Deep dive · module 26 of 100
Local AI
Models that run on your own computer or server.
After this lesson
You understand when local AI is useful and what the limitations are.
In view · How it connects
What local AI needs
The model requires local computing power and sufficient memory.
Choose a model that fits your task and your hardware.
Check whether the application still uses external services or storage.
Part 1 of 3 · Own device or server
Your preference for less movement is on. Choose each part yourself.
When does local help?
A local model may suit a defined task on local hardware. First, check which components actually remain on the device. A local app may still use a cloud model, search service, or synchronisation function; the name of the app does not prove offline operation.
Compare quality and effort
Achievable quality and speed depend on model, task, memory, computing power, and settings. A local model is not automatically worse or better than a cloud model. Compare the same fictitious tasks and also measure management, control, and total costs.
Start with a limited trial
Start with a small, non-confidential trial and check model licence, hardware requirements, and connections. Assess several common and difficult cases using the same quality threshold. Maintenance and access to the device remain your own responsibility.
New practice example
From explanation to a usable result.
A local model also requires attention to storage, calculation time, updates, and the app used.
Fictional practice situationFictional: you want to rewrite a short internal text on your own laptop. The chosen app optionally offers online synchronisation.
View a possible solution
This is how it can be done
First test with fictional text. Check model source, usage agreements, data storage, and synchronisation. Then measure the quality and waiting time on your own device.
Why this works
A model that can start is not yet a suitable working environment. The same test helps to weigh convenience, control, and performance.
Make it slightly more difficult
Make a list of data that the app can retain outside of the original text, such as history or technical logs.
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Watch and understand · with narration
Where does your prompt go?
With a computer-generated voice, on-screen text and captions in your chosen language. Start and pause whenever you like. Open the video separately.
Read the text version of the video
- In the cloud, your input travels through a connection to an external server. The model runs there.
- The answer returns to your device. Check the terms for storage and the use of your data.
- Locally, the model runs on your own computer or server. This needs memory and computing power.
- Look at the whole chain. A local app can still use external search services or synchronisation.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
A model that can start is not yet a suitable working environment. The same test helps to weigh convenience, control, and performance.
A step further
Make a list of data that the app can retain outside of the original text, such as history or technical logs.
Try & discover · no obligation
Follow the route of your data
Let the model calculate locally and then switch on online search. Does the entire application remain local?
- Model calculationOwn deviceNo external step
- SearchNo external search stepNo external step
- StorageOn the deviceNo external step
In this practice model, calculation, searching and storage remain on the device. In a real application, you also check logging, updates and telemetry.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Explain in simple language when local AI is a good idea for a small team, and mention 2 limitations. Use only the provided fictitious facts. Make missing information visible and indicate what a human must check before use.
Apply it
A team wants to summarise locally. The trial model fits in the memory, but you do not yet know how well it handles missing data.
Which next trial provides useful information?
- Test ordinary incomplete fictitious documents against the same quality threshold and check external connections.
- Assume that suitable hardware also guarantees correct answers.
- Use real files immediately to save time.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What do you determine with a small local AI trial?
- That local AI is always better.
- Whether quality, speed, data flow, and management suit the task.
- That there is never an external connection.
Extra practice
Name one use case where local AI would be useful to you, and why. Work with a fictitious situation and name one concrete control point before you would use the result.
Summarising internal documents without sending data outside. Reason: privacy of customer information. I compare the outcome with the confirmed input, leave unknown data blank, and adjust the result before use.
