“Memorization can enhance genuine reasoning abilities, enabling models to generalize better to new and varied problems. 📌 Memorization Dependency: LLMs exhibit high accuracy on training puzzles, primarily due to memorization. When these models encounter slightly altered versions
LLMs don’t do formal reasoning – and that is a HUGE problem
“Bold idea in this paper. Proves how LLMs game getting high scores on prestigious benchmarks like AlpacaEval 2.0, Arena-Hard-Auto, and MT-Bench. Proves how even a “null model” that always outputs a constant response (irrelevant to input instructions) can cheat automatic
The Shift from Models to Compound AI Systems – The Berkeley Artificial Intelligence Research Blog
“An article in the Wall Street Journal in which I express my opinion on the limitations of LLMs and on the potential power of new architectures capable of understanding the physical world, have persistent memory, can reason and can plan: four features of intelligent behavior that” / X
Overview of strong human intelligence amplification methods — LessWrong
“Yes, LLMs are going to hit a wall in a year But it’s also true that they are already smarter than most humans The last mile in AI automation is not intelligence, it’s plumbing!” / X
“This is something I tell companies & organizations. When you talk to folks at the frontier labs, many genuinely believe they are on the road to AGI and genuinely think it is coming quite soon. You don’t have to believe they are right, but you shouldn’t assume its all marketing.” / X
“@DanielColson6 @OpenAI @GoogleDeepMind @AnthropicAI @TIME One thing I regularly tell journalists who ask me for story ideas is that they should take the AGI aspirations of the frontier labs literally, rather than covering AGI as a kind of laughable marketing stunt.” / X
The AI Boom Has an Expiration Date – The Atlantic
Opinion | AI, Aging and Shifts Globalization Will Shock the American Economy – The New York Times
“Large Language Models don’t reason. Says Apple Paper. 🤔 Reveals LLMs lack robust mathematical reasoning, relying on pattern matching rather than genuine conceptual understanding. Now generally till now, LLMs have shown impressive performance on grade-school math tasks like
Apple Engineers Show How Flimsy AI ‘Reasoning’ Can Be | WIRED
Leave a Reply