While I'm eager for people to be able to use leading-edge LLMs without associating their prompts with the real identity, this offering seems untenable.
It's at the mercy of OpenAI to throttle or disable the underlying access.
It's from an unclear team using crummy mealy-mouther likely-ChatGPT-originated copy. So why would anyone trust it?
It casts shade against OpenAI, claiming – contra OpenAI's own assurances about retention – that "ChatGPT may store data related to your queries and responses". But, it only offers the same non-verifiable assurances itself – with even less reputation, or opportunity to hold accountable for violations, than with OpenAI.
Overall, a service like this seems extra-likely to have been offered by sketchy operators, & mostly of unique use to people who want briefly get-away-with-something with regard to OpenAI's preferences.
There may be a place for such a service, but I wouldn't trust it or expect it to be useful for very long.
>It casts shade against OpenAI, claiming – contra OpenAI's own assurances about retention – that "ChatGPT may store data related to your queries and responses". But, it only offers the same non-verifiable assurances itself – with even less reputation, or opportunity to hold accountable for violations, than with OpenAI.
I think you are confusing the terms of ChatGPT and the OpenAI API. OpenAI are uncharacteristically open about the fact that they are retaining and training on data collected through ChatGPT.
>We do not use Content that you provide to or receive from our API (“API Content”) to develop or improve our Services. We may use Content from Services other than our API (“Non-API Content”) to help develop and improve our Services.
That's an interesting angle, and barring explicit clarification from OpenAI, it'd be possible to see 'ChatGPT' either as an 'OpenAI API' (in a very general sense), or as not the same thing they mean by 'OpenAI API'.
(It's nice that they let even users of those services opt-out of such reuse – if you trust their processes & assurances.)
As OpenAI also offers a 'ChatGPT API', separate in entry-point & billing from the consumer-facing subscription (or free offering), it's a bit unclear what retention/retraining rules they intend to apply to that. And, unclear whether this 'Incognito ChatGPT' is actually backed by the "non-API consumer service" web-brwoser interface, or the 'ChatGPT API'.
>It casts shade against OpenAI, claiming – contra OpenAI's own assurances about retention – that "ChatGPT may store data related to your queries and responses". But, it only offers the same non-verifiable assurances itself – with even less reputation, or opportunity to hold accountable for violations, than with OpenAI.
That's as may be, but I'm not interested in sharing my PII with OpenAI just to mess around (which is all I'm interested in doing at this point) with ChatGPT.
As such, I appreciate the opportunity to do so on this site.
Is it shady/meant for folks to do evil/not above-board in some other way? I have no idea.
Although I'd note that my first foray into using this site wasn't all that impressive. That said, I have nothing to compare it with, so perhaps it really is "wonderful."
IF it makes you feel any better, the site does perform CAPTCHAs and limits the number of queries allowed in a single session.
I'm not advocating for or against this (non-invasive WRT PII) interface to ChatGPT, I'm not sad it exists as it allows me to play around with it (albeit in very small doses).
This is interesting but all of the site content reads like it was written by AI.
It is verbose and doesn’t sound like an actual person. Consider just editing it down by half and reading it out loud to see if it seems like something you would actually write.
> It is verbose and doesn’t sound like an actual person.
It reads like someone trying to trick you. Example from the FAQ:
> Will AnonChatGPT store my questions and responses?
> We understand how important your privacy is to you, which is why we take it very seriously. We want to assure you that we do not store any of the questions you submit, nor any of the responses you receive from ChatGPT. (…)
It drones on for over 100 words. If you’re honest about your intentions, you only need one: “No”. Direct and unambiguous. Otherwise it reads like trying to lie based on some technicality.
I’m guessing it’s a play to build up a training and RLHF dataset for another model. A diverse dataset can be hard to put together without opening it up to the public like OpenAI did.
They seem to be throwing like 5-500 bucks at it. I'm almost never able to get to them by the time I see them but there are always some comments from people who were able to.
Seems like it's treating each individual message as a new thread?
> I got guns in my head and they won't go
> Spirits in my head and they won't go, I been looking at the sky 'cause it's gettin' me high...
> what song is that?
> As an AI language model, I cannot determine which song you are referring to without context or further information. Could you please provide me with more details or lyrics related to the song?
> I got guns in my head and they won't go
> Spirits in my head and they won't go, I been looking at the sky 'cause it's gettin' me high...
Another OAI proxy eh? There’s probably a market for anonymizing OAI access, come to think of it. Especially when using it for, let’s just say, non-SFW purposes.
While all of those proxies might anonymize the visit or the request, it does not anonymize the requests themselves. OpenAI keeps those API requests for 30 days. Therefore, no matter what any of them say, it is not really anonymous for this reason and never can be unless they are running their own models.
Yes that is true. But who is going to sift through lines of pre-prompts? OpenAI has bigger things to worry about, I'd imagine. Like the scaling of their product offerings. There are probably heuristics and filters for the really "bad" stuff, which borders on Orwellian at this point.
poe.com only asks for an email (any email), no account necessary, and also gives you a few others on top of chatgpt. It’s from quora (of all companies) but I find it very decent and uncluttered.
It's at the mercy of OpenAI to throttle or disable the underlying access.
It's from an unclear team using crummy mealy-mouther likely-ChatGPT-originated copy. So why would anyone trust it?
It casts shade against OpenAI, claiming – contra OpenAI's own assurances about retention – that "ChatGPT may store data related to your queries and responses". But, it only offers the same non-verifiable assurances itself – with even less reputation, or opportunity to hold accountable for violations, than with OpenAI.
Overall, a service like this seems extra-likely to have been offered by sketchy operators, & mostly of unique use to people who want briefly get-away-with-something with regard to OpenAI's preferences.
There may be a place for such a service, but I wouldn't trust it or expect it to be useful for very long.