New Feature Proposal - Leverage ChatGPT AI for Automatic Book Summary and Analysis - Page 2

mikhail_fil · 06-03-2023, 03:45 PM

Quote:

Originally Posted by BetterRed

One of the participants in the thread I referenced in post #6 poses the following question:

I get the impression you didn't even look at the thread. Rather than arguing with those who say something can't/shouldn't be done, you could try engaging with someone who's already done it.

BR

The irony. I have not made even one argument with anyone here, so it appears that you didn't really look through this thread as you would have noticed that I only made the initial question.

I did go to that thread, and may have missed that point. I am completely new to this "forum" format, and just learning how everything works here.

I appreciate you pointing out that someone made that plugin, and I will go there now.

Just because my question elicited an argumentative response, this doesn't mean that this was my intent.

BetterRed · 06-03-2023, 05:07 PM

Quote:

Originally Posted by mikhail_fil

The irony. I have not made even one argument with anyone here, so it appears that you didn't really look through this thread as you would have noticed that I only made the initial question.

I did go to that thread, and may have missed that point. I am completely new to this "forum" format, and just learning how everything works here.

I appreciate you pointing out that someone made that plugin, and I will go there now.

Just because my question elicited an argumentative response, this doesn't mean that this was my intent.

I didn't think, even for a moment, that argument was your intent, if I had I wouldn't have posted a reply. Look for the recent replies in the plugin ideas thread from member "Bradles".

Added: I see you found it

BR

mikhail_fil · 09-30-2023, 04:23 PM

Came back to agree that all who said it was a crazy idea were right - ChatGpt bulshits so much, it's scary. If you don't know the topic, you won't be able to decipher the lies either.

Sent from my SM-S908U using Tapatalk

Wiggo · 09-30-2023, 05:23 PM

Quote:

Originally Posted by mikhail_fil

Came back to agree that all who said it was a crazy idea were right - ChatGpt bulshits so much, it's scary. If you don't know the topic, you won't be able to decipher the lies either.

I find it really good that you tried it out, drew your conclusions and shared them with us. Thank you!

Comfy.n · 09-30-2023, 05:30 PM

Quote:

Originally Posted by kovidgoyal

Quite apart from the dead-end nature of the current approach to AI, this isn't feasible even if the AIs were actually intelligent since they can only be prompted with very limited amounts of text at the moment. Maybe that will change in the future, but currently as far as I know there is no way to ask the AI a question involving thousands of pages of text.

Even local language models can't be prompted with large texts, I found recently, to my dismay. That's the key point, as Kovid pointed out.

As for AI hallucinating, in my experience with small texts, I find it does quite a good job

Quoth · 10-01-2023, 06:13 AM

AI doesn't hallucinate. It fails.
It doesn't learn. It's fed data, either files or via web links, by parasite owners.
It produces plausible junk because it has no concept of fact or fiction, because that's too much work for the human owners. There is no understanding.

All the terms are chosen to misdirect.

Comfy.n · 10-01-2023, 07:40 AM

I hear Bard now has a fact-checking button:
https://wondertools.substack.com/p/5...to-use-bard-ai

But I use mostly OpenAI's ChatGPT.

Comfy.n · 10-01-2023, 09:06 AM

Quote:

Originally Posted by Quoth

AI doesn't hallucinate.

um.. I've just witnessed an example of AI hallucinating..

Click image for larger version

Name: MWSnap 2023-10-01, 08_57_34.png
Views: 158
Size: 76.7 KB
ID: 204031

Comfy.n · 10-01-2023, 09:09 AM

...and:

Click image for larger version

Name: MWSnap 2023-10-01, 08_42_34.png
Views: 167
Size: 44.2 KB
ID: 204032

Quoth · 10-01-2023, 10:48 AM

It a failure of the software. It's a program + database that's incapable of thought or hallucination. That's a marketing term to excuse that it's an over hyped system created by a parasitical corporation.

DNSB · 10-01-2023, 12:47 PM

Quote:

Originally Posted by Quoth

It a failure of the software. It's a program + database that's incapable of thought or hallucination. That's a marketing term to excuse that it's an over hyped system created by a parasitical corporation.

Don't be coy. Tell how you really feel about AI.

Quoth · 10-01-2023, 03:15 PM

Quote:

Originally Posted by DNSB

Don't be coy. Tell how you really feel about AI.

https://futurism.com/critics-microso...ain-ai-drought

Edit:
Then there is the electricity.

After the "training" the queries (end user) also use a lot of cooling water and electricity.

MSWallack · 10-02-2023, 11:24 AM

So you want to feed AI more copyrighted work for the AI to enhance its own data repository from which it can create even more copyright-violating derivative works? I'd rather not put Calibre in the spotlight of authors and publishers who are finally beginning to challenge AI that uses their works without consent and without royalties.

10-01-2023, 06:13 AM	#21
Quoth Still reading Posts: 15,114 Karma: 111120239 Join Date: Jun 2017 Location: Ireland Device: All 4 Kinds: epub eink, Kindle, android eink, NxtPaper	AI doesn't hallucinate. It fails. It doesn't learn. It's fed data, either files or via web links, by parasite owners. It produces plausible junk because it has no concept of fact or fiction, because that's too much work for the human owners. There is no understanding. All the terms are chosen to misdirect. Last edited by Quoth; 10-01-2023 at 06:15 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Feature Request - Tag Proposal in Metadata Editor	Dirk-71	Calibre	3	07-22-2020 03:06 AM
Calibre - Feature proposal for server	Mr. Nietzsche	Server	2	08-21-2019 10:48 AM
calibre new feature proposal - avoid breaking paragraph	rjwse@aol.com	Conversion	9	03-17-2016 09:26 PM
Feature proposal: attachments	chrisberkhout	Calibre	1	08-07-2013 11:40 PM
Automatic fill "Summary" text at ADD BOOK function	ttacta	Library Management	0	10-10-2012 09:13 AM

09-30-2023, 04:23 PM	#18
mikhail_fil Enthusiast Posts: 25 Karma: 10 Join Date: Jul 2016 Device: PC / ipad air	Came back to agree that all who said it was a crazy idea were right - ChatGpt bulshits so much, it's scary. If you don't know the topic, you won't be able to decipher the lies either. Sent from my SM-S908U using Tapatalk

10-01-2023, 07:40 AM	#22
Comfy.n want to learn what I want Posts: 1,686 Karma: 7908443 Join Date: Sep 2020 Device: none	I hear Bard now has a fact-checking button: https://wondertools.substack.com/p/5...to-use-bard-ai But I use mostly OpenAI's ChatGPT.

10-01-2023, 09:09 AM	#24
Comfy.n want to learn what I want Posts: 1,686 Karma: 7908443 Join Date: Sep 2020 Device: none	...and:

10-01-2023, 10:48 AM	#25
Quoth Still reading Posts: 15,114 Karma: 111120239 Join Date: Jun 2017 Location: Ireland Device: All 4 Kinds: epub eink, Kindle, android eink, NxtPaper	It a failure of the software. It's a program + database that's incapable of thought or hallucination. That's a marketing term to excuse that it's an over hyped system created by a parasitical corporation.

10-02-2023, 11:24 AM	#28
MSWallack Right, Except When Wrong Posts: 364 Karma: 4323767 Join Date: Aug 2007 Location: Indianapolis Device: Kindle Oasis 3 (sometimes iPad Mini).	So you want to feed AI more copyrighted work for the AI to enhance its own data repository from which it can create even more copyright-violating derivative works? I'd rather not put Calibre in the spotlight of authors and publishers who are finally beginning to challenge AI that uses their works without consent and without royalties.