Tim Berners-Lee
Sir Tim Berners-Lee, one of the inventors of the WWW and the hypertext protocol, a professor at the Massachusetts Institute of Technology and the University of Southampton, is both an advisor Government of Great Britain. Therefore, he must soberly look at the world and assess what is happening around and what it can lead to.
Arguing about Artificial Intelligence, many experts think about the emergence of a strong AI and technological singularity, but if we reason with reason, what is the use People from a strong AI, why invest billions of dollars in its development? The benefits of having a strong AI are vague, and the risks are too great.
It is much more logical to assume that first weak versions of AI will appear, which are better than people will perform narrowly specialized tasks. Due to these versions of AI, people will try to gain a competitive advantage over other people for the sake of the main goal that motivates humanity – for the sake of money. *
Some signs of the coming future are already visible today. The most talented programmers and brilliant mathematicians go to work where? That’s right, in investment funds. There they create innovative mathematical models, new trading strategies, exotic ingenious derivatives on non-ordinary formulas – and they look for that alpha, that is, profit above the market. Who has higher intelligence – he gets alpha. The middle peasants get beta, and losers suffer a loss. The financial market, free from all material goods, subcards intrigues – is a struggle of intellects in pure form, according to clearly established rules. Fight for money. This is the only area of human activity in the world where a brilliant mathematician can earn steadily hundreds of thousands of dollars a month, especially if he receives a percentage of the alpha that he brought to his fund.
A reasonable step will be to introduce computer decision-making systems – the most advanced Version of AI – primarily in the financial world. About this, Sir Tim Berners Lee spoke in his speech on April 10, 2017 at the Global Summit on Financial Innovation.
Systems of weak AI from IBM and other companies are already being implemented in banks and investment funds. What will happen next? Tim Berners-Lee believes that the turning point will be when the AI starts making important decisions. For example, assess the benefits of investment and decide which company is more profitable to buy. Assess the credit history of people – and decide who to issue a mortgage loan, and who can not. Modern computer systems are already close to this functionality.
Tim Berners-Lee says that the next logical stage in the development of financial AI after making decisions about buying the most profitable companies will be the management of companies, the creation of holdings and the generation of child copies of its own program for management These new companies.
This will be a natural process, because in a competitive market, AI-controlled companies will have an advantage over companies run by people. In the end, only AI-controlled companies will remain on the market, which will compete with each other. It will be hard for people to understand and make sure that fair competition is observed: “And how to explain the concept of honesty to the computer, in any case?” – asks Tim Berners-Lee. This raises the question of respecting ethics and morality-rules of behavior that people themselves do not fully understand (and disagree with each other), how can this be explained to the computer?
Again, there are risks that a limited version AI will start to develop in the wrong direction. Created initially for a narrow problem, in the end, it can calculate that the entire surrounding reality is described by the conditions of this task. This opinion of Berners-Lee reminds the fears expressed by Ilon Mask in an interview with Vanity Fair:
“Suppose you created a self-improving AI to collect strawberries and it gets better and better in collecting strawberries, and collects more and more , And it improves, and all he wants is to collect strawberries. So in the end the whole world will turn into strawberry fields. Strawberry fields forever [Strawberry fields forever].
In general, there is something to think about. If the program can self-learn, a weak AI with a narrow specialization can be even more dangerous than a full-fledged AI with normal self-awareness. Perhaps, moral principles are an inherent quality of any intelligent being with intelligence and self-awareness.
* In Western civilization, the availability of legally earned money means the existence of power, in other societies it is not always the same concepts.
As a small advertisement: The user ZiingRR informed us that in the end of May, on the basis of the HSE in the IT-lecture hall, the first in Russia relatively large lecture on the AI safety theme will take place (the language of the speech is English). It is also planned to conduct a similar lecture at the Arche. ZiingRR promises that a video will be published from both lectures.
