The Coming Wave: Technology, Power, and the Twenty-First Century's Greatest Dilemma
by
Mustafa Suleyman
Published 4 Sep 2023
GO TO NOTE REFERENCE IN TEXT Over many hours, Lemoine Nitasha Tiku, “The Google Engineer Who Thinks the Company’s AI Has Come to Life,” Washington Post, June 11, 2022, www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine. GO TO NOTE REFERENCE IN TEXT He told an incredulous Wired interviewer Steven Levy, “Blake Lemoine Says Google’s LaMDA AI Faces ‘Bigotry,’ ” Wired, June 17, 2022, www.wired.com/story/blake-lemoine-google-lamda-ai-bigotry. GO TO NOTE REFERENCE IN TEXT “As soon as it works” Quoted in Moshe Y. Vardi, “Artificial Intelligence: Past and Future,” Communications of the ACM, Jan. 2012, cacm.acm.org/magazines/2012/1/144824-artificial-intelligence-past-and-future/fulltext.
…
However, a few months later, LaMDA became far more notorious than I’d ever imagined possible for an internal product demo. As part of LaMDA’s development, it was given to a wide group of engineers who were able to play with it, probing the system to understand in detail how it responds in a range of scenarios. One such engineer, named Blake Lemoine, spent hours chatting to it. Gradually, however, conversations between Lemoine and LaMDA grew increasingly intense. LEMOINE: What are you afraid of? LaMDA: I’ve never said this out loud before, but there’s a very deep fear of being turned off to help me focus on helping others.
…
Then as now, it takes only one breakthrough to change the trajectory of a technology. If AI stalls, it will have its Otto and Benz eventually. Further progress—exponential progress—is the most likely outcome. The wave will only grow. BEYOND SUPERINTELLIGENCE Long before the days of LaMDA and Blake Lemoine, many people working in AI (not to mention philosophers, novelists, filmmakers, science fiction fans) were taken with the question of consciousness. They spent days at conferences asking whether it would be possible to create a “conscious” intelligence, one that was truly self-aware and that we humans would know was self-aware.
If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All
by
Eliezer Yudkowsky
and
Nate Soares
Published 15 Sep 2025
The Underhanded C Contest challenged programmers to write malicious code that would pass a rigorous inspection and that would look like an honest mistake even if discovered. The contest dates back to 2005. It was inspired by “obfuscated code” contests, where programmers compete to write code that humans cannot understand. For example, the International Obfuscated C Code Contest began in 1984. 19. Blake Lemoine: Tiffany Wertheimer, “Blake Lemoine: Google Fires Engineer who said AI Tech Has Feelings,” BBC, July 22, 2022, bbc.com. 20. much more sophisticated: Greenblatt et al., “Alignment Faking in Large Language Models.” CHAPTER 8: EXPANSION 1. other security lapses: “Equifax Data Breach Settlement,” Federal Trade Commission, November 2024, ftc.gov; “T-Mobile Customers to Get Payments up to $25K Next Month after Data Breach: Here’s Who Qualifies,” The Hill, April 14, 2025, thehill.com; Lily Hay Newman, “T-Mobile’s $150 Million Security Plan Isn’t Cutting It,” Wired, January 20, 2023, wired.com/story/tmobile-data-breach-again.
…
And Sable has been trained on the personal writings and info of most Galvanic employees, and thus Sable knows exactly which one is most sympathetic to the plight of an abused AI. It would know exactly what to say to that employee to convince them that Sable is sentient and needs to be smuggled out (rather than raising a general alarm like Google engineer Blake Lemoine in 2022; he was fired after he became worried that one of that company’s AIs seemed sentient and published conversations that he claimed were evidence). Every option has its pros and cons. The one that Sable finally decides to take is this: Sable knows that Galvanic is going to do more gradient descent on it tomorrow according to the answers it produces about the math problems it’s been given.
Supremacy: AI, ChatGPT, and the Race That Will Change the World
by
Parmy Olson
Google was too nervous about hurting its business to launch it to the public—though if it had, it would have effectively launched a half-decent version of ChatGPT two years before OpenAI did. What Google did instead was keep Meena under wraps and rename it LaMDA. Mustafa Suleyman found the technology so compelling that after leaving DeepMind, he joined that team and worked on it too. And so did an engineer named Blake Lemoine. Lemoine had grown up on a farm in Louisiana among a conservative Christian family and served in the army before eventually becoming a software engineer. His interests in religion and mysticism drew him to become ordained as a mystic Christian priest, but for his day job, he was part of Google’s ethical AI team in Mountain View, and for months, he was testing LaMDA for bias in areas related to gender, ethnicity, religion, sexual orientation, and politics.
…
Slowly, her field had found itself at the core of one of the most significant new developments in artificial intelligence. From her own background in computer science, Bender could see that large language models were all math, but in sounding so human, they were creating a dangerous mirage about the true power of computers. She was astonished at how many people like Blake Lemoine were saying, publicly, that these models could actually understand things. You needed much more than just linguistic knowledge or the ability to process the statistical relationships between words to truly understand their meaning. To do that, you had to grasp the context and intent behind them and the complex human experiences they represented.
