All it does is unearth statistical correlations in a sea of data, giving it a form that (in the case of artificial intelligences that create text or images) seems similar to the product of human intelligence, but which in reality has nothing to do with it. see with what are the distinctive features of our intelligence.
Deep learning can certainly still progress, especially from a quantitative point of view (more data, more computing power, even larger neural networks). To reach AGI, however, we would need a qualitative leap, of which there is not even a shadow yet. In short, for the moment, artificial general intelligence is nothing more than an idea: a kind of technological holy grail that everyone is pursuing, but no one knows if it will ever be truly achievable or even what path to take to achieve this. science fiction objective.
These are aspects that some of the greatest experts in the field have been insisting on for years: from the Italian philosopher Luciano Floridi (who already dismantled this thesis in 2016 with arguments that remain totally valid today) to the computer scientist Gary Marcus, according to the which deep learning (which is the basis of everything we now call artificial intelligence) cannot lead us to human-like intelligence.
As James Bridle, an artist trained in cognitive computing (and author of the important essay New Dark Age), “ChatGPT is inherently stupid: you’ve read most of the web and know what human language should sound like, but it has no relation to reality. […] He is very good at producing something that sounds like it makes sense, and especially at producing clichés and clichés, but he is unable to meaningfully engage with the world for what it is. “Be wary of anyone who tells you that this is an echo, even an approximation, of real intelligence.”
In summary, we have created a very advanced and amazing technology capable of computer and statistically simulating certain aspects of human intelligence. And now we assume that this same simulation should at some point, inevitably, become a real form of intelligence.
What awaits us
None of this, of course, excludes the possibility that one day (we don’t know how far away) a true form of artificial general intelligence will emerge. Nothing can be ruled out a priori. What is important to note, however, is that at the moment it is not even glimpsed, nor is the necessary technological advance that could lead us to this achievement. As computer scientist Andrew Ng (former head of Google Brain) said, “worrying about artificial general intelligence is like worrying about overpopulation on Mars before you’ve even set foot on it.”
The reason why the story of artificial general intelligence has caught on so much is that there are enormous economic interests behind it. For the computer scientist and Nobel Prize winner Demis Hassabi, founder of DeepMind (whose mission, by statute, is the creation of AGI), there are entrepreneurs like Sam Altman, Elon Musk, Satya Nadella and countless others, for whom artificial general intelligence does not It’s more than a useful narrative device for pursuing your private goals.
#general #publicity #stunt
