A new “generation” has entered the artificial intelligence (AI) lexicon.Meanwhile, NVIDIA CEO Jensen Huang recently claimed Although it has been suggested that “artificial general intelligence” could emerge within the next five years, many experts believe that this new technology will not pose a threat to human jobs or lifestyles in the near future. I am.
Huang, the leader of the world's top company that makes AI chips, answered a question at a Stanford University economic forum about the timeline for achieving Silicon Valley's enduring ambition to develop computers that can think like humans. Ta.
He explained that timelines are heavily influenced by how the goals are characterized. He noted that if being able to pass tests on humans is the benchmark, then artificial general intelligence (AGI) may be here soon.
“If I were to give an AI… every test you can imagine, create a list of those tests and put it in front of the computer science industry, and in five years, all the It will do well on the test: Last week, the company's market capitalization reached $2 trillion, Huang said.
What is AGI?
There is widespread confusion in the public about the difference between AGI and AI. AI follows specific instructions. AGI aims for autonomous intelligence, Duncan Curtissaid the senior vice president of AI products and technology at AI data company Sama in an interview with PYMNTS.
Models like ChatGPT struggle with creating entirely new content and combining disparate information to solve unfamiliar problems, Curtis said. They can sound overconfident, if not completely accurate. AGI, on the other hand, is designed to effortlessly meet new challenges and continuously learn and improve.
“This confusion makes it easy for the public to mistake models like ChatGPT for intelligent AGI,” he added. “A better definition of what both AI and AGI can do is necessary for safer interactions between the public and AI, and will help reduce unrealistic expectations and hype for both AGI and AI. is also helpful.”
Some researchers claim that current AI models show signs of AGI. Last year, Microsoft researchers Published a paper He said GPT-4 is a major step toward achieving AGI. The research team notes that although GPT-4 is purely a large-scale language model, early versions include abstraction, understanding, coding, vision, mathematics, law, understanding human motivation, medicine, and even emotion. They pointed out that it had amazing capabilities in a variety of fields and tasks. .
far from humans
Despite Huang's recent predictions, many observers believe we won't see AGI anytime soon.At this time, even the term “AGI” is unclear, as there is no precise definition of what constitutes intelligence, the technical advisor said. Vaclav Winkarek said in an interview with PYMNTS.
“This is a subject that has been debated in scientific and philosophical communities for thousands of years,” he added. “Because the 'I' is not clearly defined (or understood), it is impossible to create a specification on which technology people can build systems.”
Yann LeCun, chief scientist at Meta and a deep learning pioneer, has previously disputed Huang's statements that AGI is on the horizon.Lucan expressed His belief is that current AI systems are still decades away from reaching a level of perception that includes common sense.
LeCun suggested that society will tend to develop AI closer to the intelligence level of cats and dogs before reaching human-level AI. He also pointed out that the technology industry's focus on language models and textual data is not enough to create the advanced AI systems that experts have long dreamed of.