interesting @jackclarkSF assessment of the effects of OpenAI announcing GPT-2 but not concurrently releasing it https://
importai.substack.com/p/import-ai-37
5-gpt-2-five-years-later
…
SAFETY
-
GPT-2 Release Strategy: OpenAI’s Delayed Disclosure Impact
By
–
-
California AI Bill Requires Pre-Creation Model Safety Certification
By
–
Not a joke: California's AI bill would require companies to certify that models are safe before they're even created.
-
Evals Broken by Marketing Incentives and Goodhart’s Law
By
–
evals were not broken when evals werent marketing goodhart, bad incentives
-
Vibrations Help Unwinding Bolts: Matroid Ensures Aircraft Safety
By
–
Vibrations are surprisingly good at unwinding bolts. This is why spacecraft and aircraft manufacturers use Matroid to check their bolt safety wires, which tie bolts down. pic.twitter.com/WFFD6VtYzP
— Reza Zadeh (@Reza_Zadeh) 3 juin 2024Vibrations are surprisingly good at unwinding bolts. This is why spacecraft and aircraft manufacturers use Matroid to check their bolt safety wires, which tie bolts down.
-
Personal Universes: Solution to Multi-Agent Value Alignment
By
–
Personal Universes: A Solution to the Multi-Agent Value Alignment Problem Roman V. Yampolskiy : https://
arxiv.org/abs/1901.01851 #AGI #ASI #ArtificialSuperIntelligence -
Art, AI exploitation and controversy: ethical implications explored
By
–
If it's still Art, and the abuse/explotation is a known component of it, then the outrage and backlash is expected… Willing participant getting millions of views out of the controversy.
-
Broader evaluation coverage needed for AI models
By
–
but even these evals are already fairly specific, would be interesting to see a broader eval coverage.
-
Corporate Exploitation and Ethical Concerns in AI Development
By
–
There are so many misguided people on both sides and it's turning nasty *because* of the underlying corporate exploitation. You can't just be innocent on the AI "side", as there are many sociopaths and assholes that pleasure in abusing others no matter what — part of their fun.
-
Alignment’s Hard Problem: Two Decades of Misrepresentation
By
–
Alternate take: Alignment started with the obvious observation that alignment might be hard, but the implications of this idea were outside the Overton Window, so a variety of different Pretend Serious groups spent two decades seizing on whoever could be misread for an excuse.
-
Octopus Thought Experiment: Critical Analysis of AI Understanding
By
–
This is a decent summary of the octopus thought experiment from Bender & @alkoller 2020, with two glaring exceptions, right at the start: https://
techcrunch.com/2024/06/01/wtf
-is-ai/
… >>