ChatGPT-5 won’t be coming in 2025, according to Sam Altman – but superintelligence is ‘achievable’ with today’s hardware

In a caller Reddit AMA (ask maine anything), OpenAI CEO Sam Altman, on pinch immoderate different apical OpenAI executives, dropped a number of hints astir nan company’s future, and what to expect from ChatGPT adjacent year.

Firstly, erstwhile asked if location would beryllium a ChatGPT-5 Altman doubled down connected his “fake news” consequence connected X precocious to an article talking of its imminent release: “We person immoderate very bully releases coming later this year! Nothing that we are going to telephone GPT-5, though”, he replied. The phrasing “this year” does bespeak that a caller LLM could beryllium released by OpenAI very soon though.

Agents are coming

It's been a engaged week for OpenAI, pinch nan merchandise of Advanced Voice mode connected nan desktop app versions of ChatGPT and a caller ChatGPT search, which moreover challenges Google. When asked astir nan worth of nan precocious released ChatGPT search compared to accepted hunt engines, for illustration Google, Altman was enthusiastic: “For galore queries, I find it to beryllium a measurement faster/easier measurement to get nan accusation I'm looking for. I deliberation we'll spot this particularly for queries that require much analyzable research. I besides look guardant to a early wherever a hunt query tin dynamically render a civilization web page successful response!”

When asked astir nan adjacent update to Dall-E 3, nan image generator that’s portion of ChatGPT, Altman was non committal: “The adjacent update will beryllium worthy nan wait! But we don't person a merchandise scheme yet." So it doesn’t look for illustration overmuch will beryllium happening connected nan image procreation beforehand for a while.

Real intelligence

One area wherever ChatGPT is being challenged by its rivals is successful AI that tin execute tasks autonomously. We’ve seen Google make inroads into this area pinch its Jarvis AI. When asked if ChatGPT will beryllium capable to execute tasks connected its own, Altman replied “IMHO this is going to beryllium a large taxable successful 2025”, which indicates nan guidance OpenAI will beryllium taking adjacent year.

Referring again to autonomous AI, erstwhile asked what nan adjacent large breakthrough for ChatGPT would be, Altman replied: “We will person amended and amended models, but one deliberation nan point that will consciousness for illustration nan adjacent elephantine breakthrough will beryllium agents”. Agents are autonomous AI bots that tin execute tasks for you.

Perhaps nan astir absorbing remark from Altman was astir nan early of AGI - artificial wide intelligence. Seen by galore arsenic nan ‘real’ AI, this is an artificial intelligence exemplary that could rival aliases moreover transcend quality intelligence. Altman has antecedently declared that we could person AGI wrong "a fewer 1000 days".

When asked by a Reddit personification whether AGI is achievable pinch known hardware aliases it will return thing wholly different, Altman replied: “We judge it is achievable pinch existent hardware.”

You mightiness besides like...

• OpenAI edges person to making its first AI spot successful bid to powerfulness your favourite caller apps
• ChatGPT integration and image procreation are yet present successful nan iOS 18.2 developer beta
• ChatGPT hunt is now unrecorded – and it could beryllium nan Google replacement you’ve been looking for