Significant harm (multiple deaths, spread of fascism, decline of democracy etc). Net is a separate question.
@garymarcus
-
Hybrid AI: A 30-Year Argument for Advanced Intelligence
By
–
exactly!! that’s a big part of why I have been arguing for hybrid AI for 30 years 🙂
-
Smug Archive Cherry-Picks Technology Successes Ignoring Failures
By
–
The Smug Archive
cherry-picks
ignores nukes & tech that has actually cause major harm (submachine guns, pesticides, etc)
ignores importance of regulation (seatbelts, airplane certification etc)
ignores tech that failed (full self-driving, Facebook M, dirigibles etc) https://
x.com/ylecun/status/
/ylecun/status/1614674331023413248
… -
AI brittleness persists despite regular patching approaches
By
–
in older approaches to AI, people applied to bandages, to address the brittleness, but they were embarrassed about it and knew it wasn’t really working. In new approaches, people regularly apply patches, without shame, but the brittleness remains.
-
Modern AI Brittleness Problem Nobody Seems Notice
By
–
The problem with old-fashioned symbolic AI was that it was brittle. The problem with modern AI is that it is brittle, and hardly anyone seems to care.
-
LLMs: Commercial Viability Yet Problematic Detour from Trustworthy AI
By
–
with respect to the criteria I laid out, it has. Check the 10 issues I raised. Which have been solved? 1/10? In ten years, people will see LLMs commercially viable (yet problematic) but a detour relative to trustworthy AI. And even commercial value is still speculative.
-
Open Architecture vs Fair Data Compensation in AI
By
–
Seems to me that there are two orthogonal issues:
– should your architecture be open? – where do you get your data from/are contributors justly compensated? Both are tricky. But virtue in one doesn’t logically entail virtue in the other. -
LLMs Lack True Compositional Reasoning Capabilities
By
–
Read http://
rebooting.ai; it went to press just before LLMs became popular but lays out many challenges for genuine intelligence. I don’t see how LLMs resolve any of them. It’s not compositional and can’t reason, so it’s just a giant but superficial statistical approximator -
Superficial AI Progress Masks Deeper Problems and Misinformation Risks
By
–
not in the areas that I have emphasized for over two decades. in my view we have tons of superficial progress, but little progress in the deeper problems that would get us to genuine and trustworthy intelligence. And are entering a mess of misinformation we are ill prepared for
-
Distribution Shift Remains Core Challenge in AI Development
By
–
I predicted driverless cars would be harder than expected, language comprehension would be poor, reasoning would be poor; they still are. I based these predictions around distribution shift; importance of that is now recognized. You put words in my mouth that I did not say