recently the people of Open AI ended up releasing GPT-4, the new version of its Artificial Intelligence that constitutes the soul and central nucleus of ChatGPT.
From own demonstration As a result of the improvements, it became clear to us that this evolution of the AI would represent a serious leap from what we knew before.
But now Microsoft has triggered some warning signs, by publishing the results of a scientific investigation where it concludes that GPT-4 would actually be showing signs of human intelligence.
It is a goal that previously seemed like the material of science fiction stories and that would finally be achieved with this Artificial Intelligence, with all the implications and risks that this entails.
Under OpenAI’s own description, GPT-4 is now a much larger “multimodal (accepting image and text input, and emitting text output)” language model, which, while less capable than humans in many scenarios from the real world, he exhibits human-level performance in various academic and professional benchmarks.”
But the conclusions of the investigation developed by Microsoft seem to find indications that the AI would go a little further than what its general description suggests.
GPT-4 and ChatGPT would show real indications of being a human-like Artificial General Intelligence (AGI).
A group of Microsoft researchers have published in arxiva new investigation titled Sparks of Artificial General Intelligence: Early experiments with GPT-4which translates as “Sparks of Artificial General Intelligence: first experiments with GPT-4”.
In said document, the team of scientists affirm that GPT-4 has already shown early signs of being an AGI, which would mean that it now has capabilities that are at the same level as that of a human or in some cases slightly above:
“GPT-4’s performance is surprisingly close to human-level performance and often far exceeds previous models like ChatGPT.
Given the breadth and depth of GPT-4’s capabilities, we think it could reasonably be seen as an early (but still incomplete) version of an Artificial General Intelligence (AGI) system.
We show that, beyond its language proficiency, GPT-4 can solve novel and challenging tasks spanning math, coding, vision, medicine, law, psychology, and more, without requiring any special prompting.”
These are some of the conclusions that the team of researchers has conceived after testing the capabilities of GPT-4. Where the document itself dynamite part of these ideas. Since in several sections the limitations of the AI are listed.
While it is repeatedly emphasized that the claims about GPT-4 and its evolution into an AGI do not mean that it is capable of doing anything a human can or that it has unknown internal motivation, consciousness, and goals.
Even so, it will be worth paying attention to the evolution of what happens with this model.