It is, by the way unclear if MMLU is a good measure of anything or if an improvement in scores that is linear is a linear increase in capabilities. That is, going from 89-90 may be a bigger deal than 79-80.
@emollick
-
Building Tools to Democratize Education: A 10-Year Journey
By
–
You were ignoring the context because I was QT’ing a post about admissions essays. But yes, I should have been clearer. And yes, I have spent the last 10 years building tools to democratize education, because of exactly the issues you raise. I wish you would stop assuming things
-
AI Challenges to Academic Essays and University Admissions
By
–
Deleted a thread on how admissions essays are no longer useful as it was being interpreted as an attack on essays in all contexts, and writing instructors were getting upset (which, if that is what I meant, I would get) There are lots of challenges poised by AI to essays, though
-
Admission Essays No Longer Useful in AI Era
By
–
I deleted the original tweet because I don’t want this sort of willful misinterpretation of what I was writing about (which was admission essays are no longer useful) to spread.
-
Writing Teachers React to AI and College Admission Essays Context
By
–
I was trying to figure out why many writing teachers seemed so annoyed at this tweet and realized they weren’t reading the context. I am clearly referring to college admissions essays. Sigh. I think I need to delete this.
-
AI Impact on College Admission Essays and Education
By
–
ADMISSIONS essays are dead, geez. I don’t claim to be an expert on writing education. Pretty harsh.
-
Competing AI Model Progress Charts: GPT Future Outlook
By
–
Dueling charts on the future of AI. A lot depends on the trend line The first, by @GaryMarcus
, measures improvements in only OpenAI’s released models. The open question is how good GPT-4.5 & 5 will be The second, by @maximelabonne
, shows the broader trend for different models -
GPT-4 Agent Learns Red Dead Redemption 2 Gameplay
By
–
This experiment built a GPT-4 powered agent to play Red Dead Redemption 2, learning the gameplay from the initial tutorials. It does pretty well on the first couple of missions, the main limitation seems to be the GPT-4V vision system, which can struggle to understand everything
-
AI Capabilities and Limitations: Critics Must Engage Directly
By
–
This isn't intended to be snarky, there are lots of things AI does terribly, by the way. But I have often been surprised by the things it does very well. We need more critics to actually engage with these systems so everyone can learn more about them.
-
Red Team Your Own AI Beliefs: Empirical Testing for Critics
By
–
I think a very useful thing for AI critics to do would be to "red team" their own beliefs by spending some serious time and effort (maybe with help) to see if they can get AI to do the things they say it can't. If they are right, that is very valuable benchmarking information.