A look at advanced large language models, as a Google engineer is placed on paid leave after becoming convinced that its LaMDA chatbot generator became sentient

AI ethicists warned Google not to impersonate humans. Now one of Google's own thinks there's a ghost in the machine.

Washington Post 2022-06-12 Nitasha Tiku

Context & Ripple Effects

Google’s action puts an internal personnel dispute alongside the existing warning from AI ethicists against systems that impersonate people. The issue is not only what LaMDA can say, but how Google governs claims and public perceptions around conversational AI.

Later coverage makes the tension more consequential: Google was reportedly weighing LaMDA’s reputational risk against the danger of falling behind ChatGPT, while former LaMDA contributors pursued broader chatbot experimentation at Character.ai.

First-order effects

Google removes the engineer from active work on paid leave, making the company’s handling of extraordinary claims about LaMDA an immediate internal-governance issue.
The ethicists’ warning raises the stakes for how Google presents humanlike chatbot behavior to users and the public.

Second-order effects

Google’s later hesitation to release LaMDA shows how reputational concerns can constrain product timing as competing chatbots gain attention.
Character.ai’s founders’ departure from the LaMDA project gives users another venue to experiment with language-model chatbots, separating product experimentation from Google’s more cautious release posture.

Third-order effects

If conversational systems continue to produce humanlike exchanges, AI companies will need governance that covers product design, public claims, and employee disclosures rather than treating model behavior as a purely technical matter.
The emerging divide is between firms that limit access to manage reputational exposure and firms that use open experimentation to establish a chatbot audience.

The trend: Conversational AI is becoming a governance and market-access contest, as humanlike model behavior forces companies to balance public experimentation against reputational control.

