A OpenAI launched the new GPT-4 to succeed the GPT-3.5 (which has already been rolled out to Microsoft Teams) officially. The resource arrives to feed ChatGPT and, according to the developer, it is the most advanced system ever made. Microsoft itself had already announced that it would launch the new version this week.
The creators even claim that the new GPT will be more creative and expands its limit from 3 thousand to up to 25 thousand words. Likewise, technology is able to provide “safer and more useful answers“. The tool can also solve difficult problems with greater precision and will still have the ability to perform tasks more optimally.
OpenAI even ran some capability tests with GPT-4 and talked about the results:
We spent 6 months making GPT-4 more secure and aligned. GPT-4 is 82% less likely to respond to requests for prohibited content and 40% more likely to produce factual responses than GPT-3.5 in our internal assessments
According to the website, more than 50 experts from the sector were consulted, who were in charge of analyzing questions related to technical aspects. In addition, details related to security and protection against violation of AI rules have also been worked out in the new stage of technology.
OpenAI also stated that the new GPT surpassed the current one by obtaining higher scores in the Biology Olympiads and in the Uniform Bar Exam, something equivalent to the OAB exam. Despite being presented at this time, the application of GPT-4 was already being done in apps such as Duolingo.
