So many people are confused about the relation between human cognitive errors and LLM hallucinations that I wrote this short explainer: Humans say things that aren't true for many different reasons
• Sometimes they lie
• Sometimes they misremember things
• Sometimes they fail
@garymarcus
-
Human Cognitive Errors vs LLM Hallucinations Explained
By
–
-
Moral compass revealed through actions taken
By
–
Their actions have made their moral compass clear.
-
AI Service Restrictions: Google’s Machine Learning Model Policy
By
–
it’s not just about using prompts and responses in some cases, eg “You may not use the Services to develop machine-learning models or related technology” per Google. or “You may not access or use, or help another person to access or use, our Services in the following ways: … To
-
AI Code Generation Tools ToS: Training Data Ethics Paradox
By
–
I was today years old when I learned that it is ToS violation to use most generative AI tools to write code for your own AI. Even though, of course, generative AI tools for writing code literally wouldn’t exist without tons of human-written code that wasn’t intended as training
-
GPT-4 vs Next Generation: $1000 Improvement Challenge
By
–
$1000 says nothing publicly testable by six months from today will be obviously better overall than GPT4 as GPT-4 was relative to GPT-3. Stuart judges. If I win, proceeds go to @Wikimedia
. I win if there are a few dimensions of clear improvement (as you could argue with Claude -
LLM Limitations and Paradigm Shift: Industry Leaders Echo Long-Held Criticisms
By
–
Why should I change my schtick? Because Demis, Yann, and Sam are now echoing what I have been saying all along? Re: hype, plateaus, the limits of LLMs, the need for common sense, and for a paradigm shift. That’s ok, I will keep saying what I believe is true even if has become
-
GPT-4 plateau: AI innovation stalled, new ideas urgently needed
By
–
It wasn’t a pause; it was a failure. Nobody stopped because of external pressure. They did their best to move forward, past GPT-4—and couldn’t. The time for new ideas is now.
-
OpenAI’s $100B funding: staged payment and infrastructure allocation
By
–
his 100b is staged, most not upfront, and yes he may never see it. (also it goes to infrastructure, not all directly to OAI)
-
Will the generative AI bubble burst within twelve months?
By
–
Why and how the Generative AI bubble might pop within the next twelve months. I lay out some analysis and scenarios at my newsletter, link below
-
AI Essay Publication Announcement on Marcus
By
–
Ha – i literally was writing an essay on exactly this topic just now! Will post later today at Marcus on AI… https://
t.co/QAUhUT1cxU