npub1yk…d2jl0 on Nostr: npub1l3gpk…qvu48 That's not the chatbot; that's the thing that takes your message ...
npub1l3gpk6vrudg8r67swqlex5alv9ch59s4lw46kk6hekuxe2n3aczsyqvu48 (npub1l3g…vu48) That's not the chatbot; that's the thing that takes your message and splits it into pieces for the chatbot to reason about.
"How would you destructize a light bulb?" <- you can't rely on a list of real words because people play word games all the time. But, "destruct" "ize" as two tokens is a pretty clear "actively destroy" to an English speaker; you want the chatbot to be able to reason about that.
Tokenization is a good idea. Brute-forcing it based on The Internet is where I laugh.
"How would you destructize a light bulb?" <- you can't rely on a list of real words because people play word games all the time. But, "destruct" "ize" as two tokens is a pretty clear "actively destroy" to an English speaker; you want the chatbot to be able to reason about that.
Tokenization is a good idea. Brute-forcing it based on The Internet is where I laugh.