OpenAI Offers Beta Version of New AI Model
OpenAI has revealed a model artificial intelligence The new generative “O1” is able to think and answer more complex questions, especially in the field of mathematics, compared to previous models.
She said The company In a statement, it said, “O1 thinks before giving an answer.” The launch of this model represents a new step for the company towards achieving its goal of developing “general” artificial intelligence, that is, artificial intelligence similar to human intelligence, according to Agence France-Presse.
The company provided the trial version of “o1”, which will be available in the first phase to those who use “Chat GPT” for a paid fee.
Business Insider reported that the new model performs more complex tasks than previous models, and can solve more difficult problems in science and mathematics.
The site’s report indicated that the “O1” model “can compete with doctoral students in difficult tasks in physics, chemistry and biology.”
In a practical example, the new model beat the “GPT-4o” model, which was unveiled by OpenAI in May, in the qualifying test for the International Mathematical Olympiad, by a large margin.
The “O1” model was able to solve 83 percent of the exam correctly, compared to the other model that solved only 13 percent.
Agence France-Presse tested the model by asking it simple logical questions, and it achieved the same results as “GPT-4o,” but it took more time and gave more details about its answers, instead of generating them almost instantly.
Another difference is that the new model is currently unable to process or generate content other than text.
OpenAI said it tested its new model on solving mathematical problems or generating lines of code, and that it “competes with human experts on a large number of standardized tests that require strong thinking ability.”
She added that the model placed “among the top 500 students” in a US high school math competition.
OpenAI explained that the model “uses a series of thoughts, much like a human who can think for a long time before answering a difficult question.”
“He learns to recognize and correct his mistakes. He learns to break down difficult steps into simpler ones. He learns to try a different approach when the first one doesn’t work,” she explained.
The launch of “O1” comes at a time when OpenAI is seeking to raise subscriptions that will raise its value to about $150 billion, making it one of the most expensive unlisted companies in the world, according to American media.
The list of underwriting investors includes Microsoft and chip manufacturing giant Nvidia.
Last month, OpenAI said its chatbot, ChatGPT, now has more than 200 million weekly active users, double the number it had last fall.
The company said that “92 percent of the 500 largest companies in the United States use its products.”
