Mike Shea

It's when companies like ChatGPT are valued at billions of dollars and then claim like, well, we don't really owe anything to the people that generated all of the training material that we were able to get that billions of dollars from. That pisses me off.

It's when companies like ChatGPT are valued at billions of dollars and then claim like, well, we don't really owe anything to the people that generated all of the training material that we were able to get that billions of dollars from. That pisses me off.

right because it's like well if it's if we're not valuable to us then don't use us oh well if we did that then we don't have any money at all because imagine how good chat gpt is without any training data so obviously that training data is worth it to them but then they're like yeah but we're not going to pay for it so that that kind of pisses me off i will say one other thing though which is i bet you you're not getting the value out of it that you think you're getting because i think there is a very poor recall on the data that it returns i've seen it myself and

right because it's like well if it's if we're not valuable to us then don't use us oh well if we did that then we don't have any money at all because imagine how good chat gpt is without any training data so obviously that training data is worth it to them but then they're like yeah but we're not going to pay for it so that that kind of pisses me off i will say one other thing though which is i bet you you're not getting the value out of it that you think you're getting because i think there is a very poor recall on the data that it returns i've seen it myself and

right because it's like well if it's if we're not valuable to us then don't use us oh well if we did that then we don't have any money at all because imagine how good chat gpt is without any training data so obviously that training data is worth it to them but then they're like yeah but we're not going to pay for it so that that kind of pisses me off i will say one other thing though which is i bet you you're not getting the value out of it that you think you're getting because i think there is a very poor recall on the data that it returns i've seen it myself and

And an example is if you feed a product in that has 20 gods described in the product and then say, give me a list of the gods in this product, it's not going to come up with 20 and it's not going to be accurate. So in the world of search and retrieval and natural language processing and AI, there's a thing called precision and recall.

And an example is if you feed a product in that has 20 gods described in the product and then say, give me a list of the gods in this product, it's not going to come up with 20 and it's not going to be accurate. So in the world of search and retrieval and natural language processing and AI, there's a thing called precision and recall.

And an example is if you feed a product in that has 20 gods described in the product and then say, give me a list of the gods in this product, it's not going to come up with 20 and it's not going to be accurate. So in the world of search and retrieval and natural language processing and AI, there's a thing called precision and recall.

Precision is how many of the results that you got returned to you were accurate to the query that you gave it. And recall is of the total amount of data that was out there, how much of that data did you get that you should have gotten? Recall is really hard to measure. It's sort of the dog that didn't bark problem. You don't know what you didn't get. Right. It's really hard.

Precision is how many of the results that you got returned to you were accurate to the query that you gave it. And recall is of the total amount of data that was out there, how much of that data did you get that you should have gotten? Recall is really hard to measure. It's sort of the dog that didn't bark problem. You don't know what you didn't get. Right. It's really hard.

Precision is how many of the results that you got returned to you were accurate to the query that you gave it. And recall is of the total amount of data that was out there, how much of that data did you get that you should have gotten? Recall is really hard to measure. It's sort of the dog that didn't bark problem. You don't know what you didn't get. Right. It's really hard.

And what I have seen with a lot of the AI bots that supposedly they call it RAG, like retrieval augmented generation, that they retrieve a bunch of information. You read a bunch of information and then you use the rest of the model to help you query that information. It's not bad if you're like, hey, give me three adventure prompts and you really don't care if it's accurate.

And what I have seen with a lot of the AI bots that supposedly they call it RAG, like retrieval augmented generation, that they retrieve a bunch of information. You read a bunch of information and then you use the rest of the model to help you query that information. It's not bad if you're like, hey, give me three adventure prompts and you really don't care if it's accurate.

And what I have seen with a lot of the AI bots that supposedly they call it RAG, like retrieval augmented generation, that they retrieve a bunch of information. You read a bunch of information and then you use the rest of the model to help you query that information. It's not bad if you're like, hey, give me three adventure prompts and you really don't care if it's accurate.

The problem is it didn't really read and process and understand everything you gave it, and it might only have read half of it. How do you know which half? And how do you know what stuff it's not building off of for it? So I don't think it's that accurate. So what I would recommend you do instead is read the book. Highlight the book. Use little tabs. Highlight stuff.

The problem is it didn't really read and process and understand everything you gave it, and it might only have read half of it. How do you know which half? And how do you know what stuff it's not building off of for it? So I don't think it's that accurate. So what I would recommend you do instead is read the book. Highlight the book. Use little tabs. Highlight stuff.

The problem is it didn't really read and process and understand everything you gave it, and it might only have read half of it. How do you know which half? And how do you know what stuff it's not building off of for it? So I don't think it's that accurate. So what I would recommend you do instead is read the book. Highlight the book. Use little tabs. Highlight stuff.

Copy and paste text and put it in your own text documents. That's what I've been doing when I've been doing my Dragon Empire campaign. I didn't run anything through AI, right? I read the book. So my feeling is I get, and I have friends who love it. I have friends who love to shoot ideas on this thing. But I don't think it's actually giving you the kind of value you think you're getting from it.

Copy and paste text and put it in your own text documents. That's what I've been doing when I've been doing my Dragon Empire campaign. I didn't run anything through AI, right? I read the book. So my feeling is I get, and I have friends who love it. I have friends who love to shoot ideas on this thing. But I don't think it's actually giving you the kind of value you think you're getting from it.

Copy and paste text and put it in your own text documents. That's what I've been doing when I've been doing my Dragon Empire campaign. I didn't run anything through AI, right? I read the book. So my feeling is I get, and I have friends who love it. I have friends who love to shoot ideas on this thing. But I don't think it's actually giving you the kind of value you think you're getting from it.