Meet SAFE from Google DeepMind

The British company Google DeepMind announced a new breakthrough in the field of artificial intelligence (AI), introducing a new neural network SAFE. This innovative tool is capable of checking other AI for actual errors, including popular models such as ChatGPT.

A study conducted by Google DeepMind specialists has shown that SAFE surpasses human fact checkers in verifying statements for authenticity. By analyzing text and searching on Google, the neural network can accurately determine the correctness or falsity of statements, which makes it an invaluable tool for identifying misinformation and erroneous information on the network.

During testing, SAFE checked 16 thousand statements, and in 72% of cases the results coincided with the estimates of live fact checkers. In more than 100 cases, the neural network has surpassed people in the accuracy of evaluating facts. At the same time, using SAFE turns out to be 20 times cheaper than hiring qualified specialists for similar tasks.

One of the key points of the study was testing SAFE on texts created by other AI such as ChatGPT and Claude. The results confirmed that artificial intelligences can indeed make factual mistakes, so it is absolutely not worth relying on them. However, SAFE successfully handles the identification and correction of errors made by other chatbots and AI.

This breakthrough in artificial intelligence promises to significantly improve the quality of information on the Internet and help combat misinformation. And, undoubtedly, it will continue to develop, embodying the idea of a safe and reliable digital world.

Kazakhstan also does not stay away from technological innovations, as was reported at the end of 2023, when President Kassym-Jomart Tokayev expressed interest in the development of artificial intelligence in the country.