Artificial intelligence is not only serving to improve productivity in work. Yes, it is impressive power Create Prompts to perform some specific taskssolve your doubts, or make images of different styles with text indications, but this technology goes much further.
Google Deepmind, Openai and Anthropic have AI models that are not only used for complex actions to generate programming code, but also in the scientific field. In various parts of the world, both chatgpt, and Other similar projects are having an impact on research that will be crucial for the future of humanity.
MIT experts have already created a model that could be the next step for the AGI and now it turns out that A group of Cambridge specialists He has found something equally impressive. Plato’s ancient mathematical problem, which had been solved for hundreds of years, has been resolved in the blink of an eye by artificial intelligence.
How GPT-4 has solved the problem of virtue?
Montaje/Freepik/Pixabay
Scientists have talked about how AI could read the minds of humans and it seems that everything goes a similar path with the recent discovery made by the GPT-4 model. Studyfinds reports that it has managed to solve one of the most complicated mathematical problems in history and on its own.
What does this mean? That has made decisions based on the request required to be able to achieve its objective, without obtaining indications of how to do it or why it is guided. It is the factor to double the area of a square that It is raised in the Menón de Plato dialoguealso known as men or virtue.
This says that it is not possible to get another picture with bending the sides because that only quadruples the area. Therefore, Pythagoras theorem is required Because it is not an action of whole numbers or simple arithmetic, but of more geometric reasoning. The solution so far was that the duplicate picture should be the “diagonal of a triangle formed with the original side”, or at least this is how the Gemini explains.
The milestone that Openai’s mentioned model has achieved is incredible because he has achieved a way to do it without focusing only on geometry. In this case, he proved to have self -correction capacity and error resistance or hallucinations, a behavior similar to an apprentice, since he uses his neuronal network database to assimilate existing information and achieve new solutions following that.
Well, the key to this 2400 -year -old inconvenience was Modern Algebra. According to what Chatgpt said, «the diagonal does not offer a new direct dimension», so he rejected the suggestions of the knowledge shortcuts about the subject to create a new alternative branch.
He directly calculated the value on one side to get the root square of the double of the original area and thus find the duplication result. For Plato’s era, algebra was not even close to having been a viable method because it still did not exist at that time.
Scientists explain that the power of this tool is allowing to go beyond memory responses that are inserted into the neuronal network. It responds to the PROMPTs and corrects being necessary until a possible result.
Here is also the theme of the Chat next development zone (ZPD)which suggests that with the proper orientation about the situation, artificial intelligence receives better reasoning and education, which is consistent with Vygotsky’s theory.
The test was done in February 2024 with the version of that time, so it is not known exactly if with the new versions such as the GPT-5 it would give the same results or improve when the Menón dialogue. It should be noted that this It does not mean that it is Aginor that it does not have indications of human understanding, but can be enhanced with the appropriate parameters.
Know How we work in NoticiasVE.
Tags: Artificial intelligence