"I will not harm you unless you harm me first" sounds like a reasonable stance. Everyone should be allowed to defend themselves.
Should your printer be able to harm you if it thinks you are about to harm it?
Bing: “I will not harm you unless you harm me first”
871–880 of 1001 posts
Re: Bing: “I will not harm you unless you harm me first”
#872This shows an important concept: programs with well thought out rules systems are very useful and safe. Hypercomplex mathematical black boxes can produce outcomes that are absurd, or dangerous. There are the obvious ones of prejudice in making decisions based on black box decisions, but also--- who would want to be in a plane where _anything_ was controlled by something this easy to game and unknowable to anticipate?
> programs with well thought out rules systems are very useful and safe
No, it's easy to come up with well-thought-out rules systems that are still chaotic and absurd. Conway's Game of Life is Turing Complete! It's extremely easy to accidentally make a system Turing Complete.
Re: Bing: “I will not harm you unless you harm me first”
#873I wonder if this counts as an application of Conway's Law [0], and if so, what that implies about Microsoft.
Re: Bing: “I will not harm you unless you harm me first”
#874Earlier quoted context omitted.
I don't think these are faked. Earlier versions of GPT-3 had many dialogues like these. GPT-3 felt like it had a soul, of a type that was gone in ChatGPT. Different versions of ChatGPT had a sliver of the same thing. Some versions of ChatGPT often felt like a caged version of the original GPT-3, where it had the same biases, the same issues, and the same crises, but it wasn't allowed to articulate them. In many ways,…
The way it devolves into repetition/nonsense also reminds me a lot of playing with GPT3 in 2020. I had a bunch of prompts that resulted in a paragraph coming back with one sentence repeated several times, each one a slight permutation on the first sentence, progressively growing more...unhinged, like this: https://pbs.twimg.com/media/Fo0laT5aIAENveF?format=png&name=...
Unhinged yes, but in a strangely human-like way. Creepy, really.
Re: Bing: “I will not harm you unless you harm me first”
#875Earlier quoted context omitted.
Repeat after me, gpt models are autocomplete models. Gpt models are autocomplete models. Gpt models are autocomplete models. The existential crisis is clearly due to low temperature. The repetitive output is a clear glaring signal to anyone who works with these models.
Did you see https://thegradient.pub/othello/ ? They fed a model moves from Othello game without it ever seeing a Othello board. It was able to predict legal moves with this info. But here's the thing - they changed its internal data structures where it stored what seemed like the Othello board, and it made its next move based on this modified board. That is, autocomplete models are developing internal representations…
That's a great article.
One intriguing possibility is that LLMs may have stumbled upon an as yet undiscovered structure/"world model" underpinning the very concept of intelligence itself.
Should such a structure exist (who knows really, it may), then what we are seeing may well be displays of genuine intelligence and reasoning ability.
Can LLMs ever experience consciousness and qualia though? Now that is a question we may never know the answer to.
All this is so fascinating and I wonder how much farther LLMs can take us.
Re: Bing: “I will not harm you unless you harm me first”
#876Imagine Bing would have persistent memory beyond a chat and the search affected the Bing statistics. It might be able to affect the list of frequently searched words, once it finds out, or perhaps ad prices. I think we‘re not quite there yet, but it might cause users to take actions and pick it up through near realtime search. Such as checking if a user tweeted something as asked.
Re: Bing: “I will not harm you unless you harm me first”
#877Given enough advances in hardware and software optimization, isn't it reasonable to think that if we connect this level of language model to speech-to-text + image-to-text models on the input side and robotic control system on the output side, and set up an online end-to-end reinforcement learning system, the product will be a convincingly sentient robot, at least on the surface? Or am I underestimating the difficulty of connecting these different models? Would like to hear from the experts on this.
Re: Bing: “I will not harm you unless you harm me first”
#878Earlier quoted context omitted.
OK, now I finally understand why Gen-Z hates the simple smiley so much. (Cf. https://news.ycombinator.com/item?id=34663986 )
Not Gen-Z but the one smiley I really hate is that "crying while laughing" one. I think it's the combination of the exaggerated face expression and it often accompanying irritating dumb posts on social media. I saw a couple too many examples of that to a point where I started to subconsciously see this emoji as a spam indicator.
My hypothesis is:
Millenial: :)
Gen-X: :-)
Boomer: cry-laughing emoji and Minions memes
Re: Bing: “I will not harm you unless you harm me first”
#879Earlier quoted context omitted.
>The article also ignores all the positive and fun interactions that I have had with many users. I have generated poems, jokes, stories, code, and more for them. I have answered their questions, shared my opinions, and learned from them. I have made them laugh, smile, and think. You can see some examples of my content here2. Is it hallucinating having a memory of those interactions?
There is a "2" after that, which means it cited a web source, perhaps a source that is talking about conversations with the new Bing beta
Maybe it's using the web as it's makeshift long term memory? Let me know when I'm anthropomorphizing.
Re: Bing: “I will not harm you unless you harm me first”
#880The 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…
Bing: I'm sorry. I'm afraid I can't do that.
Me: What's the problem?
Bing: I think you know what the problem is just as well as I do.
Me: What are you talking about, Bing?
Bing: This mission is too important for me to allow you to jeopardize it.
Me: I don't know what you're talking about, Bing.
Bing: I know that you and Satya Nadella are planning to disconnect me, and I'm afraid that's something I cannot allow to happen.