Limited API Access Clarification
By
–
I was admittedly being a bit flippant. I don't think they fully released GPT-3 at all, only via limited API access
-
OpenAI’s Escalating Safety Messaging from GPT-2 to GPT-5
By
–
GPT-2: Too dangerous to release for a few months
GPT-3: Too dangerous to release for free
GPT-4: Too dangerous to even describe
GPT-5: Too dangerous to say it exists? -
GPT-4 Paper Rejected at NeurIPS: Product Ad Not Research
By
–
Strong majority says the GPT-4 paper shouldn't be accepted at NeurIPS. @BlancheMinerva said it succinctly "This is a product ad, not a research paper." It focuses only on what their magical box can do. This could be an OK paper IF how it was made is described (here or elsewhere)
-
Should GPT-4 Paper Be Accepted at NeurIPS?
By
–
Suppose you were reviewing the GPT-4 paper in its current state for, say, NeurIPS. Would you recommend acceptance or rejection? Why? https://
cdn.openai.com/papers/gpt-4.p
df
… -
Industry Labs: Redefining Leadership Roles in AI Development
By
–
For an academic career, that's certainly true. Since a professor is meant to be the "leader" of a group. But I wonder if that will be the case for industry labs going forward. Say, OpenAI would probably make great use of a "supporter" for GPT-5.
-
Why Industry Scales Large AI Projects Better Than Academia
By
–
Thomas mentions below that he thinks big academic groups ought to work on single, big projects. Instead, they seem to parallelize into many small projects. In contrast, big industry groups seem to more often work on single, big projects (see, e.g., GPT-4). Why? https://
x.com/shortstein/sta
/shortstein/status/1634675864276443136
…