When Opeli introduced ChatgPT to the world in 2022, he introduced generative artificial intelligence into the mainstream and began the snowball effect, which led to rapid integration with the industry, scientific research, health care and everyday life of people using this technology.
What next for this powerful but imperfect tool?
Bearing in mind this question, hundreds of researchers, business leaders, teachers and students gathered in Mit's Kresge Auditorium for the inaugural symposium MIT Generative AI Impactium (MGAIC) September 17 to share insights and discuss the potential future of general AI.
“This is a key moment – generative artificial intelligence moves quickly. Our task is to make sure that as technology progresses our collective wisdom maintains the pace,” said Mit Provost Anantha Chandrakasan to start the first Symposium of MGGAIC, a consortium of industry and researchers, the myth launched in February to use the generative force for a good society.
Emphasizing the critical need for this joint effort, President Mit Sally Kornbuth said that the world is counting on lecturers, researchers and business leaders, such as MGGAIC, to meet the technological and ethical challenges of generative artificial intelligence as technology develops.
“The part of the responsibility of the myth is to keep these progress for the world. … How can we manage magic (generative AI) so that everyone can probably rely on this in the case of critical applications in the real world?” Kornblleng said.
To speak, the speaker of Yann Lecun, the main scientist of AI in Meta, the most exciting and significant progress in generative artificial intelligence will probably not result from further improvements or expansion of large language models, such as LAMA, GPT and Claude. Through training, these huge generation models learn patterns in huge data sets to get new outputs.
Instead, Luniun and others are working on the development of “world models”, which they learn just like a baby – by seeing and interacting with the world surrounding them through a sensory contribution.
“The 4-year-old has seen the same number of data as the largest LLM. … The world model will become a key element of future AI systems,” he said.
A robot with this type of global model can learn to perform a new task without training yourself. Lecun believes that global models are the best approach for companies to make robots intelligent enough to be generally useful in the real world.
But even if future AI generative systems become smarter and more human by including global models, Lecun is not worried that robots run away from human control.
He said that scientists and engineers would have to design handrails to maintain future AI systems, but as a society we have been doing it for millennia, developing the rules for adapting human behavior with common good, he said.
“We will have to design these handrails, but thanks to construction the system will not be able to escape from these handrails,” said Lecun.
The main speaker of TYE BRADY, the main technologist from Amazon Robotics, also discussed how generative artificial intelligence can affect the future of robotics.
For example, Amazon has already included AI generative technology in many of its magazines to optimize the way they travel and transfer material to improve order processing.
He expects that many future innovations focus on the use of generative artificial intelligence in cooperation by building machines that allow people to become more efficient.
“Genai is probably the most influential technology that I have witnessed through all my career of robotics,” he said.
Other presenters and panelists discussed the influence of generative artificial intelligence in enterprises, from Largescale, such as Coca-Cola and analog devices, to startups such as Care Care Company Abridge.
Several members of the MIT Faculty also talked about their latest research projects, including the use of artificial intelligence in order to reduce noise in ecological image data, designing new AI systems that alleviate bias and hallucinations, and enabling LLMS to learn more about the visual world.
After a day spent studying the new generative AI technology and discussing its consequences for the future, MGAIC of the Faculty, Vivek Farias, Professor Patrick J. McGovern in the mit Sloan School of Management, said that he hopes that the participants left “a sense of possibilities and urgency to make this possibility.”
