According to Dr.Roey Tsezana, "We make real breakthroughs every year, and optimize the capabilities of artificial intelligence significantly, with the most important breakthrough being in the field of machine learning and first and foremost in Deep Learning
There is nothing to fear that artificial intelligence will take over the world by itself, it is more dangerous that people will misuse it, says Dr.Roey Tsezana, a researcher at the Blavatnik Center for Interdisciplinary Cyber Research at Tel Aviv University, and the author of the book "The Rulers of the Future". Dr. Cezana lectured on this issue at the Bloomfield Museum of Science in Jerusalem, at the night of the scientists last week.
According to Dr. Cezana, the most important field is DEEP LEARNING, in-depth learning that is achieved through the creation of artificial neural networks that roughly resemble the way biological nervous systems work. In fact, we learn from nature to create virtual structures that mimic the way biological nervous systems learn, but according to him, we do this at certain points in a more successful and focused way than evolution did. We are not trying to produce a brain in a computer, but only to imitate certain ways in which the human and animal brain achieve their great achievements in areas such as image processing, inference, learning from a large number of examples, and so on.
Thanks to these developments, computers can now do things that they were unable to do in the past, for example identify objects in images with a good level of certainty, as well as perform face recognition, voice recognition, understand human speech and provide an appropriate response, deal with games and more. These are areas that artificial intelligence has not dealt with in the past. And in general, in situations that the AI has not been exposed to so far, but that it can now learn from other situations what it should do.
"We see that the artificial intelligence even succeeds in programming itself in certain places, when we allow it," said Dr. Cezana and wondered "what is the meaning of learning if not self-learning?!".
Cezana claimed in his speech that artificial intelligence can greatly advance the field of robotics. He told the audience about a very interesting study during which we gave the artificial intelligence based on deep learning different types of robot models and required it to understand how the models should move in the simulation in order not to fall. It would take a lot of time and a lot of work for human engineers to find the solution to these problems, but the artificial intelligence could run a lot of simulations, find the unique parameters and assign optimal movement ability to each type of robot, according to its structure.
However, Dr. Cezana also explains that when we want to control robots and allow them to move through the streets, we must take into account additional things, and this time limitations - a robot with four legs that loses a leg or a stick gets stuck in its wheels, or even if a screw just fell from one of its legs, its entire way of moving, that you have learned up until now, is completely useless. This is in contrast to a person who, if his leg is injured, can hold on to a stick or limp more slowly.
"We also see that artificial intelligence can perform tasks that humans are not capable of and develop strategies that humans did not even know they needed to develop in the past. A good example of how artificial intelligence can develop higher strategies and intuition than humans can is the game of GO, which artificial intelligence has specialized in over the past two years. The way the AI mastered the game, which is much more complicated than chess, was that it played a million games against itself and from each such game it learned, and each time it got better for the next game. "This is an amount of games that a person cannot play in his entire life, and the artificial intelligence did it in months," said Cezana.
When the AI faced off against Lee Sedol, one of the world's most prominent GO champions, he was confident that he would beat her because he remembered her from six months ago, when she was still in the very early stages of her training. But what he found out very quickly was that in those six months she did a series of training that a person doesn't do in their entire life and that's how she managed to beat him 4-1, while using strategies that humans hadn't thought of before.
"It is said that one of the GO champions who was watching the games just started crying at one point, because he saw how the artificial intelligence was showing humans that everything they think they know about a game that has existed for thousands of years does not compare to what the artificial intelligence can bring to the table," said Cezana.
"This understanding that artificial intelligence can improve our performance, our industries, our private lives and even our governments is beginning to permeate everyone and everyone wants to know how and what it will do, but this question is still open," he said. "Putin's statement that the country that will be the strongest in the field of artificial intelligence will rule the world is exactly the point. We can talk about a future in 50 years, where artificial intelligence acquires its own desires and threatens to eliminate humans. In the distant future. In the here and now what is important is that everyone is in a race to improve artificial intelligence also because it has the ability to improve everyday life."
According to him, the estimates say that by 2030 artificial intelligence will increase the world economy by 16%. out of nowhere Simply because artificial intelligence enters all fields and optimizes our conduct - driving cars, flying drones, treating us medically, psychologically and so on. But Dr. Cezana emphasized that in addition to this, the technology could also be responsible for cyber attacks and cyber defenses. "We see that the world is becoming more connected and those who can carry out the strongest and most effective cyber attacks can paralyze entire countries," he said.
