According to the CEO of Nvidia Jensen Huang, the company's GPUs have increased AI computing performance by at least a million times over the past 10 years, and they expect to do the same in the next 10 years. This announcement, even net of the emphasis, makes it clear how much to expect (and in how much time) from AI systems, and what impact they will have on our lives in a short time: from health to mobility, passing through industrial production.
One million and one (another) million
Without us, nothing. This is the summary of Huang's words, if you want to know mine. Nvidia claims to have made it possible to build advanced AI models like ChatGPT, which require huge amounts of processing. What interests me about its release, however, is relative to the future: to the fact that between processors, systems and algorithms Nvidia will allow AI to make another huge leap, one (another) million times* more, and all by 2033.
What are the expected changes due to the new generations of artificial intelligence? For us who are still disposing of the wonder of hearing the "first cries" (already incredible), it is not easy to hypothesize. A hand, however, is given by Huang himself by intervening in the presentation of your company's financial results. Here is his prediction.
AI factories
Within the next 10 years, this massive further improvement in systems could lead to the creation of what it calls AI "factories." What are? In summary, they are factories that do not produce physical goods, but "soft" goods such as artificial intelligence. A bit like the software houses arose at the end of the 70s: we will see small, large, mega-large ones.
And what will they sell us? I don't know exactly what to call them. Maybe we'll find a name for them that doesn't make us repeat "AI" or "artificial intelligence" all the time. Virtual Staff? Synthetic assistants? Maybe. For example, a manufacturing company could use these "soft" goods to optimize production, reduce costs, improve product quality and prevent maintenance problems. Another company might be using AI to improve the user experience of their products and services, tailoring the results to individual needs. Here we are no longer dealing with robots or algorithms that perform a precise task, but that do or plan the work of entire departments, organize entire company sectors.
And one day they take over the… no, I'm kidding. A bit of a wink though, eh?
* It's not entirely clear to me, after hearing Huang's statements, how the claimed increases in performance are measured. But it is clear that he is optimistic that AI will accelerate exponentially in the coming years.
