OpenAI has introduced its latest artificial intelligence model called ChatGPT o1-preview, which is designed to solve complex logical tasks more efficiently and to be able to “think more humanly” about the question asked, taking more time to think before giving an answer.
“We trained these models to spend more time thinking about a problem before reacting, much like a human would. Through training, they learn to improve their thinking process, try different strategies and recognize their mistakes.” stands in contribution on the OpenAI website.
In testing, OpenAI found that its latest model, o1, performed similarly to doctoral students on challenging benchmark tasks in physics, chemistry, biology, and even math or coding. While the previous ChatGPT-4o solved only 13% of the assignments correctly in the qualifying exam for the International Mathematical Olympiad, the o1 model was able to cope with 83% of the tasks thanks to its logical abilities.
The new model can thus be useful, for example, in healthcare to annotate cell sequencing data, in physics to generate complicated mathematical formulas, or in development to perform multi-step workflows.
ChatGPT o1, on the other hand, lacks, for example, the ability to browse the web or upload and work with files. Moreover, according to the magazine The Verge slower than the GPT-4o version. Although o1 focuses more on thoughtful and more logical answers, it still failed to solve a problem called “hallucinations,” a situation where artificial intelligence generates incorrect answers.
ChatGPT o1 is now available and can be selected in the version overview. At the moment message limits are set at 30 for o1-preview and 50 for o1-mini, the company is working hard to limit them. In the future, users using ChatGPT will also get access to the o1-mini version for free without a subscription.
Source: www.cnews.cz
