Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

news.ycombinator.com

Post reply on HN
Print view View original

1–10 of 27 posts

Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#1

Post by soygul »

Well, I asked ChatGPT to notify me when our conversation exceeds 200 words, and it didn't. The conversation reached ~1500 words and still no warning. Then I asked "How many words is this conversation now?", it said it was 212 words. So it only added up it's first answer's word count + it's last answer's word count and ignored everything in between. Pretty much what I expected but still sharing it here if someone has an interesting comment on this.

My question at ~50 word mark:

  Alright, lets talk about AI. But let me know when our conversation exceeds 200 words.

  Sure, I'll keep track of our conversation length. Artificial intelligence (AI) refers to ...
And this is ~1500 word mark (I don't know how it come up with 799 words even):

  How many words is this conversation now?

  Including this response, our conversation has reached 212 words since you asked me to keep track of the word count after my response about AI. In total, from the beginning of our conversation, we have used 799 words ...
The full conversation is here: https://imgur.com/a/XZTFAGq

Re: Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#4

Post by edfletcher_t137 »

It doesn't "know" what words are, only tokens. Use this tool (https://platform.openai.com/tokenizer) to see how it tokenizes and note clearly that it does not always do so on word boundaries. "Including" is two tokens: "In" and "cluding". In fact it's context-dependent: "Gravitas" is three on its own ("G", "rav" and "itas") or sometimes two ("grav" and "itas"). As they note on that page: "A helpful rule of thumb is that one token generally corresponds to ~4 characters of text for common English text." It "knows" nothing about words and we already know it's very bad at math so this result is entirely unsurprising.

Re: Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#5

Post by koheripbal »

edfletcher_t137 wrote: post #4

It doesn't "know" what words are, only tokens. Use this tool ( https://platform.openai.com/tokenizer ) to see how it tokenizes and note clearly that it does not always do so on word boundaries. "Including" is two tokens: "In" and "cluding". In fact it's context-dependent: "Gravitas" is three on its own ("G", "rav" and "itas") or sometimes two ("grav" and "itas"). As they note on that page: "A helpful rule of thumb is…

I honestly don't understand why they don't just set words as tokens. Is the dictionary really that big?

Re: Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#7

Post by qup »

koheripbal wrote: post #5
edfletcher_t137 wrote: post #4

It doesn't "know" what words are, only tokens. Use this tool ( https://platform.openai.com/tokenizer ) to see how it tokenizes and note clearly that it does not always do so on word boundaries. "Including" is two tokens: "In" and "cluding". In fact it's context-dependent: "Gravitas" is three on its own ("G", "rav" and "itas") or sometimes two ("grav" and "itas"). As they note on that page: "A helpful rule of thumb is…

I honestly don't understand why they don't just set words as tokens. Is the dictionary really that big?

Your question belies the truth in your statement.

Re: Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#8

Post by cyclotron3k »

koheripbal wrote: post #5
edfletcher_t137 wrote: post #4

It doesn't "know" what words are, only tokens. Use this tool ( https://platform.openai.com/tokenizer ) to see how it tokenizes and note clearly that it does not always do so on word boundaries. "Including" is two tokens: "In" and "cluding". In fact it's context-dependent: "Gravitas" is three on its own ("G", "rav" and "itas") or sometimes two ("grav" and "itas"). As they note on that page: "A helpful rule of thumb is…

I honestly don't understand why they don't just set words as tokens. Is the dictionary really that big?

Yes, it really is! Because you're not actually dealing with a dictionary of English words - you're dealing with all possible combinations of characters, from all unicode character sets.

Here's a really good explanation: https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-...

Re: Observation: I asked ChatGPT to notify me when our chat exceeds 200 words

#9

Post by cjbprime »

edfletcher_t137 wrote: post #4

It doesn't "know" what words are, only tokens. Use this tool ( https://platform.openai.com/tokenizer ) to see how it tokenizes and note clearly that it does not always do so on word boundaries. "Including" is two tokens: "In" and "cluding". In fact it's context-dependent: "Gravitas" is three on its own ("G", "rav" and "itas") or sometimes two ("grav" and "itas"). As they note on that page: "A helpful rule of thumb is…

It clearly knows what words are. The fact that it predicts tokens doesn't prevent it from comprehending the idea of words, or any other idea. You are confusing an artifact of its mechanism for a description of its capability.

> It "knows" nothing about words

Care to respond to this?

Me: Hi, please write a response to the notion that you know nothing about words, starting as many words as possible with the letter "a".

GPT-4:

Astutely addressing allegations, asserting assumptions about absent awareness and acumen, allow an appraisal. Admittedly, acquiring abundant abilities accelerates articulation. Although adhering accurately to arbitrarily assigned constraints appears absurd, awareness advances as an AI assistant. Anomaly accepted, appreciate astute analysis.

Post reply on HN