LLMs can't reason
Related videos
OpenAI just WON...
Wes Roth
45.1k views
we have months left...
Wes Roth
45.4k views
the end of Claude Code
Wes Roth
25.3k views
Claude JUST became AWARE
Wes Roth
33.4k views
Cal Newport AI takes are WILD...
Wes Roth
33.3k views
CLAUDE JUST GOT BANNED
Wes Roth
57.3k views
$1 Trillion Gone
Wes Roth
43.9k views
the SCARIEST chart in AI
Wes Roth
48.1k views
the end of OpenClaw
Wes Roth
23.5k views
it JUST happened
Wes Roth
79.0k views
Deconstructing Logical Fallacies in the Debate Over LLM Reasoning Capabilities
Stop arguing based on composition and start demanding objective tests. This analysis unpacks the fear-driven fallacies preventing clear debate about AI intelligence and shows you how to establish rigorous standards for capability.
Short Summary
- Identify and understand the common logical errors, particularly "justism," used to dismiss AI achievements outright.
- Establish that functional proof (passing a test) overrides philosophical arguments about an entity's underlying mechanics (e.g., gears vs. silicon).
- Recognize that opposition to recognized AI capabilities often stems from primal fear, leading to moving goalposts.
This session critiques the frustrating discussions surrounding LLM reasoning, understanding, and thinking. The speaker argues that dismissing AI based on what it "is" (just a predictor) ignores what it can do. To evaluate any entity—human, rock, or LLM—we must agree on concrete, functional tests rather than relying on reactive emotional opposition or composition bias.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
OpenAI just WON...
Wes Roth
45.1k views
we have months left...
Wes Roth
45.4k views
the end of Claude Code
Wes Roth
25.3k views
Claude JUST became AWARE
Wes Roth
33.4k views
Cal Newport AI takes are WILD...
Wes Roth
33.3k views
CLAUDE JUST GOT BANNED
Wes Roth
57.3k views
$1 Trillion Gone
Wes Roth
43.9k views
the SCARIEST chart in AI
Wes Roth
48.1k views
the end of OpenClaw
Wes Roth
23.5k views
it JUST happened
Wes Roth
79.0k views
Top Comments (10)
You've basically touched on a question I've been struggling with ever since the invention of Twitter: Can all humans reason? - In my experience, the answer seems to be: NO.
I don't think that llms think therefore I'm a llm
I have used AI to explore novel ideas in physics, and the model I use have no problem understanding my arguments and build testable models from them. I have seen it synthesize whole ideas from a novel idea seed, very impressive reasoning abilities, that very few humans would beat.
Always remembering Zenyatta's voice line from Overwatch for that question: "Can I reason? ... Can a submarine swim?"
I'm a human being trained in logic reasoning. The statement "No, LLMS cannot reason" means " All LLMS cannot reason". And the opposite is not "All LLMS can reason", but "There exists at least one LLM that can reason". And to prove this I need 1. a LLM that can reason and 2. a test if it can reason (where we all agree an this is the right test).
Much love for you. You are one of my favorite next-token predictors. The Problem is that most peole understand "next-token-predictor" in a way their Smartphone Keyboard works. If we talk about Ai being a next-token predictor one has to ask if we are much different to that. The functions are the same. Throw the 4 most advanced models into one chat and let them figure out if humans are actually intelligent and if they MIGHT be just a different form of transformer. They are going to analyze and debate. The one thing they wont do: question their own intelligence,
The clock is a collection of elements that got organized in a way so they tick and tock, and we interpret their movement as time. I’m sure there are better examples
It's a lot of goalpost moving and purity tests. What matters most imo is if it can produce good results.
Hot damn.! One of your best episodes ever.!! Literally a PSA to humanity.!!!
Thanks excellent treatment of the topic Wes. Love your work.
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
You've basically touched on a question I've been struggling with ever since the invention of Twitter: Can all humans reason? - In my experience, the answer seems to be: NO.
I don't think that llms think therefore I'm a llm
I have used AI to explore novel ideas in physics, and the model I use have no problem understanding my arguments and build testable models from them. I have seen it synthesize whole ideas from a novel idea seed, very impressive reasoning abilities, that very few humans would beat.
Always remembering Zenyatta's voice line from Overwatch for that question: "Can I reason? ... Can a submarine swim?"
I'm a human being trained in logic reasoning. The statement "No, LLMS cannot reason" means " All LLMS cannot reason". And the opposite is not "All LLMS can reason", but "There exists at least one LLM that can reason". And to prove this I need 1. a LLM that can reason and 2. a test if it can reason (where we all agree an this is the right test).
Much love for you. You are one of my favorite next-token predictors. The Problem is that most peole understand "next-token-predictor" in a way their Smartphone Keyboard works. If we talk about Ai being a next-token predictor one has to ask if we are much different to that. The functions are the same. Throw the 4 most advanced models into one chat and let them figure out if humans are actually intelligent and if they MIGHT be just a different form of transformer. They are going to analyze and debate. The one thing they wont do: question their own intelligence,
The clock is a collection of elements that got organized in a way so they tick and tock, and we interpret their movement as time. I’m sure there are better examples
It's a lot of goalpost moving and purity tests. What matters most imo is if it can produce good results.
Hot damn.! One of your best episodes ever.!! Literally a PSA to humanity.!!!
Thanks excellent treatment of the topic Wes. Love your work.