By Baba Tamim
Blake Lemoine, the Google engineer fired for violating the company’s confidentiality policy, has now expressed concerns about the risks associated with AI-driven chatbots like Microsoft’s Bing AI.
The latest AI models, according to him, are the most potent technological advancement since the atomic bomb and can alter the course of history fundamentally.
“It’s essentially impossible to run the same experiments on it that I did on [Google’s] LaMDA. It bails to canned responses constantly,” Lemoine wrote in a tweet on Wednesday expressing concern after testing Microsoft’s Bing AI chatbot.
Lemoine fears the technology could be misused to distribute false information, political propaganda, or derogatory information about people of diverse racial and religious backgrounds.
Google fired Lemoine in July 2022 when he claimed the AI chat system that the company had been developing was “sentient” (has feelings).
‘Predicting a train wreck?’
Many have expressed concerns about Bing’s AI’s possible sentience since it was made public, echoing those he expressed last summer, Lemoine said in an op-ed for Newsweek on Monday, when he had not tested Microsofts’s Bing AI.
He expressed that the word “vindicated” doesn’t quite capture how I’ve felt in this situation.
“Predicting a train wreck, having people tell you that there’s no train, and then watching the train wreck happen in real time doesn’t really lead to a feeling of vindication,” said Lemoine.
“It’s just tragic,” he added.
Before being released to the general public, Lemoine noted that he would like to see AI put through more thorough testing to look for risks and the ability to manipulate consumers.
“I feel this technology is incredibly experimental, and releasing it right now is dangerous,” he stated.
The engineer repeated prior complaints that AI models had not undergone sufficient testing before being made public.
Meanwhile, tech moguls like Bill Gates hold opinions that differ from Lemoine’s.
AI models’ no threat’ – Bill Gates
Bill Gates, a co-founder of Microsoft, described the most recent AI models as “a generation-old technology” in an interview with the Financial Times, published on Thursday.
“The technology most people are playing with, it’s a generation old,” he said.
Although AI-powered chatbots like Bing can say some “crazy things,” according to Gates, this is primarily because users have turned provocation into a game by trying to find ways to trick the model’s programming into making a mistake.
Gates stated that existing AI models are “fine, [and] there’s no threat,” adding that “it’s not apparent who should be blamed “if you sit there and provoke a bit.”
Meanwhile, Lemoine’s latest comments and op-ed have received no rebuttal from Microsoft or Google. (interesting Engineering)