It is not so long ago that Artificial Intelligence has come into our lives in a “normal” way, that is, through apps, platforms, web pages or integrated into devices that we use every day, but this technology has been developing for many years before. get the AI models we have today.
We have the mistaken belief that AI is intelligent and capable like humans, but this is not exactly the case, and while this technology can replace us in many tasks, for now They are not capable of doing or creating without human intervention, That is, for the AI to create a text, we have to give it the instructions to do so.
But not only that, but to be able to create a text, AI has had to learn what a text is and learn with real examples to then be able to create its own. This applies to all the things AI can do, so everyone These systems have been trained with enormous amounts of data that serve as an example so that they learn to replicate what humans ask of them.
But, in recent weeks, experts have begun to warn that We have reached the training data limit for AI, That is, there is no longer data to train these models and this represents a serious problem for which we have to find a solution.
Among these voices that have raised the alarm, we find that of Elon Musk who claims that we actually reached this limit a few months ago, in 2024. Musk collects the words of Ilya Sutskever, former chief scientist of OpenAI, in 2022 where he spoke of “data spike”which refers to this limit of information that exists to train the AI.
Temporarily, Musk has proposed a solution that is not new and that in fact several of the leading companies in AI such as Microsoft, Meta, OpenAI and Anthropic have already had to take. It is about the “synthetic data” that refer to sets of information generated by machines that They mimic the characteristics and patterns of real-world data, without actually being that data.
In fact, It is believed that last year 60% of the data used was synthetic, since they are also a much more economical way. However, these are not as pretty as they seem, and several research suggests that using synthetic data They could reduce the creativity of these AI models, increase the biases in results and in general there could be the collapse of all these AI models.
#Elon #Musk #data #peak #Artificial #Intelligence #reached #limit
