True or false?
Mark hears lots of claims about AI, and doesn't know which are true and which are false. Bit gathers the ten most common misconceptions, with videos that explain each one in more depth.
1 · 2
Understanding and confidence
One: AI understands like a person. Actually, it guesses the next word, very well. Two: if AI sounds sure, it's right. False, it can still make things up, which is called hallucination.
3 · 4
Memory and knowledge
Three: AI remembers everything you've ever said. Actually, it only sees what's in its context window. Four: AI searches the web every time it answers. No, mostly it uses what it learned before. For fresh information, you have to let it search.
5 · 6
Answers and models
Five: getting a different answer when you ask again means AI is broken. No, that's temperature, randomness on purpose. Six: the biggest model is always the best. For small tasks, small models are faster and cheaper.
7 · 8
Data and jobs
Seven: you can paste anything into AI. Don't paste passwords, card numbers or customer data. Eight: AI will replace every job. AI replaces tasks, and jobs change how they're done.
9 · 10
Doing and learning
Nine: AI can only chat. Now AI agents also use tools and do work for you. Ten: learning with AI means letting AI do it for you. Real learning means letting AI ask you questions, while you do the work.
This article is based on the video 10 common misconceptions about AI from the Mark học AI channel. Watch the video (in Vietnamese) to see the animations.