“Exhausted” artificial intelligence. During an interview broadcast on X on Wednesday, January 8, Elon Musk talked about generative artificial intelligence, which needs data to continue learning. For the billionaire, however, these AIs have already largely analyzed the available data, to the point that almost nothing would be left available.
“We have exhausted the accumulated sum of human knowledge in training AI,” he explains. A depletion occurred during the year 2024.
The risk of a closed vacuum
For Elon Musk, we will now have to rely on the data generated by these AI models, such as Grok or ChatGPT: “The only way to complement real-world data is to use synthetic data created by AI.”
The goal is not to create a vacuum, but to make the model qualify itself and thus move to a self-learning process.
The risk is that AI consumes erroneous or biased synthetic data and then produces an equally biased result. But the main companies in the sector did not wait for Elon Musk’s statements to start working on synthetic data. According to Gartner, 60% of the data used has already been generated by AI.
In fact, it’s no surprise to see that Microsoft has released Phi-4, an open source model based entirely on synthetic data. For some experts, this development, although it involves risks, is much less expensive, particularly when it comes to establishing alliances with media to access paid articles.
Source: BFM TV
