OpenAI est convaincu de savoir comment construire l'intelligence artificielle générale, alors que nous attendons une Siri plus intelligente.

We could always be primarily waiting for a smarter version of Siri, but the CEO of OpenAI, Sam Altman, thinks that the company already knows how to create the holy grail of artificial intelligence: general artificial intelligence (AGI).

AGI is the term used to designate an AI system capable of matching or exceeding human cognitive abilities in a wide range of domains – in other words, AI that is at least as intelligent as we are…

Altman made this statement in a blog post on his personal blog.

We are now convinced that we know how to build a general AI (AGI) as we always understood it. We believe that by 2025, we could see the first AI agents "join the workforce" and physically change business production. We continue to believe that iteratively putting large tools in people's hands leads to major results widely distributed.

We are starting to look beyond this, towards true superior intelligence. We love our current products, but we are here for future glory. With superior intelligence, we can do anything else. Superior intelligent tools could accelerate scientific discoveries and innovations on a massive scale beyond what we would be able to achieve with our own means, and that would in turn lead to an increased abundance and prosperity.

Knowing how does not mean being capable of doing it, of course, and Altman does not specify when this will be realized – he simply says that he expects important advances non-specified in the "next years".

This seems to be science fiction for now, and a bit crazy to talk about. Of course—we were already there before and we are happy to do it again. We are quite confident that within the next years, everyone will see what we see, and that the need to act with much prudence, while maximizing wide benefits and autonomy, is so important.

The post looks both back and forward, with Altman describing his surprise at the beginning.

About a year ago, on the day in question, what had gone wrong that day was that I was fired without notice during a video call, and just after hanging up, the company's blog published an article about it. I was in a hotel room in Las Vegas. This made me feel, with an intensity almost impossible to explain, as if a bad dream had ended.

Being fired publicly without notice triggered a series of terrible hours and days. The "warfare fiasco" was the most bizarre part. We had no satisfactory answer on what had happened or why.

The whole of this event was, in my opinion, a major failure of governance by well-intentioned people, including myself. Looking back, I am certainly regretful that I could have done things differently and I would like to believe that I am a better leader more reflective today than a year ago.

While Altman could predict a real AI, we are still (primarily) waiting for a smarter Siri. Although some of Apple's artificial intelligence features have already added capabilities to the intelligent assistant, we should not expect a fully conversational Siri before at least 2026.

Photo by Aidin Geranrekab on Unsplas