ChatGPT works as a marketing tool by generating inequalities in the relationship between employer and workforce, says neuroscientist Miguel Nicolelis. For him, intelligence is the result of millions of years of evolution, which cannot be computed in binary code.
Nicolelis has been working with neural networks for 30 years, the mechanism behind current machine learning algorithms. Reference in interfaces between brain and machine, he worked in the development of neuroprostheses capable of restoring body movements. During the opening of the 2014 World Cup, in São Paulo, a wheelchair user kicked the ball into the goal with the help of equipment developed by him.
Nicolelis says to Sheet that it is absurd to say that language models like ChatGPT are ten times more intelligent than a human being because they write quickly or communicate in several languages, as did Geoffrey Hinton, computer scientist who invented neural networks and was partner and advisor to Google for over a decade. “The tortoise is extremely intelligent, it’s just slow.”
Mr. criticized writer Yuval Harari. Why?
He mixes things from other areas without having deep knowledge. In Sapiens, he mixes up references and interprets our results in a way that has absolutely nothing to do with what we did. It’s a job I’ve spent 30 years of my life doing. When he says that in the future we are going to put this thing called brain-brain interface, which was something experimental that I did with rats, I did it with monkeys and we did it with human beings, for rehabilitation. But it’s not that I’m going to exchange my feelings with other people. It’s an exchange of motor commands, things suited to reducing digital logic. He interpreted it as if I was reading someone’s mind, which will never happen. He says: ‘we are going to live to be 200 years old’, ‘we are going to put an end to aging’. All of this is fantasy.
What about what Harari says about artificial intelligence?
He lives from lacration to lacration. He wrote that artificial intelligence hijacked the system; she didn’t kidnap anything. The human species is hijacking its own evolution.
Behind artificial intelligence, there are armies of people taking notes.
And it has armies of evangelists. I never liked that word, because it denotes that the vast majority of human movements have become religions. Everything looks like religion. From a scientific point of view, I’ve been saying this for years, and now Noam Chomsky uses the same phrase, artificial intelligence is neither intelligent nor artificial. It is not artificial because it is created by us, it is natural. And it’s not smart because intelligence is an emergent property of organisms interacting with the environment and with other organisms. It is a product of the Darwinian process of natural selection. The algorithm can walk around and do things, but they are not smart by definition. If he were alive, Charles Darwin would have a heart attack from this
Is calling machine learning better?
Machine learning, deep learning, machine learning, are big names that use words that we are used colloquially to use, related to the human brain or any animal, to define things that we do with binary logic. Human intelligence is not binary. Therefore, it is a misnomer.
The creator of neural networks Geoffrey Hinton says that he tries to simulate the structure of the neuron, to think about these algorithms.
He comments a lot of nonsense too. He said that artificial intelligence is already ten times superior to human intelligence, which is absurd. We have these marketers in these areas of technology who claim things that ring true. But they don’t have the proof.
He works with results. He talks about the speed with which he delivers responses, multiple languages.
The tortoise is extremely intelligent. She is slow. But what we’re talking about is trying to use the language of the market to define what life does. The market wants fast, efficient things, with infinite profit and zero expense. Intelligence has no such commitment. The organism’s intelligence is committed to making it survive as long as possible in an ever-changing environment. Just because a computer plays chess faster and beats a world champion doesn’t mean it’s smart. It’s just more efficient, because chess is a game with predetermined rules. This computer cannot survive in the Palmeiras stadium on a game night, it does not understand the reasons for a fight, because it does not have the ability to generalize its intelligence.
The researcher at the Open Philanthropy Institute, Ajeya Cotra, estimated that, in the current model of society, the human mind runs the risk of being obsolete by 2037 in terms of production for the job market. That makes sense?
