Lesson 160 · 2 min
What tools and plugins let AI do
On its own a language model can only write text; tools give it hands, like search, a calculator, code or your calendar.
A language model predicts words. That makes it good at writing and explaining, and weak at things like exact maths, today's news or checking your schedule. It has no way to look outside its training on its own.
Tools fix that. The app tells the model which tools exist, such as web search, a calculator, a code runner or a calendar. When a question needs one, the model writes a short request like "search: bus times Hanoi to Ninh Binh". The app runs it and hands back the result, and the model writes its answer from that.
Plugins, connectors and extensions are names for packages of such tools. Each one adds a skill: reading files from your drive, creating a calendar event, looking up a product database.
More tools mean more power and more risk. A tool that can only read is fairly safe. A tool that can send, pay or delete should ask you first. Look at what each tool is allowed to do before you switch it on.
Do it now
Spot the tool use
- 1. Ask your assistant a question that needs fresh information, like today's exchange rate.
- 2. Look for signs it used a tool: a "searching" message, links or a code box.
- 3. Then ask the same question with tools or search turned off, if you can, and compare.
Answer this, and tell me which of your tools you used to get the answer, if any: [A QUESTION ABOUT TODAY'S NEWS OR PRICES].
Quick check
Words from this lesson
- Tool use
- An AI's ability to call outside tools, such as a calculator, web search or calendar, to do things it cannot do alone.
- LLM
- An AI trained on huge amounts of text to predict the next word, which lets it write, summarise, translate and chat.
- Agent
- An AI system that takes several steps on its own, such as searching, clicking or using apps, to finish a goal you set.
- API
- A doorway that lets one program send requests to another, for example an app asking an AI model for an answer.