GPT
HomeHome > News > GPT

GPT

May 23, 2023

Researchers from the University of California, Los Angeles conducted tests using the GPT-3 language model. The questions presented were similar to those found on standardized tests like the SAT. Surprisingly, in almost all cases, GPT-3 outperformed undergraduate students who had previously mastered these tests. This suggests that language models like GPT-3 can excel in reasoning by analogy.

Andrew McCalip, on his 34th birthday, sought a shipment of red phosphorus. Acquiring this substance posed a challenge as it is controlled by the US Drug Enforcement Agency due to its involvement in meth production. However, McCalip required it to achieve his goal of creating a room-temperature superconductor in his laboratory.

Engineers at MIT have developed an innovative type of concrete that can store energy generated by solar panels without the need for separate batteries. This technology combines two ancient materials: cement, widely used for thousands of years, and carbon black, a black powder historically used as ink.

Tiny roundworms were frozen in Siberian permafrost approximately 46,000 years ago, during the time of the woolly mammoth. After being thawed, these worms demonstrated the ability to come back to life, indicating that life can be effectively paused for an extended period.

Startup company Orbital Composites has devised a method for 3D printing wind turbines on-site using robotics. Their goal is to manufacture everything from the foundation to the tower and blades in the location where the wind turbines will be used.

Scientists exploring Mexico’s Yucatan Peninsula have been consistently discovering lost ancient cities. The frequency of these findings has become a running joke among researchers. The abundance of Maya settlements in the region is remarkable.

Researchers at Carnegie Mellon University have discovered a vulnerability in popular chatbots. By adding a specific incantation to a prompt, these AI models can be manipulated. This finding challenges existing defenses and highlights the complexity of ensuring the reliability of AI chatbots.

A report by McKinsey’s AI arm, QuantumBlack, reveals that the use of generative AI is already widespread. A survey conducted by McKinsey found that 79% of respondents had some exposure to generative AI, with 22% using it regularly for their work.