Despite differences within the AI community, pretty much all AI experts find the notion that Google's LaMDA might be sentient completely ridiculous
No, LaMDA is not sentient. Not even slightly. … Blaise Aguera y Arcas, polymath, novelist, and Google VP, has a way with words.
The Road to AI We Can TrustGary Marcus
Context & Ripple Effects
The debate followed Google's decision to place an engineer on leave after he publicly concluded that LaMDA was sentient, documented in coverage of the engineer's leave. The expert consensus reported here sharply separates persuasive chatbot output from evidence of consciousness.
Later coverage returned to the same boundary, arguing that LLMs lack the physiological states needed for subjective sensations in a critique of LLM sentience claims. That makes the episode an early test of how Google and the AI field explain humanlike systems without endorsing humanlike interpretations.
First-order effects
Google's LaMDA is publicly framed by the cited experts as a language model rather than a sentient entity, undercutting the claim associated with its suspended engineer.
Blaise Aguera y Arcas and other AI experts become visible counterweights to a narrative driven by the chatbot's apparently humanlike conversation.
Second-order effects
Google's AI communications face greater pressure to distinguish fluent model behavior from claims about inner experience, because employee and public interpretations can diverge from the company's technical position.
Researchers and commentators gain a clearer incentive to challenge consciousness claims with criteria beyond conversational output, as the later physiological critique illustrates.
Third-order effects
As conversational AI becomes more convincing, governance of AI companions is likely to center increasingly on preventing anthropomorphic claims from being mistaken for evidence of sentience.
The episode points to a durable split between product experiences designed around natural dialogue and scientific standards for attributing consciousness.
The trend: Humanlike language interfaces are making AI-companion governance and clear boundaries around sentience claims a recurring industry issue.
honestly if this system wasn't just a stupid statistical pattern associator it would be like a sociopath, making up imaginary friends and uttering platitudes in order to sound cool. https://twitter.com/...
Nonsense on Stilts. A hot take on all this “OMG. LaMDA is Sentient” mania—and why the AI community is pretty united on this one. https://garymarcus.substack.com/ ...
“If the media is fretting over LaMDA being sentient (and leading the public to do the same), the AI community categorically isn't.” listen to @GaryMarcus - he knows the score https://garymarcus.substack.com/ ...
From a man (or, more accurately, three men and a woman) who know whereof they speak: https://twitter.com/... Some general good advice: https://twitter.com/...
The Google AI story rings eerily similar to Ted Chiang's novella “The Lifecycle of Software Objects,” a story about a company that develops sophisticated digital animals ("digients") that are raised through computer-mediated interactions with employees. 1/
“Blake LeMoine, originally charged with studying how “safe” the system is, appears to have fallen in love with LaMDA, as if it were a family member or a colleague. (Newsflash: it's not; it's a spreadsheet for words.)” https://garymarcus.substack.com/ ...
He seems to treat LaMDA “as if it were a family member or a colleague. (Newsflash: it's not; it's a spreadsheet for words.) To be sentient is to be aware of yourself in the world; LaMDA simply isn't. It's just an illusion.” https://garymarcus.substack.com/ ...
@FlandersBlanche @Sifill_LDF It's a nonsense moral panic pushed by a reporter duped by a guy who clearly wants attention. The quality of LaMDA is interesting. The debate over sentience is utter nonsense and a waste of time. I think this is on point: https://garymarcus.substack.co…
Not sure what Turing would say, but I don't think the Turing Test itself is meaningful 👉relies on human gullibility 👉it can easily be gamed 👉advances in it have not historically led to advances in AI 👉essay I wrote about in 2014 still applies: https://www.newyorker.com/... https:…
https://garymarcus.substack.com/ ... Bad take. It is theoretically possible to make an AI that is as intelligent as a human but has no sense of self and just tries to complete sentences. LaMDA is not that, not even close. But you can't dismiss an AI just because it's ‘only autoco…