"There is no doubt that the one who controls the most sophisticated artificial intelligence will in fact control the world and gain great power over the other countries that are not as advanced," concluded Dr. Cezana and said, "and that is why, in the coming decades, we do not need to fear that artificial intelligence will go out of human control, but we need to fear from those who control artificial intelligence and ask what they can do with it."
See more on the subject on the science website:
5 תגובות
The author is right in that in the coming decades there is nothing to worry about artificial intelligence taking over the world. On the other hand
In the longer term there is definitely room for concern because if they don't start worrying now it will be too late to worry.
An intelligence that will reach a level above man in all fields will no longer be controllable and from that moment the end of humanity is certain.
For that matter, even if they develop a race of genetically engineered humans who are far superior to other humans, the end of normal humans is also certain.
Lassaf agrees that it is definitely possible that the descriptions of AI will take over tomorrow as they are loved in the media and are ahead of their time, although humanity has already experienced several technological revolutions that caused a huge change in our lifestyles so this is not something that has not happened in the past, the AI systems that exist today are only at the beginning and are called AI systems Uncommon problems
But already in the narrow field in which they are found, they surpass the humans who are the smartest among us in game simulations where they are tested each time in a more complex game which is quite amazing, they are in the initial stages where they begin to enter the fields of work such as medicine, law and more and they will have a very big impact already at these stages , it is clear that if along the timeline these systems continue to improve their impact will be greater with the enormous advantages including the disadvantages that will come, perhaps one of the things a person does is mental simulations of future situations
And since these are systems with increasing complexity, we need time to prepare and try to shape the future that will be better for us that those systems will give the good they can give while reducing the disadvantages,
That's why the time to work on it already starts today because at some point the extensive AI systems will enter
And we will not have the right tools to deal with those very complex problems that will arise.
Regarding the 3D printers, they certainly have a huge potential that is only partially utilized.
, but the reason for the buzz is perhaps breakthroughs of new systems that were thought to be in the 80s but happened now, at the moment there is no barrier to the continued improvement of those systems, but maybe there will be in the future
There may be great difficulty in transitioning from narrow AI systems to broad ones.
The description of the fixed and limited robot is correct but it corresponds to the initial current state and the future knockoff,
Regarding the AI takeover, it might be more human dedication to the power of the AI than forceful takeover,
Like a barter where we will lose our independence in order to receive its enormous benefits, in fact the first buds are already visible today even before the rise of AI, today our knowledge is no longer independent at home we do not have an encyclopedia at home
We don't even have a CD with encyclopedia, everything is in the cloud which we access through the mobile which is our interface, on the one hand we get a huge benefit where the knowledge becomes available
All the time but at a price where it is not actually under our physical control, a similar trend of behavior is also seen in small companies that are bought by corporations where the company/branch step by step loses its independence and parts of it pass to the control of the corporation. Branch is pushing itself to transfer shares to the concern at the cost of loss of independence, it seems that the most problematic thing that AI will have beyond its enormous advantages is the loss of work and the transition of people from a situation where they have value to a situation where they have less real value than people in the feudal period, the value of people may be due to ideology, but The state or what will happen in the future will not need them whether for the army or for work, people may be perceived as unnecessary in the future world
As consumers of resources without any contribution, a place where an elite group that controls through AI and through sophisticated robots everything they need to exist comfortably against a group of people who, at best, live on aid money from the state,
The great ideologies of the 20th century were all based on the importance of man, even if some of them were only seen as cogs in the machine. Propaganda resources had to be invested in order to convince the people to devote themselves to the ideology. In the future, the people will for the most part be redundant. The only reason to perhaps try to convince them is that riots will not start.
But not to recruit them to some idea because they will be completely unnecessary, the whole economy we know will change because if people don't have resources because they don't work then they also don't have the purchasing power on which the entire modern economy is built, it's almost impossible to know what the future world will be like after the elite AI, maybe this is the so-called singular point.
The buzz of artificial intelligence is blown out of proportion. It is possible that artificial intelligence will be good in very specific niches, but in the visible range it is very far from the dream that is sold to everyone. It's quite similar to the buzz that XNUMXD printers had, huge potential but in practice we are very far away.