Even since long before the foundations of the artificial intelligence, humanity has “fantasized” with the idea of a society of robots that take control of the world; clearly at our expense. For this reason, every time AI is in the news somewhere in the world, the comments assuring that “they will soon come for us” do not take long to arrive, also adding that “they will take away our jobs and our freedom”.
However, although they are speculations full of fear of the future, they hide a very real question: When will artificial intelligence surpass us in our day-to-day tasks? The answer to this question is already being studied by experts in the field. In fact, the first great advances in artificial intelligence will come in this same decade.
But first, artificial intelligence has to start understanding, not just doing. As he stated to THE COUNTRY François Chollet, software engineer specializing in AI and machine learning at Google, “Something happens like in the movie terminator it’s pure science fictionNo, it’s not even good science fiction. There is a huge disconnect between what the general public perceives as AI, due to popular culture, movies, and media misconceptions, and what AI actually is,” adding:
They tell us stories about an AI aware of its own existence, with its own goals and feelings. But in practice, this is centuries away from our current capabilities. We have no idea how we could even begin to develop something like that. And it’s not just that, the main thing is that no one is trying
François Chollet, software engineer specializing in AI and machine learning at Google
The limitations of artificial intelligence
Until now, the AI has a big restriction. Despite being great tools for the human being, and being brilliant solutions in their specialized areas, they have very limited capabilities. According to the aforementioned medium, “A child in his earliest childhood is capable of learning to put the triangle in the hole of the toy triangle, he can recognize the sounds of animals and begin to apply what he learned in one scenario in other different ones” .
Nevertheless, machines do not have the ability to relate things and entities without prior training. For this reason, by not having adaptive learning to different environments, an AI is not capable of transferring what it has learned to other activities and tasks for which it has not been trained. Of course, it is a field in which progress is being made, but there is still a long way to go before it can reach human learning.
“There is no neural network in the world capable of identifying objects, images, sounds and playing video games at the same time as people.”
THE COUNTRY
A big change is coming
In its report published in 2017, the web MIT Technology Review He already told us about the changes that are predicted for artificial intelligence in the medium term.
In this document, they tell us that artificial intelligence will be able to outperform humans in simpler activities within this same decade. Katja Grace, a researcher at the Future of Humanity Institute at the University of Oxford, in the United Kingdom, brought together a group of fellow experts in the field to unravel the matter.
The surveys carried out by Grace and her team yielded surprising results. The sample collected by the researchers is based on a number of 1,634 experts who presented papers concerning Neural Information Learning in December 2015; and those who participated in the International Conference on Machine Learning in the middle of the same year. In this way, Grace has managed to collect a good number of samples for her survey results.
Of these 1,634 experts, only 352 responded. The results were quite interesting, with the average claiming that artificial intelligence will outperform humans in several activities in the next 10 years (based on the date of the study, 2017). Among these tasks we have the translation of languages (2024); high school essay writing (2026) and truck driving (2027).
Other more complicated tasks will take longer to arrive. Among them, we have the writing of a best-selling book, which the results estimated for 2049. Likewise, we have the performance of surgeries, which the study reports that we will have to wait until at least 2053.
How much longer to cross the final barrier
The experts already have the answer, or at least they think they do. During the meeting at Oxford University organized by Katja Grace and her classmates, the final verdict has been announced. Just as they already predicted that in the coming years artificial intelligence would outperform human beings in various tasks during the current decade; it seems that it is still missing at least another half century for it to surpass us in all other respects.
In the survey results, the world’s leading researchers in the field of AI confirmed that this change could happen in a few forty-five years. As of this moment, the AI would have already reached a stage of development more advanced than that of any human being on the planet.
Not all that glitters is gold
However, there is a “but”… These same experts were the ones who predicted that AI would not be able to beat a human at the game of Go until the year 2027. However, to everyone’s surprise, this came much sooner. In 2015 the first human being was defeated in Go by an artificial intelligence from Google. For this, expert predictions cannot always be taken to heart.
Likewise, forty years was the date given by scientists to achieve cold fusion for the first time. However, this field that started more than half a century ago is still a lot of sketches and nothing physical.
And what does the above have to do with the evolution of AI? Human beings are inclined to estimate great advances every forty years. This number coincides with the total length of working life of most of the population. For this reason, estimates usually show that the change will take place after the end of the working life of those currently working.
One last vote of confidence
However, the mean prediction for each of the experts yields Approximately 45 years until AI surpasses us all. In this way, in the queue of scientists we will find some estimates more accurate than others. For this, Grace prepared in advance, and analyzed the results of each participant according to her region of origin, experience and age.
The results were quite surprising. While American researchers predict that AI will overtake us in a distant 74 years; those belonging to the Asian continent do not believe that we have to wait that long. According to the results of the latter, in just 30 years the machines will be able to take the big step.
So far, there is no clear answer. But if there is one thing that all participants seem to agree on, it is that the revolution will come before the end of this century.