AI computers to run tasks both locally and in the cloud
Published
Many computer manufacturers are now working on a new type of AI computer, known as an agentic AI PC. The idea is that a computer can carry out AI tasks of various kinds and decide where each task should be done. Some tasks can run directly on the computer's own hardware, while others are sent to large AI models in the cloud.
The biggest advantage of this model is that token consumption may be reduced. Tokens are the small units of text that AI models count when they read and write text, and they are often used to measure cost and resource needs. If simpler tasks can be done locally, fewer tokens need to be sent to the cloud. This can make use cheaper and more efficient.
At the same time, several challenges remain. Local AI models are often smaller than the large models in the cloud, and they may therefore perform certain tasks less well. It is also difficult to determine in advance which model is best suited for each task. In addition, computers need enough power and memory to run AI models without the user noticing it in everyday use.
The suppliers thus envision a future in which computers and cloud services work together. How well the idea works in practice remains to be seen, however, and that will become clear when more users begin working with these computers in their everyday lives.
Sources: Computer Sweden
More Sweden news
- More people in Sweden use AI, most trust its answers
- SMHI issues several yellow snow warnings for parts of Sweden
- Sweden's parliament votes on a change that could revoke citizenship
- Planned data center in Skellefteå could become Sweden's largest electricity user
- Vietnam and Sweden want to develop cooperation in more areas
- Swedish gaming industry thriving this year
- Binance targeted in coup attempt: Report points to security weaknesses
- Sweden drops in climate ranking – new record low