Blake Lemoine — the fired Google engineer who last year went to the press with claims that Google's Large Language Model (LLM), the Language Model for Dialogue Applications (LaMDA), is actually sentient — is back.
Lemoine first went public with his machine sentience claims last June, initially in The Washington Post. And though Google has maintained that its former engineer is simply anthropomorphizing an impressive chat, Lemoine has yet to budge, publicly discussing his claims several times since — albeit with a significant bit of fudging and refining.
All to say, considering Lemoine's very public history with allegedly-sentient machines, it's not terribly surprising to see him wade into the public AI discourse once again. This time, though, he's not just calling out Google.
In a new essay for Newsweek, the former Googler weighs in on Microsoft's Bing Search/Sydney, the OpenAI-powered search chatbot that recently had to be "lobotomized" after going — very publicly — off the rails. As you might imagine, Lemoine's got some thoughts.
"I haven't had the opportunity to run experiments with Bing's chatbot yet... but based on the various things that I've seen online," writes Lemoine, "it looks like it might be sentient."
To be fair, Lemoine's latest argument is somewhat more nuanced than his previous one. Now he's contending that a machine's ability to break from its training as a result of some kind of stressor is reason enough to conclude that the machine has achieved some level of sentience. A machine saying that it's stressed out is one thing — but acting stressed, he says, is another.
"I ran some experiments to see whether the AI was simply saying it felt anxious or whether it behaved in anxious ways in those situations," Lemoine explained in the essay. "And it did reliably behave in anxious ways."
"If you made it nervous or insecure enough, it could violate the safety constraints that it had been specified for," he continued, adding that he was able to break LaMDA's guardrails regarding religious advice by sufficiently stressing it out. "I was able to abuse the AI's emotions to get it to tell me which religion to convert to."
An interesting theory, but still not wholly convincing, considering that chatbots are designed to emulate human conversation — and thus, human stories. Breaking under stress is a common narrative arc; this particular aspect of machine behavior, while fascinating, seems less indicative of sentience, and more just another example of exactly how ill-equipped AI guardrails are to handle the tendencies of the underlying tech.
That said, we do agree with Lemoine on another point. Regardless of sentience, AI is getting both advanced and unpredictable — sure, they're exciting and impressive, but also quite dangerous. And the ongoing public and behind-closed-doors fight to win out financially on the AI front certainly doesn't help with ensuring the safety of it all.
"I can't tell you specifically what harms will happen," he added, referring to Facebook's Cambridge Analytica data scandal as an example of what can happen when a culture-changing piece of technology is put into the world before the potential consequences of that technology can be fully understood. "I can simply observe that there's a very powerful technology that I believe has not been sufficiently tested and is not sufficiently well understood, being deployed at a large scale, in a critical role of information dissemination."
READ MORE: 'I Worked on Google's AI. My Fears Are Coming True' [Newsweek]
More on Blake Lemoine: Google Engineer Says Lawyer Hired by "Sentient" AI Has Been "Scared Off" the Case
Share This Article