On X, one of OpenAI's AI researchers suggests that the firm would have things to tell us very soon. Should we see this as an imminent announcement for ChatGPT-5? The question may in any case arise.
Le logo de ChatGPT // Source : ilgmyzin via Unsplash
Should we expect an upcoming deployment of a new version of Chat-GPT? A year after the introduction of GPT-4, the question may begin to arise, especially since Leopold Aschenbrenner, one of the Open AI researchers, posted an enigmatic post a few days ago on X… suggesting that this year of relative calm at Open AI would soon be over.
“One year since the release of GPT-4. I hope you all took advantage of some time to relax,” he commented, before adding that “these twelve months will have been the slowest in the progression of AI for a long time time to come.” His tweet suggests that this 12-month period without major news for ChatGPT is unlikely to happen again. In other words, announcements could punctuate the year 2024.
ChatGPT-5 on the horizon?
For context, Leopold Aschenbrenner is assigned to Open AI's Superalignment division. The latter is responsible for guiding extremely intelligent AI systems (sometimes nicknamed superintelligence) so that they act in accordance with human values and very specific objectives. His role at Open AI certainly puts him in a good position to know when and how the company will release new AI models. His tweet therefore deserves to be read carefully, even if it does not give us any concrete information as it stands.
We can nevertheless imagine that Leopold Aschenbrenner is trying to warn us, without saying too much, that ChatGPT-5 is about to materialize. However, it is difficult to be certain at this stage. We simply know that the next version of Open AI's conversational AI has been in the works for many months. There is therefore nothing crazy about believing in a deployment this year.
This is all the more intriguing given that Microsoft recently added, by mistake, a mention of a certain “ChatGPT-5 Turbo” on its Copilot Pro subscription page, and that Sam Altman himself suggested at Lex Friedman ( between two “honestly, I don’t know”), that GPT-5 would surely arrive during this year 2024. Looking back, the CEO of Open AI even laughs, in this interview given on March 18, about performance of ChatGPT-4, stating that this version “kinda sucks”.
At Microsoft, Mikhail Parakhin for his part indicated that GPT-5 would be powerful and intelligent enough to decipher a 3,000-year-old language that has never been deciphered. Enough to whet many appetites.