The latest upgrade of OpenAI’s chatbot is making both its free and paid versions seem more emotionally in-tune with users, Wired reports. The life-like repartee it’s capable of comes via a new, more powerful AI model called GPT-4o.
Demos show the new version conversing in a pithy, natural-sounding voice and responding flirtatiously when complimented. “Oh stop it, you’re making me blush,” was one line the chatbot deployed. It also correctly guessed the mood of an OpenAI staffer who showed it a selfie: “It looks like you’re feeling pretty happy and cheerful,” the bot guessed.
On Monday, OpenAI’s chief technology officer Mira Murati announced that ChatGPT’s enhanced capabilities would be rolled out “to everyone” over the next few weeks.
[See more: A quarter of Macao residents now use the likes of ChatGPT]
In a blog post on the same day, OpenAI’s CEO Sam Altman wrote of the upgrade, “It feels like AI from the movies; and it’s still a bit surprising to me that it’s real.” He added that the chatbot’s “human-level response times and expressiveness turns out to be a big change.”
Some AI researchers have warned that deploying tech that mimics and responds to human emotions could have unwelcome implications long-term. Researchers at Google DeepMind found that AI chatbots could become extremely persuasive and addictive, for example – to users’ psychological detriment.
Most tech companies have so far stopped short of anthropomorphising their offerings for reasons along these lines. Bing’s Copilot chatbot, for instance, reminds its users that it is just a computer program when asked certain questions. Other companies have forged ahead regardless, even offering the likes of AI boyfriends and girlfriends.
