With AI growing at a rapid rate, many are awaiting the creation of a sentient AI. This is why many were quick to assume the worst when a transcript allegedly proved that a Google AI had already reached sentience.
However, as it turns out, the killer transcript was heavily edited to be more “enjoyable to read.” Granted, transcripts are a chore to go through, but the fact that such a high-profile AI story was edited for the purpose of fear-mongering is as terrifying as if it was real.
Edited for your enjoyment
Former Google engineer Blake Lemoine made wild claims that a Google AI, named LaMDA, is sentient. Lemoine recalled that he had numerous conversations with the program, stating that they had conversations about religion and law.
A transcript that details conversations between Lemoine and the AI was leaked to The Washington Post. However, new details reveal that the transcript was also edited so that it would be more enjoyable to read.
According to the leak, the transcript was edited down in order to summarize the situation. Conversations with the AI supposedly were comprised of nine interviews that took place at different times on two different days. Even the specific order of the transcript was “shuffled around” because conversations with LaMDA would lead to tangents or meandering. This was done so as to not waste anyone’s time.
Both Lemoine and the contributor also claim that LaMDA changes its personality a lot during these conversations, making a transcript hard, adding another excuse to edit it down. It seems that each conversation with the AI would lead to a personality change, making this endeavor a very difficult one.
"The nature of the relationship between the larger LaMDA system and the personality which emerges in a single conversation is itself a wide-open question.”
Edited but faithful
Despite how sketchy it is to hear that a transcript was edited, Lemoine and his associate claim that it "was faithful to the content of the source conversations." Obviously, many people will have a hard time believing that this is the case since no one would believe that an edited transcript is legitimate.
Google is not a fan of all this AI talk as they claim that several engineers have had “conversations” with LaMDA and think Lemoine’s claims are baseless. According to the billion-dollar company: “The evidence does not support his claims."
"Hundreds of researchers and engineers have conversed with LaMDA and we are not aware of anyone else making the wide-ranging assertions, or anthropomorphizing LaMDA, the way Blake has,” Google told Business Insider.
It will be interesting to see if this is the start of more sentient AI getting revealed, or if this is nothing more than a baseless rumor. Either way, it’s not boring.