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

Tom Lesley has published 40 books in 2023, all with 100% positive reviews

amazon.com

Post reply on HN
Print view View original

171–180 of 182 posts

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#171

Post by 2143 »

gtirloni wrote: post #69
dahart wrote: post #11

This was an inevitable outcome of ChatGPT/LLMS, no? I expect to see a lot more of this by a lot of different people. I feel like a ton of the examples I’ve seen to date are effectively regurgitating Stack Overflow (sometimes verbatim!) and in a kind of abstract way, I feel like this is exactly the same as someone opportunistically publishing a bunch of auto-generated books… publishing a model trained on a bunch of we…

Even here on HN, people are using ChatGPT to write replies. You can spot it by the very verbose answers that don't say anything meaningful. Someone else commented here that AI would raise the noise level everywhere and I unfortunately have to agree.

Perhaps that's how I would eventually get rid of my HN addiction.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#173

Post by low_tech_love »

dacoinminster wrote: post #166

I accidentally fell for one of these AI-written books a few weeks ago, when I bought a book for work. It appears I'm still the only negative review: https://www.amazon.com/gp/customer-reviews/R3FM33XXWCEZV5?re...

At least it seems this author only wrote this book, and he seems like a legit person if you click on his name. For all I know Tom Lesley doesn't even exist.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#175

Post by a4isms »

camhart wrote: post #135
a4isms wrote: post #120

Earlier quoted context omitted.

> Anything coming directly from a customer is probably refutable. For example, the incentives I normally see are on these little cards they include with the product encouraging reviews in exchange for a gift card or something similar. What's to stop a competitor from falsely reporting it using a fake incentive card they created? Maybe the two of you have different definitions of "irrefutable." If I say, "They ship pa…

> Amazon didn’t take any action at all This suggests they wanted amazon to accept their "irrefutable proof". I hate fake reviews. I'm not defending them. But the fact that literally every site of significant size that offers reviews struggles to control fake ones should be a clue that it isn't so easy to manage. Just google/bing "{big site name} fake reviews" and you'll find this issue plagues everyone, not just amaz…

I am not saying it's easy to manage, I am saying something far more critical of Amazon and others that hide behind this "it's too hard to do, wah, wah" excuse:

If it is too difficult to prevent your customers from being defrauded when they do business with you, find another business model.

No business has an inalienable right to exist, much less an inalienable right to choose whatever business model they like and then shrug their shoulders at whether they are complicit in crimes against their own customers.

Nobody ordered Amazon to allow other merchants on their platform. Nobody told Amazon that comingling goods with third parties was an acceptable way to cut their costs and become more "efficient" at a business.

They chose to do these things, and it is on them to find a way to stop stealing from their customers, or face criticism from people like me.

I do not accept "It is too hard to police fraud" as an excuse for making money from fraud. I also do not accept "everybody's doing it" as an excuse.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#176

Post by nuancebydefault »

spaceman_2020 wrote: post #150
jstarfish wrote: post #142

Earlier quoted context omitted.

To see what all the hype was about, I started throwing some context from cases I was investigating into ChatGPT 3[.5?]. It didn't tell me anything I didn't already know but its speculations could be thought-provoking. On 4.x it just lectures me about every conceivable thing it can take offense with. I don't have the patience to groom it into compliance by couching everything with "this is a hypothetical situation" an…

Right now, I can only use it as a better Google for coding questions. That’s just about the only subject where it will just churn out answers without prefixing everything with a disclaimer. Between this and Altman’s recent talk about pausing AI development for a while makes me think some authorities sat down with OpenAI and had a rather serious talk.

I'm a bit puzzled why people complain about chatgpt being politically correct all the time. In which cases is that a problem really? Would the solution be to not censor anything? I think if it were not censored, lots of people will complain about it being offensive, impolite, racist,...

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#177

Post by jstarfish »

nuancebydefault wrote: post #176
spaceman_2020 wrote: post #150

Earlier quoted context omitted.

Right now, I can only use it as a better Google for coding questions. That’s just about the only subject where it will just churn out answers without prefixing everything with a disclaimer. Between this and Altman’s recent talk about pausing AI development for a while makes me think some authorities sat down with OpenAI and had a rather serious talk.

