Amid a broad discourse at Kresge Auditorium, OpenAI CEO Sam Altman and MIT President Sally Kornbluth chatted on the potential and dangers of the swiftly advancing technology
What are the implications of the evolving field of artificial intelligence for the future of work, education, and humanity? On May 2, MIT President Sally Kornbluth and OpenAI CEO Sam Altman engaged in an extensive dialogue on the MIT campus, during which they addressed the topics above and more.
The success of OpenAI’s ChatGPT large language models has sparked significant investment and innovation in artificial intelligence. Since its release at the end of 2022, ChatGPT-3.5 has amassed the fastest-growing consumer software application user base, with hundreds of millions of individuals employing it.
Since then, OpenAI has partnered with Microsoft and showcased AI-powered image, audio, and video generation products.
In a full Kresge Auditorium, the event encapsulated the current frenzy surrounding AI while shedding light on future developments.
“I believe the majority of us recall our initial reaction upon seeing ChatGPT: ‘Oh my god, that is so cool!'” Kornbluth declared. “Our objective is to ascertain the nature of the forthcoming generation amidst this collective turmoil.”
For his part, Altman is content with the elevated anticipations surrounding his organization and the domain of artificial intelligence in its entirety.
“I find it remarkable that after two weeks of widespread concern over ChatGPT-4, by the third week, everyone was like, ‘Come on, where is GPT-5?'” Alston stated.
“That, in my opinion, says something remarkable about human ambition and expectation, which is why I believe we must all strive to improve the situation.”
Difficulties with AI
At the outset of their dialogue, Kornbluth and Altman examined the myriad ethical quandaries that artificial intelligence presents.
Altman stated,
“I believe we’ve made good progress in determining how to align a system with a set of values.” “Despite the common belief that these things are unusable due to the constant release of toxic waste, GPT-4 behaves reasonably as expected, and we’ve managed to program it to adhere to a specified set of values—not exactly flawlessly, but far exceeding my initial expectations thus far.”
Sam Altman
In many situations, there is disagreement among individuals regarding the precise behaviour that an AI system ought to exhibit, according to Altman. This complicates efforts to establish a universal code of conduct.
“How do we determine the proper values for a system?” Altman inquired. “How do we determine the appropriate behaviour of a system?” To what extent does society impose limitations instead of placing trust in the user regarding these tools? Although not everyone will utilize them to our liking, such is the nature of instruments.
Delegating significant autonomy to individuals is crucial; however, there are specific actions that a system ought not to perform, and those must be agreed upon collectively.
Kornbluth concurred that eliminating bias from AI systems and similar tasks would be challenging.
“It is intriguing to consider whether or not it is possible to create models that are less biased than humans,” she stated.
Kornbluth also raised privacy concerns about the enormous amount of data required to train today’s large language models. Altman asserts that these issues have been confronting society ever since the inception of the internet; however, AI is elevating the complexity and gravity of such concerns. Moreover, the prospect of advanced AI systems generates new inquiries in his mind.
“How shall we navigate the tradeoffs between privacy, utility, and safety?” Altman inquired. “The point at which we all individually determine those tradeoffs and the benefits that may be gained if an individual allows the system to train on them for the rest of their lives is a novel concept for society to navigate.” Unaware of the forthcoming responses.”
According to Altman, concerns regarding the energy consumption and privacy of AI systems will be alleviated as future iterations of AI models make strides.
Altman stated, “What we want from GPT-5, 6, or whatever it may be is for it to be the best reasoning engine possible.” “At this time, the only viable approach is to train the model using enormous amounts of data.
During this process, it learns something about performing highly, extremely limited reasoning or cognition, or whatever you wish to name it. However, considering its capacity for data memorization and the fact that it stores data in its parameter space, we might later conclude that it was rather extravagant with its use of resources.
It is anticipated that a solution will eventually be found to isolate the reasoning engine from the requirement for large amounts of data or storing said data in [the model], enabling them to be considered distinct entities.
Kornbluth further inquired as to how AI might result in the elimination of jobs.
“One aspect of AI professionals that irritates me the most is when they say with a straight face, ‘This will never result in the elimination of jobs.'” This is merely an additive. “Everything is going to be fantastic,” Altman exclaimed. “This will eliminate a significant number of existing jobs, alter how a substantial number of existing jobs operate, and generate an entirely new workforce.” This is a constant consequence of technology.
The potential of AI
Altman is convinced that advancements in AI will render it worthwhile to address the field’s present challenges.
Altman stated, “It would be a tremendous victory if we invested one percent of the world’s electricity in training a powerful AI that helped us figure out how to transition to non-carbon-based energy or improve deep carbon capture.”
Additionally, he stated that scientific discovery is the application of AI, which he is most intrigued by.
Altman stated,
“I contend that scientific discovery serves as the fundamental catalyst for human advancement and is the sole means by which sustainable economic growth can be propelled. The public is dissatisfied with GPT-4 They desire for things to improve. Everyone desires an ever-improving and more rapid existence, and science enables us to achieve these goals.”
Additionally, Kornbluth inquired of Altman for guidance for students contemplating a career. He advised pupils against setting themselves back.
Altman stated,
“The most essential lesson to learn early in one’s career is that anything is figurable, and no one has all the answers at the outset.”
“You simply blunder your way through, maintain a rapid iteration speed, gravitate toward the most intriguing problems, surround yourself with the most remarkable individuals, and have faith that you will iterate to the correct solution…
“One can accomplish more and more quickly than one can imagine.”
A portion of Altman’s larger message regarding maintaining optimism and striving to build a better future was embodied in the advice.
“Teaching our youth that problem-solving is futile and that the world is hopelessly shattered; the only recourse is to wallow in our bedrooms while we reflect on our utter filthiness—that is an extremely unproductive approach,” Altman lamented. “I hope MIT distinguishes itself from the majority of other academic campuses. Presumably, it is.
But you must all make combating this a personal mission in your lives. A prosperous and abundant future, as well as an improved standard of living for our offspring. That is the sole viable course of action moving forward.
That is the only way to have a functioning society… and I sincerely hope that each of you strives to overcome the anti-progress sentiment and the belief that “people deserve a great life.”