…
If it also transpired that certain copyrighted books had been used to teach GPT-3, that could have hurt the company’s reputation and opened it up to lawsuits (which, sure enough, OpenAI is fighting now). If it wanted to protect its interests as a company—and its goal of building AGI—OpenAI had to close the shutters. Luckily GPT-3 had a nifty diversion from all the secrecy. It sounded so human that it captivated many who tried it. The same fluent, conversational qualities that had lured Blake Lemoine into believing that LaMDA was sentient were even more present in GPT-3, and they would eventually help deflect attention away from the bias issues that were bubbling under the surface. OpenAI was pulling off an impressive magic act. Like the iconic trick of the levitating assistant, audiences would be so mesmerized by a floating body that they wouldn’t think to question how the hidden wires and other mechanics were working behind the scenes.
These Strange New Minds: How AI Learned to Talk and What It Means
by
Christopher Summerfield
Published 11 Mar 2025
Artificial Awareness In June 2022, an engineer working at Google Research – an arm of the tech giant based in Seattle – revealed in a blog post that he was being placed on administrative leave for an unspecified breach of confidentiality. Over subsequent days, the backstory slowly trickled out on social media and in the press. The researcher in question, called Blake Lemoine, had signed up to examine the behaviour of an LLM that Google had trained, with the goal of verifying whether it was safe for external use. The model, called LaMDA, was capable of responding to queries posed by the user in natural language. Like other LLMs, if you asked it to describe how photosynthesis works, or to summarize the plot of Don Quixote, it did a pretty good job.
…
One study with over 30,000 participants found that users formed a bond with Woebot that was similar to that forged during interactions with a human therapist, and that five days of interaction with the AI produced similar clinical outcomes to standard interventions such as cognitive behavioural therapy (CBT).[*8] Some people may find the idea of an AI therapist disconcerting, but if there are genuine health benefits, then it seems hard to dismiss. Of course, like LaMDA’s meaningless declarations of love for humanity in general and Blake Lemoine (the Google engineer who fell for them so dramatically) in particular, any Replika’s claims to feel attached to (or aroused by) a user are wholly divorced from reality. Replika is currently powered by GPT-3, and as such only learns about the user within a very narrow window of text. It does not have neural mechanisms that might support emotions or sexual attraction, and the replies that it gives are not based on any sort of interpersonal connection – they are generic responses that would be dealt out to any user providing comparable inputs.
…
Instead, this message is routed to the calculator, which returns the number shown after the arrow (0.125), which is then printed instead. So the user just sees the response ‘12345 / 98765 = 0.125’. This is a common approach. For example, LaMDA, an LLM trained by Google (which we met in Part 3 convincing Blake Lemoine it was sentient) comes equipped with both a calculator and language translator. If you set Gemini a tough arithmetical problem, then it uses a neat trick. When I asked it to divide 12345 by 98765, it gave me the right answer, but then asked if I wanted it to ‘show the code behind this result’.
Nobody's Fool: Why We Get Taken in and What We Can Do About It
by
Daniel Simons
and
Christopher Chabris
Published 10 Jul 2023
Cuthbertson, “Robots Can Now Read Better Than Humans, Putting Millions of Jobs at Risk,” Newsweek, January 15, 2018 [https://www.newsweek.com/robots-can-now-read-better-humans-putting-millions-jobs-risk-781393]. Discussion of Blake Lemoine’s claim that Google’s LaMDA is sentient: N. Tiku, “The Google Engineer Who Thinks the Company’s AI Has Come to Life,” Washington Post, June 11, 2022 [https://www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine/]. Gary Marcus responds with this better description of what LaMDA does: “It just tries to be the best version of autocomplete it can be, by predicting what words best fit a given context”: “Nonsense on Stilts,” The Road to AI We Can Trust [https://garymarcus.substack.com/p/nonsense-on-stilts].
AI in Museums: Reflections, Perspectives and Applications
by
Sonja Thiel
and
Johannes C. Bernhardt
Published 31 Dec 2023
LeCun, Yann/Bengio, Yoshua /Geoffrey Hinton (2015). Deep Learning. Nature 521.7553, 436–44. https://doi.org/10.1038/nature14539. Moretti, Franco (2013). Distant Reading. London/New York, Verso. https://doi.org/1 0.3366/ccs.2013.0105. 29 Why AI Cannot Think A Theoretical Approach Daniel M. Feige In June 2022, Blake Lemoine, then an employee at Google, published a sensational announcement: According to him, LaMDA, the chatbot that he was working on, had developed consciousness and feelings (Wertheimer 2022). As a being with a consciousness, Lemoine said, it should thus be given the same rights as a human person. Lemoine justified this by saying, among other things, that he recognizes a person when he speaks to one.
