Multimodality, which involves creating text,
Posted: Sun Jan 19, 2025 6:59 am
presentative of the press service of MTS AI LLC confirmed to a ComNews correspondent that the development of large language models has reached a turning point due to a lack of data for training on the Internet - there are no more relevant texts available, the old ones have already been studied, and new ones are not appearing at the required speed: "In the future, progress will continue in a different form: development will become more business-oriented, models will be more aimed at solving applied problems and return on investment, rather than entertainment or visionary functions."
Based on this, a australia telegram representative of the MTS AI press service highlighted the directions of AI development in large language models in the near future:
1. Narrow specialization. Models will focus on solving specific problems, such as performing complex mathematical calculations and fulfilling specific business requirements.
2. audio, and video content simultaneously within a single model.
3. Increasing productivity while reducing resource intensity. Models will evolve toward decreasing rather than increasing parameters. This transformation will allow them to be launched and trained on user devices without loss of quality.
4. Performing real-world tasks. AI will be able to fulfill specific user requests, such as working on a computer or other devices, etc. Thus, models will master the field of robotic automation.
5. Open models will catch up with proprietary ones in quality. There are already examples of open source models that have shown decent results at a modest cost. Thus, artificial intelligence will continue to develop not only thanks to corporations and to solve their needs, but also with the help of representatives of the IT community.
Vladislav Kurenkov, head of the Adaptive Agents research group at the Artificial Intelligence Research Institute (AIRI), agreed that the lack of data for training in the future is the prevailing opinion among t
Based on this, a australia telegram representative of the MTS AI press service highlighted the directions of AI development in large language models in the near future:
1. Narrow specialization. Models will focus on solving specific problems, such as performing complex mathematical calculations and fulfilling specific business requirements.
2. audio, and video content simultaneously within a single model.
3. Increasing productivity while reducing resource intensity. Models will evolve toward decreasing rather than increasing parameters. This transformation will allow them to be launched and trained on user devices without loss of quality.
4. Performing real-world tasks. AI will be able to fulfill specific user requests, such as working on a computer or other devices, etc. Thus, models will master the field of robotic automation.
5. Open models will catch up with proprietary ones in quality. There are already examples of open source models that have shown decent results at a modest cost. Thus, artificial intelligence will continue to develop not only thanks to corporations and to solve their needs, but also with the help of representatives of the IT community.
Vladislav Kurenkov, head of the Adaptive Agents research group at the Artificial Intelligence Research Institute (AIRI), agreed that the lack of data for training in the future is the prevailing opinion among t