A group of Microsoft AI scientists has published research that argues that the language model GPT-4 on OpenAI used in Microsoft’s Bing shows “sparks” of artificial general intelligence (AGI), i.e. human-level intelligence. If confirmed, it would represent an exciting new step for research. But is it really so?
Is GPT-4’s AI a general artificial intelligence? Microsoft thinks so
Just days after OpenAI published GPT-4, a team of AI scientists at Microsoft released research claiming such as the OpenAI language model, which powers Microsoft’s Bing AImonsters “sparks” of artificial general intelligence (AGI), that is, human intelligence.
Scholars, however, emphasize that it is just a first step towards a series of increasingly intelligent systems, rather than fully human-developed AI. Furthermore, the model of GPT-4 considered in the study is only one preliminary version, still under development by OpenAI and not necessarily the final version used in the products.
Despite these premises, the research has stated that GPT-4 is part of a new series of highly intelligent language models, capable of independently solve several difficult tasks in various industries, such as math, programming, medicine, law, and psychology. According to scholars, the capabilities of the model come very close to those of humans and often exceed those of previous models.
However, the scholars admit that despite the excellent performance in many activities, the model does not think exactly like a human being. AND there are still limits to overcome. Finally, it’s important to note that Microsoft has partnered with OpenAI, so there may be some interest in hyping the company’s work. Despite everything, scholars believe that GPT-4 represents a true paradigm in the field of information technology and beyond.