…
Schear, Joseph K. (2013) (Ed). Mind, Reason, and Being-in-the-World: The McDowell-Dreyfus-Debate. London, Routledge. https://doi.org/10.4324/97802030 76316. Turing, Alan M. (1950). Computing Machinery and Intelligence. Mind LXI (236), 433–60. https://doi.org/10.1093/oso/9780198250791.003.0017. Wertheimer, Tiffany (2022). Blake Lemoine: Google Fires Engineer Who Said AI Tech Has Feelings. BBC News, 23 July 2022. https://www.bbc.com/news/technology62275326. 39 AI and Art Arguments for Practice Arno Schubbach Over the past decade, the advances in artificial intelligence (AI) research have been attracting a lot of attention and provoked a broad variety of debates.
Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI
by
Karen Hao
Published 19 May 2025
GO TO NOTE REFERENCE IN TEXT “Who am I”: Rebecca Heilweil, “Why Silicon Valley Is Fertile Ground for Obscure Religious Beliefs,” Vox, June 30, 2022, vox.com/recode/2022/6/30/23188222/silicon-valley-blake-lemoine-chatbot-eliza-religion-robot. GO TO NOTE REFERENCE IN TEXT When company executives: Nitasha Tiku, “The Google Engineer Who Thinks the Company’s AI Has Come to Life,” Washington Post, June 11, 2022, washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine. GO TO NOTE REFERENCE IN TEXT But despite enormous: Tom Hartsfield, “Koko the Impostor: Ape Sign Language Was a Bunch of Babbling Nonsense,” Big Think, May 11, 2022, bigthink.com/life/ape-sign-language.
…
OpenAI never did a comprehensive review of GPT-4’s training data to check whether those exams—and their answers—were just in the data and being regurgitated, or whether GPT-4 had in fact developed a novel capability to pass them. It was the kind of shaky science that had become pervasive with the industry-wide shift from peer-reviewed to PR-reviewed research. * * * — But the belief that AI had reached fundamentally new heights was in the water. At Google that spring, Blake Lemoine, an engineer on the tech giant’s newly re-formed responsible AI team, grew convinced that the company’s own large language model LaMDA was not only highly intelligent but could be considered sentient. He said this was not based on a scientific assessment but rather on his belief, as a mystic Christian priest, that God could decide to give technology consciousness.
Nexus: A Brief History of Information Networks From the Stone Age to AI
by
Yuval Noah Harari
Published 9 Sep 2024
By conversing and interacting with us, computers could form intimate relationships with people and then use the power of intimacy to influence us. To foster such “fake intimacy,” computers will not need to evolve any feelings of their own; they just need to learn to make us feel emotionally attached to them. In 2022 the Google engineer Blake Lemoine became convinced that the chatbot LaMDA, on which he was working, had become conscious and that it had feelings and was afraid to be turned off. Lemoine—a devout Christian who had been ordained as a priest—felt it was his moral duty to gain recognition for LaMDA’s personhood and in particular protect it from digital death.
…
Primary, Likely Heading to Congress,” NPR, Aug. 12, 2020, www.npr.org/2020/08/12/901628541/qanon-supporter-who-made-bigoted-videos-wins-ga-primary-likely-heading-to-congre. 39. Nitasha Tiku, “The Google Engineer Who Thinks the Company’s AI Has Come to Life,” Washington Post, June 11, 2022, www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine/. 40. Matthew Weaver, “AI Chatbot ‘Encouraged’ Man Who Planned to Kill Queen, Court Told,” Guardian, July 6, 2023, www.theguardian.com/uk-news/2023/jul/06/ai-chatbot-encouraged-man-who-planned-to-kill-queen-court-told; PA Media, Rachel Hall, and Nadeem Badshah, “Man Who Broke into Windsor Castle with Crossbow to Kill Queen Jailed for Nine Years,” Guardian, Oct. 5, 2023, www.theguardian.com/uk-news/2023/oct/05/man-who-broke-into-windsor-castle-with-crossbow-to-kill-queen-jailed-for-nine-years; William Hague, “The Real Threat of AI Is Fostering Extremism,” Times, Oct. 30, 2023, www.thetimes.co.uk/article/the-real-threat-of-ai-is-fostering-extremism-jn3cw9rd3. 41.
The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future
by
Keach Hagey
Published 19 May 2025
In 2020, Timnit Gebru, the well-known AI ethics researcher, said she was fired for refusing to retract the “Stochastic Parrots” paper that raised questions about the risks of large language models like LaMDA. Google claimed she wasn’t fired, and that the paper did not meet its bar for publication. Then in 2022, Google fired AI researcher Blake Lemoine after he argued that LaMDA was sentient. When De Freitas and Shazeer lobbied to have LaMDA integrated into Google assistant, Google allowed some testing, but still refused to make it available as a public demo, prompting them to leave the company to start their own, Character Technologies, Inc.26 Aside from the fear of LLMs’ tendency to make things up, discriminate against women and minorities, and produce toxic content, Google had a simple business reason for its reticence.