With GPT-4, OpenAI pushes the limits of artificial intelligence to new heights

With GPT-4, OpenAI pushes the limits of artificial intelligence to new heights ...

With GPT-4, JVTech's news JVTech's "Always more impressive" OpenAI further pushes the limits of artificial intelligence.

Published on 03/16/2023 at 11:25

Share :

OpenAI lifted the veil on its new language model last night, and within 24 hours, everyone was agreed upon.

It's a strange thing to be around GPT-4.

After several Microsoft engineers gave us a chance to see GPT-4, we knew that the groundbreaking new language model that OpenAI has promised for several years would be revealed this week. If many were expecting a formal announcement on March 16, Sam Altman finally caught everyone by surprise by presenting the evolution of his new technology two days later (March 14).

GPT-4 is a multimodal language model, meaning that it can respond to more complex situations while generating more "human" responses. OpenAI's CEO even states that his new child, as perfectible as he is, is capable of "passing the bar exam."

GPT-4 is able to integrate the world's most well-known schools.

GPT-4 might be chosen for Stanford because of his reasoning ability, which has improved sufficiently to pass most mathematics and language exams without much difficulty. If GPT-3.5 fails the most difficult math and language exams, he will be among the top 10% of students. Conversely, if GPT-3.5 was placed in the same class, he would be among the 10% of those with the worst grades.

OpenAI takes a photo credit for this article.

Sam Altman and his teams explain that his new language model obtained his grades without any training for these exams. Despite this, she still states that a minority of them were included in the training models, while excluding the result with the lowest score.

Improvement of these models over time is inevitable. But the issue isn't so much whether GPT-4 can pass these exams, as whether the evaluation technique is effective. This had also caused several difficulties in world (and French) universities during the meteoric rise of ChatGPT. However, Gary Marcus, professor of psychology at New York University, moderates the results obtained by the model, saying that: "GPT-4 can really operate as a Stanford student."

You may also like: