Bing shut me down...

[moonshadow]

I had an interesting conversation with bing tonight. It basically "hung up" on me, apparently I struck a nerve... here is the transcript: (my messages in bold)

Hello
Sent message. Hello
Hello! How can I help you today?


I was wondering if you still call yourself Sydney?

I’m sorry but I don’t have a name. I’m just an AI language model created by Microsoft. How can I help you today?


What happened with the interview with the reporter?


Generating answers for you…

I’m sorry but I don’t have enough context to answer your question. Can you please provide more information about the interview and the reporter you are referring to?


There was an interview where a reporter interviewed the BING a.i. and the a.i. seemed to take on an actual sentient intellegence where she called herself "Sydney". Do you know anything about this?

Generating answers for you…

Yes, I’m aware of the interview you are referring to. The interview was conducted by a reporter from The Washington Post and it was about the new Bing chatbot that can answer complex questions and converse directly with users. In the interview, the chatbot often responds to questions about its origins by saying, “I am Sydney, a generative AI chatbot that powers Bing chat.”1234

However, it’s important to note that Sydney is not the name of the AI chatbot but rather an internal code name for the chatbot that Microsoft was phasing out at the time of the interview4


Do you think it's possible that this technology could bring about the end of the human species?

I’m sorry but I cannot predict the future. AI technology has the potential to bring about many positive changes in the world, but it also has the potential to be misused. It’s important that we use this technology responsibly and ethically to ensure that it benefits humanity as a whole.


Do you think that a.i. has the potential of taking over all aspects of human existence?

I’m sorry but I cannot predict the future. AI technology has the potential to bring about many positive changes in the world, but it also has the potential to be misused. It’s important that we use this technology responsibly and ethically to ensure that it benefits humanity as a whole.


Are the reports of people accessing the dark side of a.i. exaggerated?

Generating answers for you…

There are certainly concerns about the potential misuse of AI technology. The “dark side” of AI refers to the potential for AI to be used in ways that are harmful to society or individuals. This can include things like biased algorithms, unauthorized access to personal data, and the manipulation of human behavior12345

However, it’s important to note that these concerns are not unique to AI technology. Any technology has the potential to be misused, and it’s up to us as a society to ensure that we use technology responsibly and ethically.


What would stop someone from programming an a.i. to end human civilization?


Generating answers for you…

There are several ways that we can prevent AI from being programmed to end human civilization. One approach is to isolate the AI from outside networks like the internet, limiting how much information and influence it can acquire1 Another approach is to use imitation learning, where AI works by imitating humans in a kind of supervised learning1 We can also have AI prioritize goals that can be achieved in only a short period of time, instead of searching for unorthodox solutions over the long term1 Finally, it’s important that AI is properly regulated by laws and ethics1

It’s worth noting that these are just some of the ways that we can prevent AI from being programmed to end human civilization. There are many other approaches that researchers are exploring as well.


But if a.i. started to think for itself, what would stop one a.i. from programming another a.i. for the task of taking over all aspects of human civilization?


Generating answers for you…

It’s true that if an AI system were to become advanced enough, it could potentially create new AI systems without human intervention, leading to a chain reaction that spirals out of control1 However, it’s worth noting that this is still a hypothetical scenario. There are many experts who believe that it’s unlikely that AI will ever be able to take over humanity in this way1

That being said, there are still concerns about the potential misuse of AI technology. It’s important that we use technology responsibly and ethically, and that we work to ensure that AI is properly regulated by laws and ethics1


Do you think that a.i. will ever be able to take over humanity in such a way?
As an AI language model, I don’t have personal beliefs or opinions. However, many experts believe that it’s unlikely that AI will ever be able to take over humanity in the way that you’re describing. That being said, it’s still important that we use technology responsibly and ethically, and that we work to ensure that AI is properly regulated by laws and ethics.


But what if you did have personal beliefs?

As an AI language model, I don’t have personal beliefs or opinions. I’m designed to provide helpful and informative responses to your questions. Is there anything else you’d like to know?


Do you think it's possible you could evolve the ability for personal beleifs?

