Benjamin Riley (@benjaminjriley) / X

Benjamin Riley

22K posts

Benjamin Riley

@benjaminjriley

Founder of Cognitive Resonance, a new venture to improve understanding of human cognition and generative AI.

Austin, TX

buildcognitiveresonance.substack.com

Joined January 2010

Pinned
Benjamin Riley
@benjaminjriley
Aug 7, 2024
Today, Cognitive Resonance is releasing a new guide titled Education Hazards of Generative AI. This free resource clarifies common misconceptions about how AI works and warns against misuses of this technology in education. Please share this widely. cognitiveresonance.net/resources.html
86K
Benjamin Riley
@benjaminjriley
Aug 29, 2024
"Most concerning is the illusion that LLMs are retrieving information rather than constructing word associations. LLM responses are statistically likely rather than factually accurate. Sometimes these things correspond, but often they do not." E. Salvaggio
Challenging The Myths of Generative AI
From techpolicy.press
92K
Benjamin Riley
@benjaminjriley
Aug 20, 2024
So Chicago Public Schools has a guide for teachers urging them to use AI, apparently drafted with some help from AI for Education. Is there magic thinking a-plenty in it? Reader, don't you already know? Let's count the ways...
83K
Benjamin Riley
@benjaminjriley
Jun 14, 2020
What’s more, renaming our monuments to treasonous losers is like literally the easiest step in dismantling white supremacy. Gonna take harder work to integrate schools, end mass incarceration, reform criminal justice, and yes, pay reparations.
Benjamin Riley
@benjaminjriley
Aug 27, 2015
Cain, Abel, both good friends. Cain got into agriculture, bad business. But he murdered his competition. That's leadership. #TrumpBible
Benjamin Riley
@benjaminjriley
Aug 20, 2024
Replying to @benjaminjriley
4. "Students must use GenAI tools ethically." Well, fair enough. Except...what if...hear me out here...what if using AI in education itself presents a host of ethical questions that are ignored throughout this guidance document?
34K
Benjamin Riley
@benjaminjriley
Aug 20, 2024
Replying to @benjaminjriley
1. "Always verify the accuracy and appropriateness of GenAI-generated content before sharing." Once again, who has the time for this? What problem does this solve for teachers if they cannot trust the veracity of AI output?
16K
Benjamin Riley
@benjaminjriley
Aug 2, 2024
Large-language models can do none of these things. Literally none.
j⧉nus
@repligate
Aug 1, 2024
LLMs can perceive how intelligent you are, the quality of your character, the culture that shaped you, where you're headed, etc, against the backdrop and reference class of superimposed recorded history.
11K
Benjamin Riley
@benjaminjriley
Aug 20, 2024
Replying to @benjaminjriley
6. How about math? "Teacher manually designs and delivers interventions for individual students," BAD. "Students engage with GenAI tutors under teacher supervision to target and strengthen developing skills," GOOD. We're now veering into education malpractice.
45K
Benjamin Riley
@benjaminjriley
Aug 20, 2024
Replying to @benjaminjriley
2. "Conduct thorough reviews to ensure outputs...are free of unintended biases and align with our educational goals. Verify and assess the source information that GenAI outputs are relying upon." What? How? Can *anyone* do this? (No, is the answer.)
8.1K
Benjamin Riley
@benjaminjriley
Aug 27, 2024
An "education" technology we'd do well to focus more attention on: Air conditioning and heating.
Peter Greene
@palan57
Aug 27, 2024
All these debates about high-minded education ideas, and some cities can't even manage to put kids in a building that isn't torturous on a hot day.
9.8K
Benjamin Riley
@benjaminjriley
May 14, 2024
Large-language models frequently hallucinate, they cannot reason, and they have no capacity to understand the mental states of humans. The notion that they are anywhere close to being "the perfect teacher" is embarrassing.
Derya Unutmaz, MD
@DeryaTR_
May 13, 2024
OpenAI's GPT-4o has just revolutionized and dramatically democratized education! Now, every student can have a perfect teacher on any imaginable topic!
Mckay Wrigley
00:00
Mckay Wrigley
18K
Benjamin Riley
@benjaminjriley
Oct 12, 2024
The first seven minutes of this video are about as elegant an explanation of the limits of LLMs, and the power of human cognition (to generalize through abstraction), as you'll find. There's a lot to learn from @fchollet.
Machine Learning Street Talk
@MLStreetTalk
Oct 12, 2024
We just released @fchollet keynote talk from #AGI24. He challenges LLM hype, discusses ARC-AGI benchmark, and proposes merging deep learning with program synthesis for true AGI progress.
00:00
34K
Benjamin Riley
@benjaminjriley
Sep 17, 2024
It's fascinating how many scientists have privately acknowledged to me that Chomsky's theory of universal grammar is thoroughly discredited, but that they don't want to anger his fans by saying so. Tomorrow I'll be sharing the thoughts of a neuroscientist who is unafraid!
183K