The creators of ChatGPT assert that their latest AI model possesses the capacity for logical reasoning and thought processes akin to a human.
The initial installment of the OpenAI o1 series, recognized as the first in its line, was introduced on Thursday as a preliminary version. OpenAI expressed expectations for frequent updates and enhancements in the near future, with most ChatGPT users gaining access to it incrementally.
As stated on their website by the creators of ChatGPT, these models are designed to delve deeper into problem-solving before providing a response, mirroring the thought process of an individual. Through training, they acquire the skill to refine their thought process, experiment with various strategies, and acknowledge their errors.
OpenAI provided a few instances demonstrating the capabilities of these innovative models. For instance, they can be utilized by medical researchers for the annotation of cell sequencing data, or by physicists for generating intricate mathematical formulas required in quantum optics.
Noam Brown, a research scientist at the company, elucidated the potential of the new AI models on Thursday. He publicly expressed an aspiration for future versions to ponder upon issues for seconds, hours, days, and even weeks. He acknowledged that the inference costs (such as elevated energy bills) would increase, but posed the question: "What price would you pay for a novel cancer drug? For pioneering energy storage solutions?"
On Thursday, the energy demands of AI, a pressing concern, were set to be brought up for discussion among high-ranking White House officials and leading tech executives. Participants in the meeting, as a source informed CNN, included Sam Altman, CEO of OpenAI, Ruth Porat, a senior executive from Google, and Dario Amodei, CEO of Anthropic.
Despite its limitations, the new OpenAI model lacks many elements that contribute to ChatGPT's utility, such as web browsing for information and image/file uploads. However, the company stated that it marks a significant leap in addressing complex reasoning tasks.
In trials conducted by OpenAI, o1 exhibited performance levels akin to PhD students on difficult benchmark tasks in physics, chemistry, and biology. Furthermore, in the qualifying exam for the International Mathematics Olympiad, the upgraded models accurately resolved 83% of the problems.
The tech sector is eagerly anticipating the advancements in this business of AI, with OpenAI's o1 series promising frequent updates and enhancements. In the realm of tech businesses, collaborations between entities like OpenAI and medical research bodies could lead to significant breakthroughs in fields like cell sequencing data annotation.
