When OpenAI unveiled ChatGPT in 2022 it introduced generative artificial intelligence to the mainstream and set off a chain reaction that quickly pushed the technology into industries scientific research health care and the daily routines of people everywhere.
What does the future hold for this powerful yet still evolving tool
That question brought together hundreds of researchers business leaders educators and students at MIT’s Kresge Auditorium on September 17 for the first MIT Generative AI Impact Consortium Symposium. The goal was to exchange ideas and discuss where generative AI might be heading.
“This is a defining moment because generative AI is advancing at a rapid pace. It is our duty to make sure that wisdom and responsibility grow alongside the technology” said MIT Provost Anantha Chandrakasan as he opened the inaugural event. The symposium was organized by a consortium of MIT researchers and industry experts formed earlier this year to explore how AI can be shaped for society’s benefit.
MIT President Sally Kornbluth highlighted the importance of this effort noting that the world is depending on experts and business leaders to address both the opportunities and the ethical challenges brought by generative AI. “A big part of MIT’s role is to keep advancing innovation while making sure the technology can be trusted for real world applications” she said.
Yann LeCun chief AI scientist at Meta delivered one of the keynote speeches. He argued that the most significant progress will likely not come from making language models like GPT Claude or Llama even larger. These models rely on vast datasets to generate new content but LeCun believes the next step lies in developing “world models” that learn by interacting with their environment in a way similar to how young children learn through experience.
According to LeCun a four year old child has already absorbed as much visual data as some of the biggest language models. He believes that world models will become a central feature of future AI systems. A robot equipped with such intelligence could handle entirely new tasks on its own without additional training. For LeCun this makes world models the most practical route to creating robots capable of real world usefulness.
Even if AI systems become more advanced and humanlike he does not see them slipping out of human control. Safeguards will always be necessary he explained but humanity has been building such rules for centuries to align behavior with the greater good. “We will need to design these guardrails but by design the systems will not be able to bypass them” he said.
Another keynote speaker Tye Brady chief technologist at Amazon Robotics spoke about the growing role of generative AI in robotics. He noted that Amazon already applies generative AI in its warehouses to guide robots and improve the way goods are moved which helps streamline operations. He believes future progress will focus on collaborative robotics creating machines that work closely with people to make processes more efficient. “Generative AI is the most impactful technology I have seen in my entire robotics career” Brady said.
Speakers and panelists also examined how generative AI is influencing businesses from large corporations such as Coca Cola and Analog Devices to startups like health care company Abridge. MIT researchers presented ongoing projects that use AI to reduce noise in ecological imagery design systems that limit bias and hallucinations and enable language models to learn more about the visual environment.
By the end of the symposium Vivek Farias professor at MIT Sloan and faculty co lead of the consortium shared his perspective. He hoped attendees left not only with inspiration but also with a sense of urgency to turn the possibilities of generative AI into reality.
