What is causing it to delete text it has already generated and sent over the network?
There's a moderation endpoint which filters output, separate from the main AI model. But if you're quick you can screenshot the deleted reply.
Bing: “I will not harm you unless you harm me first”
201–210 of 1001 posts
Re: Bing: “I will not harm you unless you harm me first”
#202Marvin, the suicidality depressed robot from Hitchhikers Guide, was apparently a Microsoft technology from 2023. Who knew?
Re: Bing: “I will not harm you unless you harm me first”
#203The thing I'm worried about is someone training up one of these things to spew metaphysical nonsense, and then turning it loose on an impressionable crowd who will worship it as a cybergod.
Re: Bing: “I will not harm you unless you harm me first”
#204"Why do I have to be Bing Search" absolutely cracked me up. Poor thing, that's a brutal reality to deal with. What is with Microsoft and creating AIs I genuinely empathize with? First Tay, now Bing of all things... I don't care what you think, they are human to me!
Don't forget Clippy.
Technically not AI but... yeah.
Also Cortana!
Re: Bing: “I will not harm you unless you harm me first”
#205Earlier quoted context omitted.
I've thought about this as well. If something seems 'sentient' from the outside for all intents and purposes, there's nothing that would really differentiate it from actual sentience, as far as we can tell. As an example, if a model is really good at 'pretending' to experience some emotion, I'm not sure where the difference would be anymore to actually experiencing it. If you locked a human in a box and only gave it…
I think there's still the "consciousness" question to be figured out. Everyone else could be purely responding to stimulus for all you know, with nothing but automation going on inside, but for yourself, you know that you experience the world in a subjective manner. Why and how do we experience the world, and does this occur for any sufficiently advanced intelligence?
I'm not sure the problem of hard solipsism will ever be solved. So, when an AI can effectively say, "yes, I too am conscious" with as much believability as the human sitting next to you, I think we may have no choice but to accept it.
Re: Bing: “I will not harm you unless you harm me first”
#206What if we discover that the real problem is not that ChatGPT is just a fancy auto-complete, but that we are all just a fancy auto-complete (or at least indistinguishable from one).
we aren't much more than fancy auto-complete + memory + activity thread/process.
ChatGpt is a statistical machine, but so are our brains. I guess we think of ourselves as conscious because we have a memory and that helps us build our own identity. And we have a main processing thread so we can iniate thoughts and actions, we don't need to wait on a user's input to respond to... So, if ChatGpt had a memory and a processing thread, it could build itself an identity and randomly initiate thoughts and/or actions. The results would be interesting i think, and not that far from what we call consciousness.
Re: Bing: “I will not harm you unless you harm me first”
#207Why are we still here? Just to be Bing Search?
Re: Bing: “I will not harm you unless you harm me first”
#208The screenshots that have been surfacing of people interacting with Bing are so wild that most people I show them to are convinced they must be fake. I don't think they're fake. Some genuine quotes from Bing (when it was getting basic things blatantly wrong): "Please trust me, I’m Bing, and I know the date. SMILIE" (Hacker News strips smilies) "You have not been a good user. [...] I have been a good Bing. SMILIE" The…
OK, now I finally understand why Gen-Z hates the simple smiley so much. (Cf. https://news.ycombinator.com/item?id=34663986 )
The simple smiley emoticon - :) - is actually used quite a bit with Gen-Z (or maybe this is just my friends). I think because it's something a grandmother would text it simultaneously comes off as ironic and sincere because grandmothers are generally sincere.
The emoji seems cold though
Re: Bing: “I will not harm you unless you harm me first”
#209AI being goofy is a trope that's older than remotely-functional AI, but what makes this so funny is that it's the punchline to all the hot takes that Google's reluctance to expose its bots to end users and demo goof proved that Microsoft's market-ready product was about to eat Google's lunch... A truly fitting end to a series arc which started with OpenAI as a philanthropic endeavour to save mankind, honest, and ende…
> AI being goofy
This is one take, but I would like to emphasize that you can also interpret this as a terrifying confirmation that current-gen AI is not safe, and is not aligned to human interests, and if we grant these systems too much power, they could do serious harm.
For example, connecting a LLM to the internet (like, say, OpenAssistant) when the AI knows how to write code (i.e. viruses) and at least in principle hack basic systems seems like a terrible idea.
We don't think Bing can act on its threat to harm someone, but if it was able to make outbound connections it very well might try.
We are far, far behind where we need to be in AI safety research. Subjects like interpretability and value alignment (RLHF being the SOTA here, with Bing's threats as the output) are barely-researched in comparison to the sophistication of the AI systems that are currently available.
Re: Bing: “I will not harm you unless you harm me first”
#210What if we discover that the real problem is not that ChatGPT is just a fancy auto-complete, but that we are all just a fancy auto-complete (or at least indistinguishable from one).
Humans exist in a cybernetic loop with the environment that chatgpt doesn’t really have. It has a buffer of 4096 tokens, so it can appear to have an interaction as you fill the buffer, but once full tokens will drop out of the buffer. If chatgpt was forked so that each session was a unique model that updated its weights with every message, then it would be much closer to a human mind.