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

OpenAI starts testing ChatGPT premium

businesstoday.in

Post reply on HN
Print view View original

281–290 of 333 posts

Re: OpenAI starts testing ChatGPT premium

#281

Post by KineticLensman »

Obertr wrote: post #115
mensetmanusman wrote: post #2

“However, ever since the chatbot has come into action, it has been negatively impacting students' learning.” This sentence is out of nowhere, did ChatGPT write this article? How could someone think something so new is suddenly negatively impacting students across the board.

I believe this is going to transform the way people learn. Instead of learning in a structured they are going to learn exactly what they need. You don't need to know about physics molecules until you face a task related to it. And ChatGPT can direct you to learn more about it. I am in my bachelor's right now and I see how I can skip a lot of knowledge regarding some low-level computer science skills until I really ne…

> You don't need to know about physics molecules until you face a task related to it. And ChatGPT can direct you to learn more about it

I don't think ChatGPT will really help you learn about actual 'tasks' related to molecules except generating short text paras about them

Re: OpenAI starts testing ChatGPT premium

#282

Post by powersnail »

random_upvoter wrote: post #85
dcsommer wrote: post #11

Earlier quoted context omitted.

Cumulative damage may be very small now, at the beginning of this age, but educators I talk to certainly are having a reckoning about it. Whether this will ultimately be bad for education is an open question, but with the way many classes and homework are designed now, students using ChatGPT to answer questions certainly do learn less.

ChatGPT's knowledge doesn't run that deep. If ChatGPT can write a credible essay about a given subject it means your subject was rather generic to begin with. I was playing a bit with ChatGPT and I wondered how much actual knowledge could be stored in that model. So I asked ChatGPT whether it knew the song by Franz Schubert called "Am Feierabend" and if so, if it could tell me the subject/meaning of the song. "Certai…

ChatGPT has no concept of trying to be correct with regard to world knowledge. This doesn't just apply to obscure things. For instance, when I ask it about mainstream books and TV shows, it frequently misattributes words or actions to the wrong character. But not only that, it will then proceed to explain why the character said so, and how does it reflect on the character.

It's not about awareness or limits of knowledge. From the point of view of a language model, it doesn't matter whether it was Todd or Walter White who killed Lydia, or whether it was Kinbote or Shade who invented a phrase. It only tries to generate a response to your input, such that it is a plausible continuation.

Re: OpenAI starts testing ChatGPT premium

#283

Post by shrimpx »

shagie wrote: post #192
shrimpx wrote: post #173

Earlier quoted context omitted.

Thanks! Is this basically what chatgpt is, a thin frontend on GPT3? Or is there more to it?

ChatGPT has more "memory" to it. If you tell it something, it "remembers" it across the chat session. GPT3 is stateless. You give it the prompt and it responds back. Context doesn't carry between requests unless you (the user) pass it along. That can make "chat" sessions with GPT (not ChatGPT) expensive as passing along the past context of the chat session consumes a lot of tokens as it gets longer and longer.

Makes sense. Do you know if there exits a re-implementation of a ChatGPT-like stateful conversational interface on top of the GPT-3 API? A cursory search doesn't turn up anything.

Re: OpenAI starts testing ChatGPT premium

#284

Post by shagie »

shrimpx wrote: post #283
shagie wrote: post #192

Earlier quoted context omitted.

ChatGPT has more "memory" to it. If you tell it something, it "remembers" it across the chat session. GPT3 is stateless. You give it the prompt and it responds back. Context doesn't carry between requests unless you (the user) pass it along. That can make "chat" sessions with GPT (not ChatGPT) expensive as passing along the past context of the chat session consumes a lot of tokens as it gets longer and longer.

Makes sense. Do you know if there exits a re-implementation of a ChatGPT-like stateful conversational interface on top of the GPT-3 API? A cursory search doesn't turn up anything.

If you look at

    https://beta.openai.com/playground/p/default-chat?model=text-davinci-003
    https://beta.openai.com/playground/p/default-friend-chat?model=text-davinci-003
    https://beta.openai.com/playground/p/default-marv-sarcastic-chat?model=text-davinci-003
you can see a chat "session".

The issue is that to maintain the state, you need to maintain the history of the conversation.

For example, the friend chat starts out at 28 request tokens. I submit it, get a response back, and it's at 45 tokens. I add in a bit of my own commentary... and it's 59 tokens. I hit submit and its 80 tokens.... and so on.

https://imgur.com/a/wJ7LNAf

That session wasn't particularly insightful, but you get the idea.

