ok fair, i overstated a bit. but look the whole field though pure pretraining would do well and then in November many reports emerged saying that recent models like Orion (relabeled as 4.5) and Behemoth didn’t meet expectations. And GPT-5 still hasn’t emerged. Did that change
@garymarcus
-
Critique des prédictions d’IA concentrées sur quelques années
By
–
It’s not true of you; it is true of anyone who speaks as if 95% to 100% of the probability distribution is within the next few years.
-
AGI Always Two Years Away: The Self-Fulfilling Prophecy
By
–
My guess, with near-100% confidence: many people will keep saying AGI is “two years away” every year from now going forward until the year in which it is actually achieved, literally no matter what happens, and no matter when it actually arrives.
-
What Will AI Accomplish in Five Years That Seems Impossible Today?
By
–
Gauntlet thrown: What specific, testable things will AI be able to do in 5 years that are truly impressive yet not achievable now?
-
The AGI Two-Year Away Prediction Cycle Explained
By
–
my guess btw is that many people will keep saying AGI is two years away every year from now until when it is actually achieved, literally no matter what happens.
-
Future of LLMs: Domain-Specific Fine-Tuning Over True Generality?
By
–
Wouldn’t it be funny if the future was a bunch of domain-general LLMs each fine-tuned in a different way to different domain, with none that was actually generally reliable? Narrow AI reinvented, and renamed! attn: @rao2z
-
AI Model Weaknesses: Image Generation and Logical Reasoning Failures
By
–
it has other weaknesses, eg image generation is wacky, and i crushed it in tic-tac-toe by merely replacing the labels x and o with y and z
-
Grok Model Demonstrates Solid Basic Arithmetic Performance
By
–
in fairness, though, grok seems to be more solid on basic arithmetic than some other models.
-
AI Intelligence Parity with Humans by 2025 Prediction
By
–
When I offered @elonmusk a million dollar bet that AI wouldn’t be smarter than the smartest human by the end of 2025 I was confident I would win the bet. I figured I could make a list with dozens, maybe hundreds, of things ordinary people could do that eluded many or most