This week, Microsoft developers presented the Kosmos-1 AI model to the public. It is capable of independently analyzing images, recognizing textual information on them, executing voice commands, taking IQ tests, solving puzzles, etc. As part of Kosmos-1, the softmaker’s specialists managed to combine several input modes such as text, image, sound, video, etc. They plan to use the AI model to develop “artificial general intelligence” (AGI) that can solve problems as well as a living person.
One of Microsoft’s key partners in the development of AI solutions is the North American startup OpenAI. He gained wide popularity thanks to the sensational ChatGPT neural network. The OpenAI management has already announced that the startup plans to start creating AGI.
The developers of the corporation created the Kosmos-1 model on their own without the participation of OpenAI. For her training, they used 800 GB of text found on the global network. Thanks to this, Kosmos-1 managed to bypass all modern language models. Whether this is actually the case cannot be determined, unfortunately. The source code for Kosmos-1 has not been published by the software giant.
