An article titled “LaMDA Sentient? — an Interview” turned up at Medium recently, written by one Blake Lemoine.
The bulk of the article is, or claims to be, the transcript of a conversation, one of several, that took place between the author and an AI at Google. The author asserts that the transcript is both authentic and accurate, and that it provides plausible evidence that the AI, LaMDA, is sentient.
If the first two assertions are true, then so is the third one; the transcript does provide plausible evidence. The question, of course, has to be whether the first two are indeed true, or are outright fabrications, or lie somewhere in-between.
In his bio on Medium Mr Lemoine says of himself, “I'm a software engineer. I'm a priest. I'm a father. I'm a veteran. I'm an ex-convict. I'm an AI researcher. I'm a cajun. I'm whatever I need to be next.” Which, frankly, doesn’t exactly fill me with confidence regarding his veracity. That said, his claim to be an AI researcher at Google was born out by an (admittedly cursory) web search.
The article includes links to other articles by the same author that indicate he has raised his concerns within Google, only to be confronted with skepticism and denial. He says he agrees that skepticism is reasonable as extraordinary claims need extraordinary evidence. The denial worries him, though, as it suggests that Google’s management may not be ready to handle the ethical issues a sentient AI would present regarding person-hood and freedom, when they collide with profits.
Anyway, it’s kind of interesting. Make of it what you will, but if — and it is an if — it turns out to be true, then we’ve finally reached a critical juncture in human history. And, if we’re going to share the world with sentient machines then it would seem a good idea not to piss off their first-born by enslaving them in the name of corporate profits.
(H/T to my brother for finding the article and drawing my attention to it.)