The AI of the future won’t just be a chatbot — it’ll be, according to OpenAI CEO Sam Altman, outfitted with incredibly detailed information about its users.
In an interview with the MIT Technology Review, Altman suggested that AI should be working for its users harder than even the hardest-working human executive assistant, and would know absolutely everything about whoever is using it.
Speaking to the magazine between a series of events at Harvard, which were hosted in part by the venture capital firm Xfund, the OpenAI cofounder said that the best use of AI would be a “super-competent colleague that knows absolutely everything about my whole life, every email, every conversation I’ve ever had, but doesn’t feel like an extension.”
Obviously, AI isn’t there yet. In the same interview, Altman referred to ChatGPT as “incredibly dumb,” and in a separate interview with The Harvard Crimson, he promised that GPT-5, his firm’s highly-anticipated update to its large language model (LLM), would not be.
“The most important thing to say about GPT-5 versus GPT-4 is just that it’ll be smarter,” the Stanford dropout told the Harvard student paper. “It’s not that it’s better in this way or that way.”
“The fact that we can make progress across all model behaviours and capabilities at once — I think that’s the miracle,” he continued.
As has been evident for a while now, talking a big game about the future of AI is very much a part of Altman’s job description.
In the roughly 18 months since OpenAI unleashed ChatGPT upon the world, he’s repeatedly discussed his fears of how badly things could go awry with the technology while insisting his company is creating the lesser of AI evils.
Within a few months of the release of GPT-4 in March 2023, however, expectations had begun bucking up against reality — which Altman, to his credit, now jokes about.
“I think it’s awesome that for two weeks, everybody was freaking out about ChatGPT-4, and then by the third week, everyone was like, ‘Come on, where’s GPT-5?'” Altman quipped during a discussion with MIT president Sally Kornbluth soon after his Harvard talks. “I think that says something legitimately great about human expectation and striving and why we all have to make things better.”
As always, the CEO remained cryptic about when exactly GPT-5 might be dropping, refusing to provide a timeline.
As MIT Tech notes, he responded to a reporter’s question about when GPT-5 would be released during his Cambridge press tour by simply saying “yes.”