“I really like Things that, if they work, really count – even if they have no very high chances of work, “said Sam Altman, co -founder and CEO of Openai, to a host of students who filled the ‘Commemorative church to hear him speak on May 1. , the coolest, most important and exciting thing. And so, it was worth continuing. »»
Altman was in Cambridge for a series of events organized by XFund, a venture capital company in the stadium Founded in Harvard in 2011 (but it operates independently and without funding from the university). During the Memorial Church event, David Parkes, dean of the Paulson School of Engineering and Applied Sciences of Harvard, presented to Altman the “Xfund Experiment Cup” 2024, which is awarded to “extraordinary founders of the best universities in the world” .
The event moderator – managing the general partner of Xfund Patrick Chung ’96, JD -MBA ’04, which invested in the first company of Altman – in the crowd of more than a thousand excited students, who started to queue for the event in a case where Harvard Yard an hour before his start. “They look at you and they aspire, they dream of having the type of impact you have had on the world,” said Chung. What advice would he give to their age?
“I think you can just do things in the world,” said Altman in response, with characteristic and deliberative breaks between his words. “You don’t need to wait, you don’t need to get permission. You can – even if you are totally unknown in the world, without almost no resources – you can always accomplish an incredible amount. »»
Some have criticized Altman’s incessant thrust for progress in the face of the potential dangers of advanced AI. Altman himself signed, last year, a letter describing AI as a Risk of extinction for humanity; Testimony before the CongressHe said that “if this technology is going badly, it can go wrong.” But it is skeptical about the slowdown in progress is the means to alleviate these threats. By working on AI 20 years ago, he did not know that it would become the technology he is today – and “even now, the most critical decisions we make, we are not aware of their importance “in the moment, he said.
Progress, he continued, is unpredictable and difficult to regulate. “There is someone, somewhere, to Openai at the moment, making an important phenomenal discovery – I don’t know what it is, I just know that it will happen statistically – which could very much shape the future “He said. “I totally agree, on the surface, that we should feel a huge responsibility and get big decisions, but you don’t always know when it happens.” THE results Progress is what must be checked: “Decide to deploy the GPT-5 or not, deciding where the threshold should be-we have made extreme care,” he said.
A large part of the decision -making power on the speed with which the development and deployment of a new AI rests in the hands of private for -profit companies like Openai – and some are concerned about this power which is concentrated among if Few individuals. These concerns spread to the OpenAi board of directors in November, when Altman was ousted as CEO –partly on security problems– Before being restored a few days later. (Professor of Eliot University and the president of Harvard Emeritus Larry Summers was Bring to the new board In the wake of the crisis, probably to deepen his leadership experience.)


During an interview Harvard magazine And other publications before his speech, Altman spoke of the challenges of AI development in the private sector. “I think it is exciting that (the development of AI) occurs in private industry,” he said, “but at a different time or a different configuration of the world, it would happen to government.” As is not the case, a “shared understanding” between the government and the industry is important. Now, he said, there is too much difference between the two groups about what IA progress will look like in the years to come. At the end of April, Altman joined the Artificial Intelligence Security and Security Council of the Ministry of Internal Security To help facilitate these conversations.
However, Altman declared during his speech, if he could return to 2015, when he co -founded Openai, he would start it as a for -profit society – not as a non -profit organization, like the organization was until 2019. The development of advanced AI simply needs too many resources, Openai website said, to be supported by a non -profit model.
With the transition to a profit-oriented structure, Openai has become a closed source: its underlying code and its implementation details are not made available to the public. This makes many in the academic world to study the model, and The researchers turned to bypassing measurements such as the analysis of models in chatgpt outputs. In response to a question of Harvard magazine About these challenges,, Altman said that Optai led university partnerships: “To give an example, we give university researchers access to the weights of the GPT-4 basic model,” he said, although there is No information accessible to the public on this program. “But finding ways to collaborate and unlock the academic world, for lack of a better word, seems really important.”
A question that some members of the academy seek to address is how IA models produce their results. Chatgpt interior mechanisms are currently not understood – and although university researchers work on the problemIt is difficult to do it with the chat of the closed source in particular. Some believe that it is important to understand these mechanisms before developing autonomous models or Apply AI in high issues parameters like health care.
“We continue it. I have nothing specific, as we fell for it Harvard magazine On the question of whether Openai also studies the problem. But to “understand” how the Chatppt produces outings might not be an achievable objective, he said: “Do I think we can understand these systems importantly? Yes, “he said. “Do I say that we can understand exactly what every artificial neuron does?” I’m not sure. But I’m not even sure that this is the right question on which to concentrate. »»
Meanwhile, the reason for profit continues to progress. After Altman’s conversation in Memorial Church, he heard eight Harvard and MIT students’ launches – selected by a thousand bids – as part of a competition to win a $ 100,000 investment Xfund.