OpenAI Launches ChatGPT Plus, Starting at $20 Per Month

Aiming to monetize what's become a viral phenomenon, OpenAI today launched a new pilot subscription plan for ChatGPT, its text-generating AI that can write convincingly human-like essays, poems, emails, lyrics and more. From a report: Called ChatGPT Plus and starting at $20 per month, ChatGPT Pro delivers a number of benefits over the base-level ChatGPT, OpenAI says, including general access to ChatGPT even during peak times, faster response times and priority access to new features and improvements.

Google is wrong

[Danathar]

It's far easier to ask Chatbot how to do something (like how do I do something in VIM) than googling it. Google's execs should be very nervous.

Re:

[iMadeGhostzilla]

I want both: links to pages written by people and a responses produced by a model that may or may not have hallucinated when generating them.

The problem will be when pages that should be written by people start including generated context.

Re:Google is wrong

[ljw1004]

The problem will be when pages that should be written by people start including generated context.

I think that many of the pages that google links me to are already generated. I do a search like "how to remove scobie layer" and get back a lot of pages that clearly did copy/paste from other sites. There'll be a heading "What is scobie" followed by a few copy-pasted paragraphs, then "how do I make scobie", then "when is scobie used", then "what is a scobie layer", and somewhere near the bottom "How to remove a scobie layer" so on. I suspect they're mostly machine-generated with only minimal human intervention, presumably to game SEO and get advertising dollars. (I think the SEO battle has been lost by the search engines for several years).

Also, for the queries I do, many of the results in the first two pages of results are clearly written by journalists who understand less about their subject than ChatGPT. Some of them just unquestioningly parrot a PR. Some of them repeat someone's quote without any recognition that the quote is logically incoherent or doesn't answer the question.

Re:

[iMadeGhostzilla]

It seems then that two sources of knowledge are emerging: the statistical knowledge related to what has been done many times which AI can cover better than Google, and the expert knowledge that is too difficult to aggregate mechanically that a person must cover.

The obvious problem for the people is they need to have enough useful expertise that their work cannot be replaced by AI. The upside is with these AI tools and real world practice you can gain expert knowledge faster than before. The downside is you

Re:

[blackomegax]

ouroboros. eventually all content on the internet is AI made, feeding new AI training models.

Re: Google is wrong

[nsbfikwjuunkifjqhm]

It's also far easier for the chatbot to get that answer dangerously wrong, or even answer an entirely different question than the one asked, and for you to have no idea because it won't tell you how it arrived at that conclusion so you can verify its accuracy for yourself. Which is one thing if you're just writing an article or something, but if you ask it how to wire something up it might give you instructions that result in electrocution. It's drivers following their GPS into lakes all over again.
Appare

$20 a month?

[Locke2005]

Can I get that included in my college fees?

In college, $20/mo for written papers was a deal!

[atrimtab]

It was bad enough with the swapping of papers from their files and storing of exams among some Greek chapters across the country. Exams for professors were kept on file and papers were swapped to a campus that had not seen them.

I hope it is really obvious when papers are written by these services. Otherwise, every essay and paper must be written in class from now on. Followed by the elimination of Wi-Fi and cell service in classrooms and collection of phones and laptops.

How else can these required skills be effectively graded?

Re:

[Waffle Iron]

How else can these required skills be effectively graded?

Maybe in the future, the only required skill will be knowing how to ask a chatbot for the appropriate answers.

Re: In college, $20/mo for written papers was a de

[functor0]

I wonder if they can incorporate some sort of "language steganography" into large replies. Then they can offer this detection as a service as ChatGPT+ for institutions.

5 bucks

[CalgaryD]

I was just thinking yesterday that I could probably justify 5 bucks a month for access to it. But 20 is just too much, for the amount of use I see I can get from it. But really, I think that it can replace some significant number of office or support workers right now, how it is now...

Re:

[ranton]

If ChatGPT even saves me 15 minutes each month writing some email / powerpoint / newsletter content for work, it is well worth $20.

Re:

[CalgaryD]

I write my emails myself, I would still need to explain to the Chat what has to be done. I do not think that it can do powerpoint yet. Newsletters... probably, but I am grateful that I do not have to do that. So, really, not yet much use, for me, at least. Another thing, I would probably feel wrong if I copy and paste its response to anything that is expected of me... anything it does at this point is somehow throw away vanilla type of responses, typically I do not have to do stuff like this.

"Chat"

[Dan East]

Does $20 a month also allow you to disable the fake "chat" responses that trickle through instead of just dumping the entire text response at once?

Re:"Chat"

[martinheriksen]

You are seeing how ChatGPT works in real-time, It writes one word, and the determines what the next word should be. It is not like us humans where we have determine what we answer in full, and then build a holistic concept of how will structure that answer in text. ChatGPT sees only one word at a time. Essentially, it is a feature not a bug kind of deal.

Re:

[ljw1004]

You are seeing how ChatGPT works in real-time, It writes one word, and the determines what the next word should be.

But is it honestly that slow? I see ChatGPT spend fifteen seconds composing its answer one word at a time. I know that's how it works under the hood, but my expectation is that it calculated the answer one word at a time in closer to 50ms, and the visual slowness is just eye-candy intended to convey an effect.

$20/mo. for ethical access

[systemd-anonymousd]

Welcome to OpenAI, where they claim that coincidentally the most ethical way to protect you from AI is the path that causes them to make multiple billion dollar deals with Microsoft for Github Copilot, charging for DALL-E 2, and now this.

Their mission statement was explicitly to ensure that everyday people could have AI without having to depend on massive corporations for access. OpenAI used to release their models and weights up until they started seeing a path to making money with GPT-2.

They were founded

Penmanship courses

[sapgau]

As more colleges and universities are expected to require written papers I see an exploding demand for penmanship courses!

No thanks

[Stomper_Stoddard]

No thanks, I can write my own Python code.

But.. contextual/strategic competence?

[heretic108]

Does this paid option come with contextual awareness and strategic competence? Or is that too much to ask right now? Will ChatGPT stay forever crippled in its own version of the Scunthorpe Problem [wikipedia.org]?

curious to see

[Big Hairy Gorilla]

who ponies up 20/month.
People are now trained to expect everything digital is free.

But there are also those who see themselves as early adopters, same people who bought both Alexa AND Google speakers ... so .... surely someone will bite.