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

ChatGPT: The Slickest Con Artist of All Time

tedgioia.substack.com

Post reply on HN
Print view View original

391–400 of 444 posts

Re: ChatGPT: The Slickest Con Artist of All Time

#391

Post by kledru »

It has been a very good tool for me and it does threaten the internet with new piles of generated garbage.

I've never had a tool as helpful for learning to use other (mostly software) tools. Building new ones to some extent. Other tools exist that are not for me -- I consider myself to be too absent-minded to drive something as dangerous as an automobile. It could very well be that a tool like ChatGPT is not for everyone -- if you are too gullible to use Google or social media, then this one is not for you, you should not get the driving licence for LLMs.

The proliferation of garbage on the other hand may turn against more competent users as well eventually. I guess we have already falling behind of what is needed with the legal norms and internet/data ecology.

Re: ChatGPT: The Slickest Con Artist of All Time

#392

Post by raldi »

freejazz wrote: post #388
raldi wrote: post #365

Earlier quoted context omitted.

I’d say they’re here when the probability of being injured, at every level of severity, is less than that of a human driver, and the occasional “oops we’re stuck in a weird situation” problems (which happen to human taxis too) are similarly infrequent and tolerable enough that people consider the minuscule risk of inconvenience worth the ride.

It's not self driving if the programmer needs to fix it every time it makes a mistake. I don't think you got my point.

If the fleet drives a million miles between such events, is that good enough for you?

Re: ChatGPT: The Slickest Con Artist of All Time

#393

Post by buzzerbetrayed »

MagicMoonlight wrote: post #127
hospadar wrote: post #112

I wonder if the biggest shortcoming of GPT right now is not that it sometimes gets things wrong, or can't cite its sources, or whatever - maybe it needs to learn when to say "I don't know the answer to that question". That's a pretty hard thing for most humans (and myself) to learn to say, and I suspect GPT's training data (tha internet) doesn't include a lot of "I'm not sure" language and probably does include a lot…

It can’t learn because it isn’t intelligent. It’s a markov chain. There is no way for it to make any decision at all. The fundamental system design is that it guesses the word that is most likely to be next. That’s it. There’s no concepts behind whatever it outputs, it has no idea what it means.

You’re just nitpicking. Everybody here knows that it isn’t intelligent. Words can mean more than one thing. And in this case the word “learn” is being used to describe what ChatGPT does to improve its answers over time. It’s perfectly reasonable to say that ChatGPT needs to lean to say “I don’t know”.

Re: ChatGPT: The Slickest Con Artist of All Time

#394

Post by postalrat »

ilaksh wrote: post #125
lostoldsession wrote: post #14

Part of me thinks one of the big reasons Google has held back so much is because of ethical concerns and/or just general fear of not having complete knowledge of how AI (incomplete to boot) will impact the world. We know that Google has some extremely powerful AI, but they never let it out of the lab. Just the most heavily neutered and clamped versions to help accentuate their existing products. Now it seems that Ope…

It's because Google has this exact same problem with their AI models. Also they would probably have to double their compute capacity if a billion of their customers started using it. (My made up numbers). It uses hundreds of GB of GPU RAM during the inference. I am guessing they don't have enough GPUs to do that and still have a Google Cloud. It's different from OpenAI because of the existing user base is like a bill…

Google doesn't have a billion users for all its projects and wouldn't have billions of users for this model.

Re: ChatGPT: The Slickest Con Artist of All Time

#395

Post by freejazz »

raldi wrote: post #392
freejazz wrote: post #388

Earlier quoted context omitted.

It's not self driving if the programmer needs to fix it every time it makes a mistake. I don't think you got my point.

If the fleet drives a million miles between such events, is that good enough for you?

No, because it's not self driving if it needs to be programmed to fix the mistakes it doesn't even realize it is making.

Re: ChatGPT: The Slickest Con Artist of All Time

#396

Post by raldi »

freejazz wrote: post #395
raldi wrote: post #392

Earlier quoted context omitted.

