Five experts answered the question 'Will the day come when AI acquires the same level of intelligence as humans?'

With the rapid development of interactive AI 'ChatGPT' and image generation AI 'Stable Diffusion', there is a report that 'these generative AIs

will increase global GDP by 7% and affect 300 million jobs. ' Or , Bill Gates, the founder of Microsoft, predicted that ``AI will teach children to read and write within 18 months,'' and various predictions about the impact of AI on the world were raised. I'm here. The Conversation, a non-profit news media, asked five AI and technology experts the question, 'Will AI ever reach human-level intelligence?'

Will AI ever reach human-level intelligence? We asked five experts

The success of GPT-4 and ChatGPT developed by OpenAI has accelerated the research and development of many other companies and AI development organizations, and the AI field is growing rapidly. Due to its rapid development, ``AI poses a serious risk to society and humanity,'' said Elon Musk, one of the founders of OpenAI, and Steve Wozniak, co-founder of Apple. , has signed an open letter calling for an “immediate cessation of development of AI beyond GPT-4 for at least six months.”

More than 1,300 people, including Earon Mask and Steve Wozniak, signed an open letter asking all engineers to stop the immediate development of AI beyond GPT-4 for 6 months due to the fear of ``loss of control'' - GIGAZINE

With the development of AI, more than 40% of experts in the cybersecurity field answered that 'AI will take away their jobs by 2030.' While boosting gross domestic product (GDP) by 7%, a ``major labor market disruption'' will occur where the jobs of 300 million full-time workers in major economies will be affected by automation. I'm here. While there are concerns about ``jobs being taken away by AI,'' AI is currently active in areas such as programming and writing, generating some illustrations and photographs, and playing tactical games such as chess and Go. etc. is limited.

The main and ultimate goal of AI research is the ability of AI to understand, learn, and perform all kinds of intellectual tasks that humans can perform. )” is an idea. AGI has not yet been realized, but The Conversation asked five experts and researchers at a university in Australia, 'Do you think that AI will reach human intelligence completely, that is, AGI will be born?' When interviewed, all five answered 'yes'.

Below are detailed comments from each expert.

・Paul Formosa (AI ethics, philosophy)
Formosa, a professor of philosophy at

Macquarie University , said, ``If enough time has passed, it seems likely that AI will grow into AGI, that is, acquire ``human-level intelligence.'' . Some people think that AI is not conscious, does not understand what it is doing, and cannot be called 'intelligent.' Existence can do intelligent things such as learning, reasoning, writing and using tools.The rise of AI in recent years has suggested that it is possible to have intelligence without consciousness. I'm here.

・Christina Maher (computational neuroscience, biomedical engineering)
Dr. Maher, who studies computational neuroscience, which looks at the brain as a computer and examines its functions at the University of Sydney , and biomedical engineering, which applies engineering knowledge to medicine, believes that AI will achieve human-level intelligence. , immediately concludes that it is not possible. The inference, problem-solving, and decision-making abilities of human-level intelligence require adaptability, sociality, and the cognitive ability to learn from experience, and AI is beginning to equip these elements. Therefore, ``AI will reach AGI if it can understand the inherent characteristics of humans, such as critical reasoning and emotional nuances,'' Maher said. It's about learning 'the emotions that humans are born with.' So the big question is not whether AI will achieve human-level intelligence, but when and how.' I'm here.

・Dana Rezazadegan (AI, data science)
Rezazadegan, a lecturer in AI and data science at Swinburne University of Technology , said AI rivals human intelligence in many ways, but how much smarter it gets will depend heavily on advances in quantum computers. increase. The human sense of creativity, emotional intelligence, and intuition cannot be analyzed with today's datasets and limited computing power. But when quantum computers power AI, they can feed AI models with vast amounts of data about how humans interact with the world, and make sense of it with fast and accurate analytics. ``AI algorithms running on stable quantum computers are likely to reach something close to general human intelligence,'' Rezazadegan concludes.

・Marcel Chart (Machine Learning and AI Alignment)
Mr. Sharuto, who gives a lecture on machine learning and ' AI alignment ' that controls AI in the desired direction for humans at the University of Sydney, said, 'I think there is a high possibility that AGI will be realized, but there are strong uncertainties at that time. I think,' he replied. For AI to match human intelligence or surpass the computational power of the human brain, the challenge for us is to understand what 'intelligence' is and to build the domains that characterize AGI, according to Schart. to gain knowledge for Metaculus , one of the well-known platforms for making predictions about AI, puts the median forecast period to reach AGI at 2032, but Scharto said the road to AGI now has unpredictable breakthroughs. The forecast is optimistic because of the need for and innovation, and they support the 2022 expert survey , which believes there is a 50% chance of AGI by 2059, being more plausible.

・Saydari Mijariri (AI, swarm intelligence research)
Mijarilli, who specializes in swarm intelligence, the field of AI about the collective behavior of AI and self-organized systems at Torrens University , said: , content creation and other tasks have proven to be feasible for machines due to advances in technology.AI algorithms are advancing rapidly, and if we follow the same trajectory, we can achieve previously unimaginable levels of AI demonstrates intelligence and automation.” After showing the infinite possibilities of AI development, he added, “The birth of more generalized AI is not a possibility, but a certainty of the future. ” says. Mr. Mijarilli, like Mr. Maher, admits that work based on intuition and empathy is an area where AI is not good, but says that it can be achieved by breakthroughs in algorithms. In addition, Mijariri points out that when AI begins to acquire cognitive ability, AI will improve AI itself, ``automation of intelligence'' will occur, and the world will change greatly from there.

'The development of AGI is a big technical challenge, but it has to be approached very carefully ethically and socially if we are to make significant progress towards AGI,' Mijarilli said at the end of his answer. said.

in Science, Posted by log1e_dh