When I talk to artificial intelligence (AI) researchers, many say that the end goal is artificial general intelligence or strong AI. Of course, there are many researchers who are skeptical of the AI results that are being made now, saying that they still have a long way to go. This is also true, but many technologies that seemed impossible in history have become a reality. My dream is to contribute to creating the first general-purpose AI.
General-purpose AI is also called strong AI, but strong AI refers to a system with general purpose intellectual ability at or above human level that can think and solve certain problems. Weak AI, on the other hand, is AI focused on solving a specific problem. Face recognition AI, AI driving, AI that understands sentences, etc. that we are currently encountering are almost among AI. In fact, there are already many areas where AI has already surpassed humans in certain areas. And most researchers predict that the percentage will increase in the future.
Then, why do so many scientists say that their dream and the ultimate goal of their research is general purpose AI? When I talk with outstanding researchers, they all yearn for impactful research. In other words, we want to produce research results that have a profound impact on our society. The impact of general-purpose AI is enough to nurture researchers' willingness to research and is so great that it is beyond everyone's imagination. A robot that thinks, talks, and judges like a human. What happens in the movie will happen in real life.
The impact of general-purpose AI can be explained by the idea of an ‘intelligence explosion’. In 1965, at the height of early AI discussions, British mathematician and statistician Irving John Good predicted an intelligence explosion in the future.
"Let's define an ultra intelligent machine as a machine that can far surpass all human intellectual activities," he said. can do. Then there will undoubtedly be an intelligence explosion, and human intelligence will be far behind. Therefore, the first superintelligent machine is the last invention that humans need to make.”
Numerous science fiction films also deal with the bleak future of mankind caused by this intelligence explosion. The development of AI technology is fearful because it is the future that has already been portrayed in so many movies such as ‘The Matrix’, ‘Terminator’, ‘Ghost in the Shell’ and ‘Transcendence’.
In fact, building a machine that surpasses the human brain still seems like a reckless challenge. However, if you think about it, many of the technologies that appeared in science fiction movies just 30 to 40 years ago have been realized. Mobile phones, tablets and simultaneous interpretation machines, for example, were dream technologies from the Star Trek movie. Therefore, it would be a wise approach to imagine what could happen before and after the technology is realized and prepare for it, rather than dismissing the technology in the film as impossible today.
A key characteristic of general purpose AI is, of course, its versatility. Although there are differences from person to person, humans can perform various tasks through simple education. You can lift heavy luggage, pay for change at the counter, remember the faces of your friends and call them by name.
Versatility is currently a major focus of major AI research. Google DeepMind's DQN (Deep Q Network) algorithm proved the potential of general-purpose AI in that one algorithm can learn many Atari games, and decorated the cover of Nature, a world-renowned academic journal. The research was evaluated as paving the way for general-purpose AI and caused a huge wave of waves.
Recently, GPT-3 of OpenAI, a natural language processing technology, is drawing attention from the AI industry. GPT-3 shows some general abilities in its field, such as writing novels, articles, and even programming. As such, today's AI researchers are taking one step at a time towards universal AI.
I remember jokingly saying that sometimes I wish I had a robot to do my job. As all technologies have double edges, if general purpose AI is invented, it will certainly change our lives innovatively, but AI that has gone beyond the stage of intelligence explosion may no longer just listen to human words and perform them. Just like ants do not understand humans, we cannot understand AI, and at that moment, AI may become an existence that overwhelmingly surpasses human intelligence.
As most will admit, there is still a long way to go. Some say that the first thing to consider is what ‘intelligence’ is that distinguishes humans from animals. However, by the time we reach some conclusions about this, AI may already outperform humans in many areas. As we experience, technology is fast. And the pace of progress is accelerating, and the flow of commercialization cannot be stopped either. How to prepare and prepare the social and institutional foundation to respond flexibly to these changes and to accept them developmentally will be an important task.
