Microsoft researchers have published stunning research on the capabilities of GPT-4, the language model that makes up today’s ChatGPT. In summary, the research revealed that GPT-4 is the closest artificial intelligence to thinking like a human.
It would be a mistake today to deny that AI technologies are starting to resemble humans. Language model developed by OpenAI GPT-4 also stands before the eyes as the greatest evidence against this error. Alright, Can GPT-4 really think like a human?
The answer to this question was given by Microsoft, which provided access to the language model before GPT-4 was even released. Microsoft researchers shared the results they obtained after their scientific work with GPT-4 in their research published in April 2023. GPT-4, The closest artificial intelligence to think like a human was evaluated.
Let’s find out the critical result in their own words: On some missions, it performed better than humans!
“Despite being a purely language model, GPT-4 demonstrates remarkable abilities in a variety of fields and tasks, including abstraction, comprehension, vision, coding, mathematics, medicine, law, understanding human motives and emotions, and more.
We asked GPT-4 to write an essay on infinity of prime numbers in the form of a poem, draw a unicorn in TiKZ (a language for graphing in LATEX), create a complex animation in Python, and solve a high school level math problem. . It accomplished all these tasks easily and produced output that was basically indistinguishable from (or even better than) what humans could produce.”
However, these results are not proof that the GPT-4 can think like a human:
Researchers say GPT-4’s ‘artificial general intelligence (AGI)’, that is, a big step towards an artificial intelligence that can do the mental tasks of a human. stressed that. On the other hand, he reminded that GPT-4 has not yet been able to do anything that people do.
It was shared that GPT-4 still has many limitations and biases. Moreover basic arithmetic errors He said he did too.
Another point the research emphasized was the idea that the GPT-4 had common sense. It was shared that the GPT-4 gave much better results than other models that had previously entered the common sense test. But still that his ‘intelligence’ is not like a human being expressed.
GPT-4 is the language model accessible to all ChatGPT users today.
RELATED NEWS