The issue is that I did three requests with 28 + 59 + 88 tokens. As I keep going, this can add up. To maintain that context, my previous history (and the chat response) remains in the prompt for the next message growing arithmetically.

---

(edit)

Working from the chat instead of friend chat... a bit more interesting conversation.

https://imgur.com/Dog2jOT

Note however, that at this point the next request will be at least 267 tokens. We're still talking fractions of a penny ($0.002 / 1k tokens) - but we are talking money that I can look at rather than even smaller fractions. And as noted - it grows each time you continue with the conversation.

Re: OpenAI starts testing ChatGPT premium

#285

Post by speedgoose »

Balgair wrote: post #218
speedgoose wrote: post #182

Earlier quoted context omitted.

Dall-e 2 is completely overpriced and almost obsolete since MidJourney v4.

Hey, I'm getting into using these images more and more. I'm a bit gunshy of all the work that goes into midjourney. Do you know of any good tutorials on how to use it?

I kinda learned while playing with the technology.

Looking at what others are doing is very helpful.

This prompt book has also been the best reference for me, even though it’s not specific to MidJourney: https://openart.ai/promptbook

Re: OpenAI starts testing ChatGPT premium

#287

Post by Balgair »

speedgoose wrote: post #285
Balgair wrote: post #218

Earlier quoted context omitted.

Hey, I'm getting into using these images more and more. I'm a bit gunshy of all the work that goes into midjourney. Do you know of any good tutorials on how to use it?

I kinda learned while playing with the technology. Looking at what others are doing is very helpful. This prompt book has also been the best reference for me, even though it’s not specific to MidJourney: https://openart.ai/promptbook

Awesome book! Thank you!

Re: OpenAI starts testing ChatGPT premium

#288

Post by lancesells »

Al-Khwarizmi wrote: post #168
BoxOfRain wrote: post #61

Earlier quoted context omitted.

Yeah I'd happily pay decent money for a version that wasn't a fun sponge who'd probably remind the teacher they hadn't set any homework at the end of the lesson. I think its 'moral compass' is a bit Americentric too, it would be nice if its sense of what is proper could be customised for different countries.

Same here... but the puritan AI ethicists on Twitter are asking for even more self-censorship. Because there are ways to making it say racist things if you explicitly tell it to do so. As is normal, given that it's just a tool, the point is that it should do what we say. I guess for these people, high-school debates where one can be assigned racist/sexist/etc. positions are an abomination. I hope some non-American eq…

> American society (or at least the elites) seems totally engulfed by extreme puritanism, and happy to impose it on the rest of us (why can't we have porn apps on the iPhone, when there's porn made by perfectly legal companies who pay their workers and their taxes?)

I don't think there's as much puritanism as you think. For companies there's fear of losing money through bad press. If there was more money to be made in allowing "bad" things to be said in ChatGPT then that will certainly come. It's all walking on eggshells until 10 years from now when quarterly results aren't looking so good. The reason there's no porn on the iPhone is because they want to sell devices to the entire family and it's much easier to moderate.

I'm going to take a guess and say the elite are far more hedonistic than the average citizen.

Re: OpenAI starts testing ChatGPT premium

#289

Post by nullsense »

draugadrotten wrote: post #50
nullsense wrote: post #39

Earlier quoted context omitted.

My son is 7 years and he routinely asks me "why learn anything when you can just ask Google?". I can only imagine the impact this will have on that attitude over time.

Do you challenge him with simple questions like "How do you know the answer you get from Google is true?"

Yep. I attempt to educate him on epistemology as much as I can. He's a relatively skeptical kid, but he's also stubborn as hell and he gives somewhat of a recursive answer of "you just Google that too".

I try to impress upon him too that some things take hundreds of hours of explanation before you know enough for it to be functionally useful to you, so you can't always just Google everything. You have to learn things and build your foundations up. Alas he hates the idea of learning anything.

Re: OpenAI starts testing ChatGPT premium

#290

Post by shostack »

TurkishPoptart wrote: post #209
bilsbie wrote: post #32

My son asked if which of two animals would win in a fight and we just got a lecture about how fighting is bad. I hope they’ll give an option to turn off what I’m calling “hall monitor mode” That could be worth paying for.

Indeed, ChatBot's political biases have been intensifying. [1] [2] [1]: https://twitter.com/AlexEpstein/status/1606347326624215040?s... [2]: https://twitter.com/stone_toss/status/1606878417865129984

It's interesting considering the political leanings of Musk. If this becomes deeply embedded in society, there's absolutely an incentive to lean on the algorithm to push an agenda that can be difficult to detect bias within given how opaque it is.
Post reply on HN