Why install AI at home, and for whom

Running a language model on your own computer has become accessible: a few clicks, a download of a few gigabytes, and an AI answers without a connection. The course "Building with AI" approaches this choice from the point of view of a developer building a product. Here, we take the side of the advanced user who wants to use it themselves.

The good reasons

• Privacy. A patient file, a client contract, a personal diary: nothing leaves the machine. This is the most solid reason.
• Offline use. On the train, in a dead zone, in a place where the internet is forbidden.
• Volume without a bill. Sorting two thousand emails, summarising three hundred records: once the machine is bought, every request is free.
• Understanding. Nothing teaches you what a model is better than seeing one run, slow down and get things wrong, at home.

The bad reasons

• Having "the best AI". The best models of the moment remain those of the big online services, and what runs on an ordinary computer is one step, often two, below.
• Saving on a subscription when you use it ten minutes a day: the time spent installing and tuning costs more.
• Getting current affairs. A local model only knows what it has learned; without a search tool plugged in, it knows nothing that postdates its training.

Local does not automatically mean private

The model runs on your machine, but the application around it may send usage statistics, offer a web search or synchronise your conversations. Check the settings, and cut the connection during a test to see what keeps working. Privacy is verified, not assumed.

Key points

• Good reasons: privacy, offline use, large volumes, learning.
• Bad reasons: seeking the best model, current affairs, or saving on light use.
• The model is local; the application around it may not be. Check.