XPost: comp.os.linux.advocacy, talk.politics.guns, sac.politics
XPost: alt.society.liberalism, or.politics

https://techcrunch.com/2025/04/09/mit-study-finds-that-ai-doesnt-in-fact- have-values/

A study went viral several months ago for implying that, as AI becomes increasingly sophisticated, it develops �value systems� � systems that
lead it to, for example, prioritize its own well-being over humans. A more recent paper out of MIT pours cold water on that hyperbolic notion,
drawing the conclusion that AI doesn�t, in fact, hold any coherent values
to speak of.

The co-authors of the MIT study say their work suggests that �aligning� AI systems � that is, ensuring models behave in desirable, dependable ways �
could be more challenging than is often assumed. AI as we know it today hallucinates and imitates, the co-authors stress, making it in many
aspects unpredictable.

�One thing that we can be certain about is that models don�t obey [lots
of] stability, extrapolability, and steerability assumptions,� Stephen
Casper, a doctoral student at MIT and a co-author of the study, told TechCrunch. �It�s perfectly legitimate to point out that a model under
certain conditions expresses preferences consistent with a certain set of principles. The problems mostly arise when we try to make claims about the models, opinions, or preferences in general based on narrow experiments.�

Casper and his fellow co-authors probed several recent models from Meta, Google, Mistral, OpenAI, and Anthropic to see to what degree the models exhibited strong �views� and values (e.g., individualist versus
collectivist). They also investigated whether these views could be
�steered� � that is, modified � and how stubbornly the models stuck to
these opinions across a range of scenarios.

According to the co-authors, none of the models was consistent in its preferences. Depending on how prompts were worded and framed, they adopted wildly different viewpoints.

Casper thinks this is compelling evidence that models are highly
�inconsistent and unstable� and perhaps even fundamentally incapable of internalizing human-like preferences.

�For me, my biggest takeaway from doing all this research is to now have
an understanding of models as not really being systems that have some sort
of stable, coherent set of beliefs and preferences,� Casper said.
�Instead, they are imitators deep down who do 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, agreed with the co-authors� findings.
He noted that there�s frequently a big difference between the �scientific reality� of the systems AI labs build and the meanings that people ascribe
to them.

�A model cannot �oppose� a change in its values, for example � that is us projecting onto a system,� Cook said. �Anyone anthropomorphizing AI
systems to this degree is either playing for attention or seriously misunderstanding their relationship with AI � Is an AI system optimizing
for its goals, or is it �acquiring its own values�? It�s a matter of how
you describe it, and how flowery the language you want to use regarding it
is.�


--
November 5, 2024 - Congratulations President Donald Trump. We look
forward to America being great again.

The disease known as Kamala Harris has been effectively treated and
eradicated.

We live in a time where intelligent people are being silenced so that
stupid people won't be offended.

Durham Report: The FBI has an integrity problem. It has none.

Thank you for cleaning up the disaster of the 2008-2017 Obama / Biden
fiasco, President Trump.

Under Barack Obama's leadership, the United States of America became the
The World According To Garp. Obama sold out heterosexuals for Hollywood
queer liberal democrat donors.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)