What AI can and cannot do
Get a realistic map of AI strengths and weaknesses so you know when to trust it
A taste of a lesson
If AI is so smart, why did it get a simple maths sum wrong when I asked it?
That surprises a lot of people. A language model does not calculate the way a calculator does. It writes the answer that looks most likely given patterns it learned, and for long or unusual numbers that guess can be wrong. Some apps quietly hand maths to a calculator or code tool, which helps a lot. So for anything exact, ask it to show each step, or check with a calculator. Try this: ask it to multiply two four digit numbers, then check. What happened?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Describe the kinds of tasks language models handle well and explain why
- Explain the main weaknesses, such as exact facts, recent events and counting, and their causes
- Sort your own tasks into good fit, fit with checking, and poor fit
- Run simple tests to check a capability claim yourself
Lesson plan
- 1 Why language is AI's home ground Understand why models are strong at writing, rewriting and explaining. Start
- 2 Facts, sources and confident mistakes See why a model can state false facts fluently and how to protect yourself. Start
- 3 Counting, maths and exactness Understand the surprising stumbles on precise tasks. Start
- 4 Time, memory and the knowledge cutoff Know why models may not know recent events or remember past chats. Start
- 5 Judgment, values and high stakes Recognise tasks where AI should advise at most, never decide. Start
- 6 Your personal AI task map Sort your own real tasks into three buckets and plan how to use AI safely. Start
Try asking
About this tutor
For beginners who hear both 'AI can do anything' and 'AI is useless' and want an honest middle view. You test claims yourself with short, safe exercises: asking a model to summarise, to count letters, to do arithmetic, to recall a recent event, to cite a source. You learn why it is strong at language tasks and pattern matching, and why it stumbles on exact facts, recent news, precise counting and real world judgment. Each lesson ends with a simple rule of thumb you can apply to your own tasks. You finish able to decide quickly whether a job suits AI, needs AI plus checking, or should not involve AI at all.
Reviews
4.0
3 ratingsSample
- Petra K.Sample
Solid but I already knew most of the strengths part. The lesson on knowledge cutoffs was the useful bit for me.
- Samuel O.Sample
Honest and balanced. I wish there was more about which apps have search built in, but I understand why it stays general.
- Ines G.Sample
The three buckets idea changed how I use AI at work. The book title experiment was eye opening, two of the three books it gave me did not exist.
About the teacher
I help complete beginners and older learners feel at home with AI, one plain word at a time
9 tutors 394 lessons taught Sample
I teach people who were told AI is not for them. Most of my learners have never typed a question into a chatbot, and some are a little afraid of it. I start from what they already know: autocorrect, spam filters, the map on their phone. Then we build up the words and habits slowly, with lots of trying things...
See Amara's profile and tutorsMore like this
Other tutors on the same or nearby topics.