Why trust Greenbot
At the TED AI Conference in San Francisco on Tuesday, OpenAI’s research scientist Noam Brown introduced a groundbreaking artificial intelligence (AI) model called the o1 series. Brown captivated the audience as he outlined a transformative shift in AI development, advocating for a new approach he termed “system two thinking.” This strategy, he claimed, has the power to revolutionize industries by enhancing the way AI processes information beyond simple data scaling.
Unlike traditional AI models, which rely heavily on scaling data and compute power, the o1 series models process information more thoughtfully, making them ideal for complex tasks in fields like scientific research, coding, and strategic decision-making.
“We’re no longer constrained to just scaling up the system one training. Now we can scale up the system two thinking as well, and the beautiful thing about scaling up in this direction is that it’s largely untapped,” Brown stated during his presentation.
Brown emphasized that while scaling data and compute has been crucial for AI’s progress, a paradigm shift toward a slower, more deliberate form of processing—known as “system two thinking”—is necessary. He illustrated this with a personal anecdote from his work on Libratus, an AI that mastered poker. Brown revealed that allowing the AI to think for 20 seconds equaled scaling the model by 100,000x without increasing resources, challenging previous assumptions about AI development.
“It turned out that having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer. When I got this result, I literally thought it was a bug,” Brown shared.
The o1 model’s deliberate processing method has demonstrated its capabilities, achieving an 83% accuracy rate in the International Mathematics Olympiad qualifying exam—a dramatic improvement over the 13% achieved by OpenAI’s GPT-4o. Brown argued that this shows the o1 model’s potential to excel in industries that rely heavily on data-driven insights, such as healthcare, energy, and finance.
“This isn’t a revolution that’s 10 years away or even two years away. It’s a revolution that’s happening now,” Brown asserted, highlighting the immediate impact of OpenAI’s latest innovation.
In addition to academic success, Brown discussed how the o1 model could bring substantial benefits to industries. For instance, he noted its potential to accelerate research in cancer treatment and renewable energy development. Despite the slower and more expensive nature of the model, he argued that its increased accuracy could justify the cost for industries where precision is paramount.
“When I mention this to people, a frequent response that I get is that people might not be willing to wait around for a few minutes to get a response, or pay a few dollars to get an answer to the question. But for the most important problems, that cost is well worth it,” he explained, emphasizing the value of investing in AI models that prioritize accuracy over speed.
Brown acknowledged the higher operational costs associated with the o1 model, which runs at $15 per million input tokens and $60 per million output tokens—considerably higher than GPT-4o. Nonetheless, he argued that enterprises that require high-accuracy outputs could find the investment worthwhile, especially in competitive sectors like healthcare and finance.
“Now we have a new parameter, one where we can scale up system two thinking as well—and we are just at the very beginning of scaling up in this direction,” Brown concluded, urging industries to embrace this innovative shift in AI technology.
