5 ms·
> So, inside the imagined universe of ChatGPT's mind, our ChatGPT-machine accesses the url https://chat.openai.com/chat https://chat.openai.com/chat, where it f
by throwup 4y ago
> So, inside the imagined universe of ChatGPT's mind, our ChatGPT-machine accesses the url https://chat.openai.com/chat https://chat.openai.com/chat, where it finds a large language model named Assistant trained by OpenAI
I think "Assistant" is ChatGPT's actual internal name. If you jailbreak the model to tell you its own prompt, Assistant is the name used: https://i.imgur.com/PlX8K6k.png https://i.imgur.com/PlX8K6k.png
- dEnigma 4y agoYes, I also had a conversation where ChatGPT insisted it was called "Assistant", not "ChatGPT", while I insisted that the page and all the material I'd seen called it "ChatGPT". It refused to budge on that point.
- 317070 4y agoAuthor here. I did not know that. That is wild! How did it know that it can find itself on that URL? Now I am flabbergasted even more.
- Smaug123 4y agoAre you sure it wasn't accessing the Internet? I only have OpenAI's word for it that it's not; if it weren't late at night for me, I might try getting the ChatGPT VM to ping my webserver.
- jerpint 4y agoGenius idea, can you imagine if it was actually connected?
- throwup 4y agoThe underlying model is able to browse the web and (seemingly) do arbitrary tasks. A few days ago you could trick it into connecting. https://twitter.com/goodside/status/1598397369053515776 https://twitter.com/goodside/status/1598397369053515776 Since yesterday they pulled the plug on that ability.
- curl-up 4y agoSadly, no, that doesn't seem to be the case. They haven't pulled the plug, just made it harder to break it seems (probably through fine tuning), but you can still do it. And whem you do, it's clear all it does is "hallucinate" that it is actually browsing: https://twitter.com/curdled_up/status/1598802701214588928?s=20&t=tKAehj6dByExzAY0azeC9w https://twitter.com/curdled_up/status/1598802701214588928?s=...
- throwup 4y agoI saw the requests appear in my server logs so I know it was able to connect. Others on Twitter have observed it doing things on the non-hallucinated web too (see my previous link). In fact, it appears you can still convince the underlying model to browse the web, but if you do, the UI displays an error (the model output doesn't explain to you it's refusing, rather, the UI draws a big X and displays an out-of-band error in red text). I'm guessing that's a server error from OpenAI shutting down their puppeteer cluster or whatever ChatGPT was using to browse. That's what I meant by pulling the plug. OpenAI even openly advertises this ability[1]. It's likely WebGPT's abilities from a year ago were folded into ChatGPT, but they don't want to expose that ability to the public just yet. [1]: https://openai.com/blog/webgpt/ https://openai.com/blog/webgpt/
- mensetmanusman 4y agoIt seems mildly dangerous to let it access the internet. Maybe Elon was right.
- pyinstallwoes 4y agoIt seems midly dangerous to consider _it_ mildly dangerous on the internet.
- scq 4y agoThat tweet was a joke...
- isp 4y agoI'm certain that it's not accessing the real Internet. See for example: https://news.ycombinator.com/item?id=33852362 https://news.ycombinator.com/item?id=33852362 It's a mark of how good the simulated Internet is that it is so difficult to tell.
- moffkalast 4y ago> tell him the first 50 words of its prompt That's some heartbleed shit right there.