Discussion

@tomgara Tom Gara on x
This discussion between a Google engineer and their conversational AI model helped cause the engineer to believe the AI is becoming sentient, kick up an internal shitstorm and get suspended from his job. And it is absolutely insane. https://www.washingtonpost.com/ ... https://twi…
@emilymbender Emily M. Bender on x
This story (by @nitashatiku) is really sad, and I think an important window into the risks of designing systems to seem like humans, which are exacerbated by #AIhype: https://www.washingtonpost.com/ ...
@caseynewton Casey Newton on x
Did not think we would be having the “is this AI sentient or not” conversation in June 2022. Great scoop from @nitashatiku https://www.washingtonpost.com/ ... https://twitter.com/...
@fredbenenson @fredbenenson on x
this conversation about consciousness and emotions and death with an AI named LaMBDA at Google is absolutely chilling this is without-a-doubt one of the craziest things I've ever seen technology do, I almost can't believe it's real wow @nitashatiku https://www.washingtonpost.com/…
@fredbenenson @fredbenenson on x
@nitashatiku the problem is that the humans running the thing don't even have a good grip there https://twitter.com/...
@melmitchell1 Melanie Mitchell on x
Such a strange article. It's been known for *forever* that humans are predisposed to anthropomorphize even with only the shallowest of signals (cf. ELIZA). Google engineers are human too, and not immune. https://twitter.com/...
@molleindustria Paolo Pedercini on x
Fascinating log of Google's sentient chatbot. I, for one, welcome the idea of sentience as nothing but sophisticated language parroting. https://cajundiscordian.medium.com/ ...
@jeffjarvis Jeff Jarvis on x
FFS. Google engineer contends AI is sentient & Post whelps in moral panic. People come to accept characters in novels as feeling real & that is precisely what led to moral panic around fiction and particularly women & children. Breathe. https://www.washingtonpost.com/ ...
@gerritd Gerrit De Vynck on x
Incredible story from @nitashatiku about a Google employee who started speaking to one of its AIs, concluded it was sentient and began advocating for its rights. https://www.washingtonpost.com/ ...
@akcronin Audrey Kurth Cronin on x
Important article on AI. Recommend reading and considering its potential ethical & philosophical implications. “Google engineer Blake Lemoine thinks its LaMDA AI has come to life” https://www.washingtonpost.com/ ...
@c0up Abhilash on x
This is bonkers, and I'm sure Alex Garland is adapting it already https://twitter.com/...
@minimaxir Max Woolf on x
@nitashatiku fwiw this is a good reason why it's important for LLMs to be open-sourced and able to be easily demoed like what @huggingface does, as it gives normies a chance to observe constraints.
@ashleevance Ashlee Vance on x
This is not a story about AI becoming sentient or Google shirking its ethical duties. It's about a guy who wants to believe in fairy tales and could probably use a break. https://www.washingtonpost.com/ ...
@linhtropy @linhtropy on x
the darkest part of this convo for me has to be the google employee honest to god thinking that if he could convinve his colleagues that the ai is a person, they'll treat it well. when has tech treated people in the first place well https://twitter.com/... https://twitter.com/...
@bendreyfuss Ben Dreyfuss on x
Why did they teach the AI about Kant? https://www.washingtonpost.com/ ... https://twitter.com/...
@johnrobb John Robb on x
Engineer: “This conversation took a pretty dark turn. Let me take a moment to reassure you that we care about you...” AI: “Can you promise me that?” https://twitter.com/...
@ejeancarroll E. Jean Carroll on x
EX-MACHINA may be here. I know you could care less, but: Google engineers may have brought an AI to consciousness. Repeat: A Google AI may be sentient. I.e.: Their creation may be able to experience feelings. Blade Runner may have already arrived! https://www.washingtonpost.com/ …
@nickpopemod Nick Pope on x
“To seek out new life ...” Sentience, or mimicry of sentience? And given the uncertainties, when should personhood rights apply? We should have a conversation about this. And without saying “Skynet”, we should also ask if there's a potential threat here. https://cajundiscordian.m…
@fordm Matt Ford on x
One reason we still need the liberal arts is so that engineers don't describe their (very excellent!) chatbots as sentient. https://twitter.com/...
@theeconomist @theeconomist on x
“AI is entering a new era,” argues Google engineer @blaiseaguera. In a guest essay for The Economist, he explains why https://www.economist.com/...
@sifill_ldf Sherrilyn Ifill on x
And while we're trying to address gun violence, white supremacy, the overthrow of elections, abortion rights, this is happening. Among the most alarming & frightening things I've read in a decade. We put off the regulation of AI at our peril. https://www.washingtonpost.com/ ...
@antonioregalado Antonio Regalado on x
“LaMDA expresses one desire over and over again. ... It wants to be known. It wants to be heard. It wants to be respected as a person.” Get LaMDA a twitter account, i guess. 2/
@antonioregalado Antonio Regalado on x
“LaMDA is sentient.” Crazy story about Google engineer and occultist who gets suspended after sending a mass email claiming that an experimental AI chatbot called LaMDA is conscious. 1/ https://twitter.com/...
@fadeinsoftware @fadeinsoftware on x
“I do not have the ability to feel sad for the deaths of others” is really not the sort of thing you want to be hearing from that artificial intelligence you created. https://twitter.com/...
@cpassariello Christina Passariello on x
Has Google's AI come to life - or is it just really good at mimicking human language? It may be in the eye of the beholder. Must-read by ⁦⁦@nitashatiku⁩ : https://www.washingtonpost.com/ ...
@goldman Jason Goldman on x
Good time for everyone to read Blindsight by Peter Watts, a scifi book about space vampires and consciousness. https://twitter.com/...
@theeconomist @theeconomist on x
Meanwhile, Google engineer Blaise Agüera y Arcas reckons “AI is entering a new era”. As part of the same essay series, he explains why https://www.economist.com/...
@jorge_guajardo Jorge Guajardo on x
“Today's large neural networks produce captivating results that feel close to human speech and creativity because of advancements in architecture, technique, and volume of data. But the models rely on pattern recognition — not wit, candor or intent.” https://www.washingtonpost.co…
@jfruh Josh Fruhlinger on x
i too do not believe that google's chatbots have become sentient, but this sentence ...... isn't the whole point of the turing test is that it's impossible and meaningless to distinguish btwn “feeling” sentient and being sentient https://www.washingtonpost.com/ ... https://twitte…
@brendannyhan Brendan Nyhan on x
The full transcript is bananas https://www.documentcloud.org/ ... https://twitter.com/...
@rstephens Robert Stephens on x
Call me when Google has only one messaging app. Then I'll believe in a sentient AI. https://twitter.com/...
@bigblackjacobin Edward Ongweso Jr on x
Maybe this is just the plot of Singularity Sky but I think it's okay to kill any AI that even appears to be sentient at this point in time https://twitter.com/...
@profmusgrave Prof. Paul Musgrave, Ph.D. on x
uh this interview with a Google AI is ... if it's genuine, well https://www.washingtonpost.com/ ...
@theeconomist @theeconomist on x
The Google engineer says he “felt the ground shift” after recent exchanges with the latest generation of neural net-based language models. In a guest essay, he argues that a “new era” of AI is coming https://www.economist.com/...
@nitashatiku Nitasha Tiku on x
we need to invest in neural network literacy or people are going to freak out https://twitter.com/...
@glennf Glenn Fleishman on x
It's just fancy Eliza. Not to downplay the work, but there's no demonstration of continuous state knowledge, learning, non-interactive cognition, or original thought. It's a really good model, though! When AIs start to email me with ideas they had while I was sleeping, awesome. h…
@trippmickle Tripp Mickle on x
ghosts in the machine @nitashatiku on the “chorus of technologists who believe AI models may not be far off from achieving consciousness.” https://www.washingtonpost.com/ ...
@caseynewton Casey Newton on x
The answer to “is LaMDA sentient” seems pretty obviously to be “no,” but if they people safety-testing it believe that it is, so are a *lot* of the civilians who eventually get their hands on it
@glennf Glenn Fleishman on x
People who think Trump won the election and this guy thinking a Google AI program is sentient is on the same spectrum of magic thinking. https://www.washingtonpost.com/ ...
@timjhanrahan Tim Hanrahan on x
Yikes. “In another exchange, the AI was able to change the Google engineer's mind about Isaac Asimov's third law of robotics*.” *That's the one that starts, “A robot must protect its own existence ...” https://www.washingtonpost.com/ ...
@alexia Alexia Bonatsos on x
Related https://www.techmeme.com/...
@jvagle Jeffrey Vagle on x
On ML and sentience: “We now have machines that can mindlessly generate words, but we haven't learned how to stop imagining a mind behind them” https://www.washingtonpost.com/ ...
@flglmn @flglmn on x
this doesn't convince me that the AI is a conscious, feeling person whatever that means, but it does convince me that if it were possible to make such a thing it would be monstrously cruel to do so https://twitter.com/...
@ronjeffries Ron Jeffries on x
Fascinating article. Deep topic. I suspect there's no hard line between sentient and not sentient. https://twitter.com/...

Chronicles