In 2017, artificial intelligence attracted $ 12 billion in investment from venture capitalists. We're just starting to discover useful AI applications. Amazon recently introduced a retail, if I may say so, store in which cashiers and checkout counters have been replaced by computer vision, sensors and deep learning. And without investment, press coverage and radical innovations, "artificial intelligence" has long been on everyone's lips. But does it exist at all?
Speaking at the World Economic Forum, Dr. Kai-fu Lee, Taiwanese venture capitalist and founding president of Google China, remarked, “I think every entrepreneur is trying to portray his company as an AI company, and every venture capitalist would gladly say, that he's investing in AI. " But some AI bubbles may burst before the end of 2018, namely "startups that presented an impossible story and tricked venture capitalists because they don't understand."
However, Dr. Lee is adamant that AI will continue to advance and take jobs away from humans. What's the difference between AI, with all its pros and cons, and bloated stories?
If you analyze several stories of supposedly AI, you quickly discover significant differences in how humans define artificial intelligence: the line is blurring between simulations of intelligence and machine learning applications.
AI experts are struggling to find consensus, but this question raises more questions. For example, when is it important to be precise with the original definition of a term and when is it necessary to stop refining the definition? Unclear. The hype often gets in the way of clarification. Plus, the hype is backed up by $ 12 billion in money invested.
This conversation is also necessary in part because world leaders have begun to publicly discuss the dangers that AI poses. Facebook CEO Mark Zuckerberg suggested that the skeptics who are trying to "come up with a scenario for the end of the world" are very irresponsible and gloomy. But the creator of OpenAI and the owner of many other companies, Elon Musk, countered Zuckerberg's statement, saying that he did not fully understand the essence of the matter. Musk said the same thing about Harvard professor Stephen Pinker in February. He wrote that Pinker does not understand the difference between functional, narrow AI and general purpose AI.
Given the fears surrounding artificial intelligence, it is important for the general public to clearly understand the differences between different levels of AI so that everyone can realistically assess the potential benefits and threats.
Promotional video:
Smarter than a man?
Eric Cambria, an expert in natural language processing, believes that “nobody is doing artificial intelligence today, but everyone says they are doing exactly that because it sounds cool. Several years ago the same story happened with big data."
Cambria notes that AI as a term originally referred to the emulation of human intelligence. “But today there is nothing even close as smart as the dumbest person on Earth. Therefore, strictly speaking, no one is engaged in artificial intelligence, does not do it, if only because we do not know how the human brain works,”he says.
The term “AI” is often used to refer to powerful data classification tools. These tools are impressive, but they operate on a completely different spectrum than human cognition. Also, according to Cambria, people are claiming that neural networks are part of the new wave of AI. This is strange because the technology has been around for fifty years.
First and foremost, AI gives access to great computing power. All of these advances are welcome, but it would be wrong to believe that machines have learned to emulate the complexities of our cognitive processes.
“Companies are just using tricks to create behavior that resembles intelligence, but it's not real intelligence, it's just a reflection of it. Such expert systems can be quite good in a certain area, but very bad in others,”he says.
This mimicry of intelligence has inspired the public imagination. Systems operating in specific domains have breathed life into a wide range of beginnings. But did it help rid the world of confusion? Rather the opposite.
Auxiliary, augmented or stand-alone
When it comes to scientific integrity, the question of precise definitions is not left out. In 1974 at the California Institute of Technology, Richard Feynman said: "The first principle is that you must not deceive yourself - you are the easiest to deceive." And further: "You don't need to deceive a layman when you speak on behalf of a scientist." He assumed that scientists should think about the fact that they too can be wrong, ripen to the root. "If you imagine yourself to be a scientist, you have to explain to a layman what you are doing - and if he decides not to support you in the given circumstances, then that is his decision."
In the case of AI, this may mean that professional scientists have a duty to make it clear that they are developing extremely powerful, controversial, profitable, and even dangerous tools that do not represent intelligence in any familiar or all-encompassing sense.
The term "artificial intelligence" may be misleading, but attempts to clarify it are already under way. A recent PwC report outlined the distinction between "assistive intelligence", "augmented intelligence" and "autonomous intelligence." Assisted intelligence is represented by GPS navigation programs that run in cars. Augmented intelligence "enables people and organizations to do things that they would not otherwise be able to do." Autonomous intelligence “allows cars to act on their own,” for example, in the case of self-driving cars.
Roman Yampolsky, an intelligence security researcher, believes that “intelligence (artificial or natural) continues to evolve, and with it the potential problems of this technology. We are accustomed to thinking that AI will one day master the full spectrum of human capabilities and become general artificial intelligence. Then he becomes superintelligent. But today we mostly use focused AI. The problem is not in terminology, but in the complexity of such systems, even at the current level."
Should people be afraid of AI? “As opportunities continue to emerge, all kinds of challenges will emerge,” Yampolsky said. There will be more and more incidents involving AI.
According to Brian Decker, founder of Encom Lab, machine learning algorithms currently work solely to meet the needs of programmers. “The marketer will say that the photodiode-controlled porch light is artificially intelligent because it“knows when it's dark outside,”while a good manufacturing engineer will point out that not a single bit in the history of computing has ever been changed, if only so was not conceived according to the logic of previous programming."
Ilya Khel
