Scientists have developed a computer system that will help create 100 times more powerful artificial intelligence

A team of scientists from the Massachusetts Institute of Technology (MIT) has presented a computer system based on the movement of light rather than electrons. The technology promises to improve the efficiency of machine learning by reducing power consumption.

According to the experiments, the new system demonstrates a more than 100-fold improvement in energy efficiency and a 25-fold increase in computing density, a measure of system power. Scientists believe that the new technology paves the way for the creation of large-scale optoelectronic processors to accelerate machine learning tasks. This approach will make it possible to include «brains» even in smartphones.

The authors of the study emphasized that the new system can be scaled up for commercial use in the coming years, as its components are created using existing technologies.

According to the scientists, current supercomputers limit the size of ML models, making their training economically unprofitable. In developing the new system, the scientists used arrays of vertical surface-emitting lasers, an innovative technology that is already used in industries such as remote detection using LiDAR and laser printing.

«We don’t know what the next generation of ChatGPT will be capable of if it is 100 times more powerful, but this discovery is what makes it possible for this technology to exist ,» the researchers said.

It is interesting that MIT researchers have developed a technology 100 times more powerful than the existing AI at a time when the public from all over the world is calling for the end of training systems that are more powerful than GPT-4. At the end of March, more than a thousand tech experts, including Elon Musk and Steve Wozniak, signed an open letter calling on all artificial intelligence labs to suspend AI training for at least 6 months. The signatories warned that uncontrolled AI systems that compete with human intelligence would pose «catastrophic risks to society» and called on OpenAI to suspend further development of ChatGPT.