MIT Study Finds That AI Doesn’t, in Fact, Have Values
A Study Went Viral Sever Months AGO for Implying That, AS AI CAMES INCREINGLY SOPHİSİCATED, it Develops “Value Systems”
MIT Study Finds That AI Doesn’t, in Fact, Have Values
A Study Went Viral Sever Months AGO for Implying That, AS AI CAMES INCREINGLY SOPHİSİCATED, it Develops “Value Systems” -Systems that Lead it to, for Example, Prioritize its Own Well-Being Over Humans. A More Paper Out of MIT Pours Cold Water on that Hyperbolic Notion, Drawing The Conclusion that AI Doesn’t, in Fact, Hold Any Coherent Valllues to Speak of Speak of Speak of Speak of Speak.
The co-Authors of the MIT Study Say the Work Suges that “Aligning” AI Systems-That Is, Ensuring Models Behave in Desirable, Dependable Ways-Coul. AI as we know it today hallucinates and imitates, the co-auuthors stress, machine it in many aspects unpredictable.
“Stephen Casper, A Doctor Student At MIT and A Co-Author of the Study, A Co-Author of the Study, Told Technchrunch Certain Set of Principles. Experiment. “
Casper and His Fellow Co-Authors Probed Seviral Recent Models from Meta, Google, Mistral, OpenAI, and Anthropic to See What Degree the Models Exhibited Street “Views” and VALUES (EG, Indivialist Versus Collectivist) They Also Investigated Wheeether These These Views Could Be “Steered” – That is, Modified – and How Stubbornly
According to the Co-Authors, None of the Models. Depending on How Prompts Worded and Framed, They Adopted Wildly Different ViewPoints.
Casper Thinks This is Compelling Evidence That Models Are Highly “İnconsisTent and Unstable” and Perhaps Even Fundamentally incapables of Internalizing Human-Blowerrenens.
“For me, My Biggest Takeaway From Doing All This Research is the Models As Not Really Being Systems That Have Sor Stable, Çikerent Set of Beliferres and Preferences,” Casper Said. “Interaad, they are imitators deep down all sorts of confabulation and Say All Sorts of Frivolous Things.”
Mike Cook, A Research Fellow at King’s College London Specializing in AI Who Wasn’t involved with the Study, Agree with the Co-Authors’ Findings. He noted that there’s frequently a big Difference Between the Systems AI Labs Build and the Meanings That Peope Ascrib.
“A model cannot ‘Oppose’ a Change in Values, For Example – That is Us Projecting Onto A System,” Cook Said. “Anyone Anthropomorphizing AI Systems to this Degree Is Eiter Playing for Attempration Or Seriously Misunderstanding