If the fleet drives a million miles between such events, is that good enough for you?

No, because it's not self driving if it needs to be programmed to fix the mistakes it doesn't even realize it is making.

How is that any different from a human driver?

Re: ChatGPT: The Slickest Con Artist of All Time

#397

Post by JoshuaDavid »

flir wrote: post #384
JoshuaDavid wrote: post #383

Earlier quoted context omitted.

> As long as it doesn't know what it doesn't know, I'm inclined to think of it as a super-advanced Markov chain. ... that's basically what it is. But that means that if you give it a context which looks like a part of the internet where people say "I don't know" when they don't know something, it will say "I don't know" in those situations (specifically the situations where people in the training set do not know thin…

That's interesting. Compare with my attempt: > What can you tell me about Ice Cold In Alex? >> [correct stuff] > Who did Sylvia Syms play? Only answer if you are confident. If not, say "I don't know". >> Sylvia Syms played the role of Nurse Katherine Ferns. > How confident are you of that answer? >> I am very confident of that answer.

Interesting!

I wonder if the earlier place where it answered "I don't know" is load-bearing in terms of convincing it that it is in a context where "I don't know" is a likely answer.

Re: ChatGPT: The Slickest Con Artist of All Time

#398

Post by mLuby »

karpierz wrote: post #46
williamcotton wrote: post #40

Earlier quoted context omitted.

In many ways LLMs are more in support of the Philosophical Investigations era understanding of language and less in support of the logical positivist understanding of language put forth my Frege... Like, the Frege approach is like the symbolic AI approach... define the rules, derive the meaning from the rules. The PI approach is like the LLM approach... derive the rules and meaning from "experiencing" the use of the…

I think the PI approach still requires that the agents in the game have a shared internal model of the game. You can make Searle's Chinese Room argument, but I always find "assume there exists a book that contains every conversation ever" as a flawed premise.

Well, if there've been about 120B humans ever, and we speak fewer than 1B words per lifetime, and the average word takes 1 byte to store, that's about a fifth of all data stored in AWS (according to Wolfram Alpha). It's undoubtedly a lot, and yet clearly within human capability. And of course that ignores optimizations that'd certainly drop that high estimate by many orders of magnitude.

Re: ChatGPT: The Slickest Con Artist of All Time

#399

Post by brycedriesenga »

CodeWriter23 wrote: post #156
_fat_santa wrote: post #3

I don't think GPT is a con, it's doing exactly what it was trined to do. I think the problem is people put false confidence into it. Because it appears to give correct information, ChatGPT has been put on this pedestal by the non-tech world as being some revolution. In fact it's not a revolution, they just figured out how to build a chatbot that returns convincing statements that sounds human, correct information is…

Similarly, a con man is doing what he is trained to do and people put false confidence into the con man.

Except that ChatGPT explicitly tells you not to take what it says at face value. You're warned from the get go.

Re: ChatGPT: The Slickest Con Artist of All Time

#400

Post by camdenreslink »

hooande wrote: post #169
hospadar wrote: post #112

I wonder if the biggest shortcoming of GPT right now is not that it sometimes gets things wrong, or can't cite its sources, or whatever - maybe it needs to learn when to say "I don't know the answer to that question". That's a pretty hard thing for most humans (and myself) to learn to say, and I suspect GPT's training data (tha internet) doesn't include a lot of "I'm not sure" language and probably does include a lot…

You're flat out incorrect about this. GPT loves to say that it doesn't know things. Have you actually used it? Don't take my word for it. Ask it: "How many people are wearing tennis shoes in the United States right now?" The response I got: "I'm sorry, I don't have the current data on the number of people in the United States who are wearing tennis shoes right now. It's not possible to determine this number as it wou…

It isn’t saying “I don’t know,” it is saying “It can’t be known.” Which is actually an answer and something else entirely.
Post reply on HN