How a Chatbot Picks Its Next Word
A chatbot is not looking an answer up. It reads what has been written so far, picks a likely next word, feeds that back in and does it again, from patterns taken out of a very large amount of text.
What a learner can do afterwards
- Describe the loop of picking one word at a time and feeding it back in
- Say where the patterns came from, and that none of them is a rule a person typed in
- Explain why the same question can come back with two different answers
1 · Read
A chatbot is not looking your answer up in a book. It reads what is written so far, picks a likely next word, tacks it on, and repeats the loop until the answer is done.
Its sense of likely comes from patterns in a huge pile of text. Nobody typed in rules for every question. Reading all that writing taught it which words tend to follow which.
Ask it to finish The cat sat on the. It may pick mat, then read the longer line and pick purring next. Ask twice and you may get rug instead, because each pick leaves room for chance.
Likely words can still be wrong words. A chatbot can sound sure while stating a false fact, so double-check important claims in a book or a source you trust. Use it as a language helper, not an oracle.
Chatbots loop over likely next words learned from vast text, so answers vary and need checking.
2 · Watch
Take it off screen
Where it sits
Where this leads
Jobs that lean on this skill. Follow one to see everything it is built on.
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.