As technology advances, industry experts say Google must decide whether to overhaul its search engine and make a full-fledged chat bot the face of its flagship service.
Google is reluctant to spread its technology widely because, like ChatGPT and similar systems, it can generate false, toxic, and biased information. LaMDA is only available to a limited number of people through an experimental app, AI Test Kitchen.
According to a memo seen by The Times, Google sees this as a struggle to deploy its advanced AI without harming users or society. At a recent meeting, an executive conceded that smaller companies would have fewer qualms about releasing these tools, but said Google must jump into the fray or the industry can continue without them, according to an audio recording of the meeting obtained by The Times .
Other companies have a similar problem. Five years ago, Microsoft released a chat bot called Tay that spat out racist, xenophobic, and otherwise dirty language, and was forced to immediately remove it from the web — never to return. Over the past few weeks, Meta has shut down a newer chatbot for many of the same reasons.
Executives said in the recorded session that Google intends to release the technology that powered its chat bot to outside companies as a cloud computing service, and that it could integrate the technology into simple customer support tasks. It will maintain its trust and safety standards for official products, but it will also release prototypes that don’t meet those standards.
It can limit these prototypes to 500,000 users and warn them that the technology may make false or offensive claims. Since its release on the last day of November, ChatGPT – which can produce similarly toxic material – has been used by over a million people.
“A cool demo of a conversation system that people can interact with over a few rounds and it feels overwhelming? It’s a good move, but it’s not what will really change society,” Zoubin Ghahramani, who heads the AI lab Google Brain, said in an interview with The Times last month, ahead of ChatGPT’s release. “It’s not something that people can reliably use on a daily basis.”