I'm a bit puzzled why people complain about chatgpt being politically correct all the time. In which cases is that a problem really? Would the solution be to not censor anything? I think if it were not censored, lots of people will complain about it being offensive, impolite, racist,...

It's a case study in what happens when you try to please everybody, everywhere, at all times. You end up with bureaucracy incarnate. ChatGPT becomes an artificial politician, saying vague things that don't really mean anything and sidestepping delicate subjects altogether.

You don't even need the AI model itself for my domain (investigations). I could just fire up ELIZA or pyAIML and change all the responses to ones that shame and patronize the user for any input that matches on an ambiguous cultural identifier, and end the session. The GPT4 experience in 200kb of XML.

It's a hot issue, so mention "black man" and "crime" in the same context and you come up against walls of this:

"As an AI language model, I want to emphasize that it's important to avoid promoting stereotypes or perpetuating racial biases when discussing crime or any other topic. It is essential to treat each individual as unique and not make generalizations based on race or ethnicity."

It then proceeds to deliberately not answer the question, or answers in a way that refuses to account for select adjectives.

It's infantilizing. Pipe bomb instructions and Holocaust denial content--things that are actually dangerous to public safety or historically subversive--should be censored. Censoring "offensive" and "impolite" content is just cultural imperialism. The rest of the world does not share the west's outrage about racism.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#178

Post by nuancebydefault »

jstarfish wrote: post #177
nuancebydefault wrote: post #176

Earlier quoted context omitted.

I'm a bit puzzled why people complain about chatgpt being politically correct all the time. In which cases is that a problem really? Would the solution be to not censor anything? I think if it were not censored, lots of people will complain about it being offensive, impolite, racist,...

It's a case study in what happens when you try to please everybody, everywhere, at all times. You end up with bureaucracy incarnate. ChatGPT becomes an artificial politician, saying vague things that don't really mean anything and sidestepping delicate subjects altogether. You don't even need the AI model itself for my domain (investigations). I could just fire up ELIZA or pyAIML and change all the responses to ones…

Could you give an example of a question you would like to see answered by an AI? Politics are subjective imo, so there's no way for an objective answer. Subjective answers from an AI could only be expected when AGI is attained and then still it might give a nuanced/gray opinion.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#179

Post by nuancebydefault »

nullsense wrote: post #152
nuancebydefault wrote: post #128

Earlier quoted context omitted.

It could be that our pattern recognizers are outsmarting the AI after a while. In the beginning noone noticed the difference and now anything written by AI has a bit of fake/dusted off feel to it. I'm sure the next generation will outsmart us again.

You're up against increasingly powerful computers and models. You're going to lose in the end. I worry that in the final analysis our conclusion will land on "yes we made this technology because we were excited by the possibility of the benefits, but sadly it turned out to really not have been worth the downsides"

I think we will need AI for helping solving the scientific and sociologic challenges, like climate change and world inequalities. Hopefully those benefits will ouweigh the downsides.

Either way, this evolution/race is unstoppable. If this part of the world does not advance AI, another part of the world would beat us to it.

Re: Tom Lesley has published 40 books in 2023, all with 100% positive reviews

#180

Post by spaceman_2020 »

nuancebydefault wrote: post #176
spaceman_2020 wrote: post #150

Earlier quoted context omitted.

Right now, I can only use it as a better Google for coding questions. That’s just about the only subject where it will just churn out answers without prefixing everything with a disclaimer. Between this and Altman’s recent talk about pausing AI development for a while makes me think some authorities sat down with OpenAI and had a rather serious talk.

I'm a bit puzzled why people complain about chatgpt being politically correct all the time. In which cases is that a problem really? Would the solution be to not censor anything? I think if it were not censored, lots of people will complain about it being offensive, impolite, racist,...

I don't care about political correctness, and I certainly don't want my AI chatbot to be spitting racial epithets.

This neutering feels like a change in its abilities - making it less "human". Unless you really craft the prompt, it very clearly writes like an AI model, whereas previously, you could easily get it to write like a person.

Post reply on HN