Microsoft researchers say AI shows how people behave, while critics accuse them of being wrong.
In a recent experiment, computer scientists at Microsoft confronted their new artificial intelligence with a complex physical puzzle. The unexpected response was an ingenious suggestion: Arrange the eggs on a book, then put the laptop upside down. This feat raised questions about an emerging form of intelligence. In March, a 155-page paper was published claiming that this system represents a step towards artificial general intelligence, rivaling human cognitive capabilities. The study has been published on an online research repository.
The AI Debate: Microsoft and Artificial General Intelligence
Microsoft has sparked heated debate by claiming that the technology industry is creating something similar to human intelligence. According to Peter Lee, Head of Research at Microsoft, This assertion at first raises doubts. Then this doubt develops into frustration and anxiety. Microsoft’s paper, “The Sparks of Artificial General Intelligence,” addresses this crucial question. Examines possible effects and risks. building a A machine equal to or superior to the human brain It presents great opportunities and risks.
However, the claim of artificial general intelligence (AGI) It can jeopardize the reputation of IT professionals. Different interpretations of intelligence can clash. The previous year, Google had fired a researcher for claiming that the AI system had bypassed Microsoft’s claims. Despite these reservations, some believe that the industry is getting closer to creating a new, inexplicable AI system. This system would be able to generate unprogrammed human responses and insights.
To explore this idea further, Microsoft has reorganized its research labs By creating custom groups. One such group is led by Sebastian Bobic, lead author of the Microsoft AGI article. These technological developments raise fundamental questions about the limits and prospects of artificial intelligence. They encourage us to reconsider our understanding of the cognitive capabilities of machines.
The power and understanding of OpenAI’s GPT-4
OpenAI’s GPT-4, the most powerful system in its class, was used by Microsoft researchers. by investing $13 billionMicrosoft has become a close partner of OpenAI in San Francisco. Under the leadership of Dr. Bobik, the team challenged the GPT-4 with various missions. For example, they asked him to write a poetic mathematical proof showing infinity for prime numbers. The result was so remarkable that Dr. Bobik was amazed: “What’s going on?” he asked during a seminar at MIT in March.
Over the months, researchers have documented the complex behavior of GPT-4. They found that he demonstrated a “deep and flexible understanding” ofHuman concepts and skills. Dr. Lee points out that GPT-4 is impressive for text generation. Its strength lies in analyzing, synthesizing, evaluating, and judging texts. Notably, when they asked the system to draw a rhinoceros in programming, it was able to do so. Even after deleting the token from the horn, he still managed to complete this task. The GPT-4 has also been successful in other miscellaneous missions. He assessed the risk of developing diabetes based on information about a person. In addition, he wrote a letter of political support to his wife modeled on Mahatma Gandhi and led a Socratic dialogue on the misuse of linguistic paradigms.
Criticisms of GPT-4 Claims
Some AI experts criticize Microsoft’s article, seeing it as An opportunistic attempt at condescending remarks On misunderstood technology. Indeed, they argue that, in theory, GPT-4 lacks the familiarity with the physical world necessary for general intelligence. Thus, there is a gap between Microsoft’s claims and the reality of the situation.
Furthermore, Martin Sapp, a researcher at Carnegie Mellon University, has criticized “artificial general intelligence sparks” asSelf and informal public relations strategy. According to him, it is often adopted by large companies. The claims in the document cannot be verified by outside experts. Tests were run on an unduplicated copy of GPT-4. Allison Gopnik, a professor of psychology at the University of California, Berkeley, raises doubts about the source of the transcript generated by GPT-4. She wonders if this is really the result of human logic or common sense. It is critical that systems such as GPT-4 not be objectified. It should not be seen as a competition for human capabilities, explains Dr. Gopnik.