It depends on what you call production and what you call obsolescence. There is a limit of digital logic. I just finished reading a book by one of the best intellectuals in the field of AI, Michael Wildridge, from the University of Oxford. It came out in 2021. In the book he says: we know that there is a limit determined by non-computable phenomena, in which there is no algorithm, there is no mathematical formula that can be solved with a program. Except that he puts two paragraphs on the most important thing in the book, and comments that researchers don’t pay much attention to it because they have too much to do. But the human mind is full of non-computable phenomena: intelligence, intuition, creativity, aesthetic sense, definitions of beauty, creativity, all of this is non-computable. What is the formula for beauty?
A young woman posted on Twitter that her uncle was accused of plagiarism because a professor took a piece of his work and asked if it was done by ChatGPT for ChatGPT. The platform is not designed to recognize whether a text was made by artificial intelligence and always responds that it is the author of any text.
In a way, ChatGPT is a great plagiarist, because it takes material made by a bunch of people, mixes it up and generates something that it calls a new product, but in reality it is largely influenced by the intellectual product of thousands and thousands of people. human beings. For the current, modern capitalist system, artificial intelligence is the great marketing tool, because it generates total inequality in the relationship with the workforce.
A boss can say: I have an artificial intelligence application, if the worker does not accept the salary I am willing to pay, which is 10% of what he earns today, I will fire him and use the application. There is a whole ideology of replacing human work, which cannot be done 100%, there is no way.
Can you say that a more utilitarian thought is gaining ground in society?
That’s the problem, this has nothing to do with the machine. What is being done is forcing human biology to follow market rules. Market rules are not divine, they are abstractions created by the human mind. What did they produce in human history? A resounding inequality of income distribution. We have people spending money to dive to see the Titanic explode in the middle of the ocean. If someone walks here from Avenida Paulista, as I did, they see tens of thousands of people dying of hunger in the streets. All of this is being ignored because these systems are convenient. They increase our productivity and our reach as a human being.
Mr. is more in line with the thinking that today these language models are more like statistical parrots?
Totally. Deep learning is nothing more than neural networks with multiple layers, more layers, more neurons and more connections between these layers. The brain does this too. However, it is impossible to simulate the biological mechanisms that the brain uses to make this decision.
The brain uses much less energy than AI supercomputers to deliver the same processing.
It’s a million-year optimization process. No wonder we came down from the trees, it took 4 million years for us to walk. It’s a much more elaborate thing: 20% of the energy your body produces goes here. [aponta para a cabeça]. The brain’s energy is enough to light a light bulb, more or less. It’s an extremely optimized section that has undergone brutal modifications since life appeared on Earth. And it is not computable. Alan Turing himself knew this, after proposing his thesis, he said: there are certain problems that my theoretical machine, which has already become a Turing machine and generated computers, will not be able to solve. And when I have this impasse, there’s only one solution. I have to call an oracle to make a decision. The oracle is a human being.
But within this logic of competition between machine and human being, do you agree with the risks for the species suggested by researchers and people in the technology industry?
The risks are tremendous. These tools have to be used under human supervision. In programming an AI system, the person asks for something, but may not consider the means to achieve the goal to be unwanted. This is what happens with the HAL computer from the film “2001 – A Space Odyssey”, by Stanley Kubrick. His mission was to arrive with the crew at a location. They just forgot to say that HAL couldn’t kill the crew. They forgot the scenarios in which the mission would be complete, but there would be no people left to see it. When someone delegates something to do a mission on their behalf, it’s not going to be able to give that thing all the restrictions we have right away because of evolution.
These mechanisms can be useful, in terms of research, as are your studies in neuroscience?
I’ve been using neural networks to interpret real neural activity patterns since the 1990s. Not the same networks as today, but simpler. It is a statistical pattern recognition method.
I don’t agree with turning a statistical tool into a new God and building a whole religion under it, as is happening. I call the church of technology.
Miguel Nicolelis, 62
He headed the Center for Neuroengineering at Duke University, before retiring as professor emeritus in 2021. A physician, he is a reference in the study of the interface between brain and machine and coordinated the scientific committee of the Northeast Consortium. He was the first Brazilian to publish an article on the cover of the scientific journal Science.