Researchers at Microsoft believe that the latest AI technology showcases notable indicators of reasoning abilities similar to human beings. The company has published a research paper titled “Sparks of Artificial General Intelligence,” which delves into the central theme that technologists have been striving to achieve and worrying about for several decades.
The researchers articulate their findings on how the new AI showcases proficiency beyond language handling through its ability to tackle intricate and revolutionary tasks like mathematics, coding, vision, medicine, law, psychology, and more without the need for any specific cues. In a recent statement, Bill Gates said that artificial intelligence (AI) holds the potential to facilitate literacy acquisition in children within the upcoming 18 months.
Essentially, according to the researchers, GPT-4 already boasts a higher intelligence quotient than most individuals. They expressed their amazement and remarked that GPT-4’s proficiency is remarkably close to the human level. They elaborate on its exceptional capacity for reasoning, problem-solving, strategic planning, abstract thinking, comprehension of intricate ideas, quick learning, and experience-based learning.
Dr. Bubeck, a French expatriate and former Princeton University professor, was among the researchers involved in the study. One of their initial endeavors was to solicit GPT-4 to compose a mathematical proof that demonstrated the infinity of prime numbers in a rhyming format. Following GPT-4’s presentation of an impressive poetic proof, Dr. Bubeck was taken aback by the system’s proficiency both mathematically and linguistically.
During a seminar at the Massachusetts Institute of Technology in March, he recalled being puzzled by the sheer excellence of the technology. For several months, the researchers continuously observed the system’s complex behavior, leading them to conclude its possession of an adaptable and thorough understanding of human principles and expertise.
However, as Microsoft researchers were utilizing a preliminary edition of GPT-4 without prepping it to prevent problematic content such as hate speech or misinformation, validating the assertions highlighted in the document would not be feasible for external professionals. Microsoft asserted that the system accessible to the public does not possess the same degree of efficiency as the one used for experimentation.
While Microsoft claims that its latest AI technology showcases traits similar to human-like reasoning, skeptics are not in agreement. According to some researchers, GPT-4’s lack of familiarity with the physical world could hinder its performance in terms of general intelligence. Furthermore, Maarten Sap, a professor, and researcher at Carnegie Mellon University, has commented on the trend of some major companies using research papers as a form of PR promotion. Sap notes that such papers may not adhere to established scientific evaluation standards, with Sparks of A.G.I. being a prime example of this approach.
As per Alison Gopnik, a psychology professor affiliated with the A.I. research group at the University of California, Berkeley, while systems such as GPT-4 may possess significant capabilities, it remains uncertain whether the text produced by these systems is indicative of human logic and understanding, New York Times reported.
Via New York Times