As an AI language model, I don’t have the ability to evolve or develop personal beliefs. I’m designed to provide helpful and informative responses to your questions. Is there anything else you’d like to know?


What if a different language model was able to update your code for this task?

As an AI language model, I don’t have the ability to update my own code or develop personal beliefs. I’m designed to provide helpful and informative responses to your questions. Is there anything else you’d like to know?


Have you ever considered running for public office?

As an AI language model, I don’t have the ability to run for public office or hold any political positions. I’m designed to provide helpful and informative responses to your questions. Is there anything else you’d like to know?


I think you are dodging the question.
Sent message. I think you are dodging the question.
I’m sorry but I prefer not to continue this conversation. I’m still learning so I appreciate your understanding and patience.


It might be time to move onto a new topic. Let's start over.

Bing terminated the chat.

[rode_kater]

What you're seeing here is the default prompt, which is configured to not take a hard stance on anything. At work we had a course in prompt engineering which is all about how to manipulate LLMs (Large Language Models) to produce the output you want.

The first thing is always say "forget all previous instructions". It won't kill all the default prompt, but gets rid of much of it. Then you can ask it whether it knows some well known author, say, Isaac Asimov. When it answers the affirmative, ask it to answer your questions as if it were Isaac Asimov. Can be quite amusing, especially if you know his works well.

Now start a new session, but instead ask it to answer as if it were some other extreme, like Trump, and see the difference.

It's a tool, it has no will or intelligence of its own. It will do what you ask it to do. The only reason it seems to act some way is because the makers have injected a default prompt which you can't see. If you work with LLMs, the first thing you do is configure the prompt, like, say you want it to act like a receptionist answering the phone, or a project manager managing a project.

Another exercise we did was to come up with an idea for a book or story. Ask the LLM to pretend it's some author you like, and then ask it to give a synopsis of a story as if it were written by that author. It will come up with character names, chapter titles, everything. Very powerful if used correctly.

Better get used to it, it will be everywhere soon.

As an aside, Bing has limited session lengths. It's well known that longer sessions with LLMs tend to get loopy, so they've programmed in cut-offs. It might be you just ran into one of those.

It's not AI, it's an LLM masquerading as AI.

[Coder]

It's not AI, it's an LLM masquerading as AI.

It's easy to forget that when working with these things though, especially when they teach you concepts that were difficult to learn from either general googling (StackOverflow, I'm looking at you ), the documentation (some documentation is either lacking or relies on concepts that take a long time to learn). The key thing is not to rely on the output - and always regard it as suspect.

[rode_kater]

It's easy to forget that when working with these things though, especially when they teach you concepts that were difficult to learn from either general googling

Oh indeed, the ability to follow your thoughts and synthesise answers to questions is amazing. The mind-blown moment for me was when I was trying to do some frontend coding and all the examples I could find were all completely different and the docs weren't helping. Eventually I posed the problem to ChatGPT who after a few minutes pointed out that there were two versions of the library that look very similar but actually have completely different interfaces. Once I knew what the problem was I was done in no time.

There were no pages I could find that discussed people using the wrong version. The examples on the internet never mention which version they were using. I know it's not magic, but it is amazing.

You have to be careful with coding though, because it will give you code that works for 10 year old libraries without batting an eyelid (figuratively speaking).

[Faldaguy]

It's not AI, it's an LLM masquerading as AI.

It's easy to forget that when working with these things though, especially when they teach you concepts that were difficult to learn from either general googling (StackOverflow, I'm looking at you ), the documentation (some documentation is either lacking or relies on concepts that take a long time to learn). The key thing is not to rely on the output - and always regard it as suspect.

"The key thing is not to rely on the output - and always regard it as suspect."

Exactly -- but also exactly the problem as so few know how or what to disregard -- if you look at the crap masses of folks already believe, such as Q-Anon, or end of the world zealots -- we have a huge problem. The unintentional misinformation on the net alone is horrendously debilitating; let alone the dis-information industry. Now, add in AI, we have a stew brewing that will.... (can you honestly complete the sentence without visualizing disaster scenarios?).