Top AI researcher dismisses AI ‘extinction’ fears, challenges ‘hero scientist’ narrative https://
bit.ly/3Jk73QJ #AI #MachineLearning #DeepLearning #LLMs #DataScience
SAFETY
-
Top AI Researcher Dismisses Extinction Fears and Hero Scientist Narrative
By
–
-
Neural Networks Fall Short: Fly Brain Complexity Reveals AI Limitations
By
–
Good reminder that vintage 2023 “neural networks” aren’t very neural.
— Gary Marcus (@GaryMarcus) 2 juillet 2023
Fly brain has 3,000+ distinct neuron types, all but entirely abstracted away from in current “neural networks”.
Fair bit of evidence for innate constraints, too, contra the AI/Ml zeitgeist. https://t.co/TrJSVcaNDQGood reminder that vintage 2023 “neural networks” aren’t very neural. Fly brain has 3,000+ distinct neuron types, all but entirely abstracted away from in current “neural networks”. Fair bit of evidence for innate constraints, too, contra the AI/Ml zeitgeist.
-
What Happens When AI Becomes Sentient Three Strategic Outlooks
By
–
What Happens When AI Becomes Sentient? 3 Strategic Outlooks
#AI #AIio #BigData #ML #NLU #Futureofwork @CRudinschi @AntonioSelas @alexjc @RobotLaunch @karpathy @andyjankowski @bobgourley @CadeMetz http://
ow.ly/QUI430svStr -
Decision importance based on human and societal impact
By
–
The importance of a decision is significantly based on how it affects you and other people.
-
Self-Supervised LLM Evaluation Framework for Deployment Monitoring
By
–
8/ Evaluations with No Labels – a framework for self-supervised evaluation of LLMs by analyzing their sensitivity or invariance to transformations on input text; can be used to monitor LLM behavior on datasets streamed during live model deployment.
-
AI Developers Ignore Non-Expressive Use Arguments Without Opposition
By
–
AI developers argue non-expressive use and nobody is challenging them on that! I see your arguments but nobody with money is doing anything…
-
Expected Utility Theory Limitations in Modern AI Systems
By
–
It looks like a restatement of expected utility theory, dealing with none of the places where I expect difficulty to lie in an EU paradigm, such as "getting any utility function whatsoever into an AI built along anything remotely resembling the modern paradigm" and "stating a
-
AI Can Achieve Both Accuracy and Transparency Simultaneously
By
–
AI Can Be Both Accurate and Transparent
#AI #AIio #BigData #ML #NLU #Futureofwork @TunstallAsc @StrategyFintech @TamaraMcCleary @TerenceLeungSF @psb_dc @thomaspower @vinod1975 @ylecun http://
ow.ly/MoqI30svSto -
Prompt Hijacking: Building Secure AI Systems
By
–
Prompt hijacking is a real thing. Build AI systems that take this into account.
-
AI Capabilities Beyond Simple Statistics and Human Understanding
By
–
It's as silly to say "AI is just predictive stats" as it would be to say "neurons are just electrochemical pulses". AI clearly can solve all sorts of very hard problems, which is amazing and awesome. AI capabilities are also, in many ways, increasingly beyond our comprehension.