Microsoft has recently created a stir in the artificial intelligence space with its latest language model, GPT-4. This revolutionary model has shown an advancement in the realm of common sense reasoning, and has the ability to predict the next word in a sentence. An impressive 155-page paper was released by Microsoft’s team of scientists and it showcased the model’s ability to demonstrate human-like common sense in its behaviours.
GPT-4 boasts skills in areas such as abstraction, comprehension, coding, mathematics, law, vision, medicine, emotion and understanding human intentions. Though it still has certain limitations, such as producing incorrect results in basic calculations, it appears to be making progress in this field of common sense reasoning.
One particular prompt given to GPT-4 produced a poetic proof, which was so well-written mathematically and linguistically that it cast doubt in the minds of the researchers if they were conversing with an AI or a human.
Not everyone is pleased with the results. Maarten Sap, from Carnegie Mellon University, has expressed his own opinion about the Microsoft team’s study. Sap believes that it is nothing more than a PR stunt which does not abide by scientific review norms. Even though the Microsoft team is aware of the informal nature of the studied processes, doubts persist about the accuracy and openness of the conclusions that have been drawn.
Whatever the opinion may be, GPT-4 is a significant milestone in the progress of AI. It has already had a tremendous impact in the field of natural language processing, problem-solving and decision making. As further progress is made in artificial intelligence research and development, it is imperative to maintain a balance between creativity and rigor, to ensure responsible and reliable applications for these new technologies.
Microsoft is an American multi-national technology company based in Redmond, Washington. Founded in 1975 by Bill Gates and Paul Allen, the company remains one of the largest software corporations in the world. In addition to its widely used Windows operating system, Microsoft produces a variety of other products and services, such as Microsoft Office, Xbox, and Azure.
Dr. Bubeck is a former professor at Princeton University and was a member of the research team that had early access to ChatGPT-4 prior to its public release. Bubeck’s extensive work in computer science and artificial intelligence has led to numerous accomplishments, such as developing the first method for constructive induction of recursive functions back in 1987. He has also worked as a researcher at Microsoft Research since 2006 and was a member of the Microsoft Research Faculty Summit in 2011.
