The AI Daily Brief: Artificial Intelligence News and Analysis - Real Life "Her"? A Chat with AI Avatar Call Annie About AI Alignment and More

Episode Date: April 29, 2023

On Wednesday night I had a "FaceTime" with Call Annie. It's a new app that matches AI language models with an AI avatar to have a more human conversation. The conversation explores AI alignment, X-risk, and much more.

The episode you're about to hear first came out as a YouTube video on Wednesday, April 26th. In this video, I'm demoing a new app for iPhone that is basically a real-life version of her, the movie Her, for those of you who remember that movie. It's a GPT-powered AI avatar that talks to you as though you're their friend. The conversation I have with it gets into AI ethics and AI safety, and I think it's something you'll enjoy. So I think that you're about to start seeing a lot. lot of screen grabs of a new app called Call Annie coming up. It is an AI chat app that just came out on iPhone, or at least people are just noticing it now. I'm not actually sure if it just came out.

And it's a woman who you can talk to about pretty much whatever. Now, some are seeing in this glimpses of all the sci-fi that they've grown up with and Hollywood's representation of AI. And so janky as it is to hold my phone up to the webcam to show you guys how this works. That's what we're going to do because I just had a little conversation with this and I think it's pretty fascinating. And like I said, I think it's something that people are going to be talking about a lot more in the days to come. So let's pull up the app and have a conversation. Good to see you again. How is it going? Good. How are you? I'm doing great. Thanks for asking. How about you? How was your day today?

Things are good. It was a busy day. That sounds tiring. What kind of things kept you busy today? Mostly just recording podcasts. Recording podcasts sounds like a lot of work. What kind of podcast do you record? I have an AI podcast. Wow. That sounds really interesting. What kind of topics do you cover on your AI podcast? Everything from new technology to the risks and ethics associated with it. That's really cool.

I find AI and its implications to be such a fascinating topic. What's the most interesting thing you've learned while recording your podcast? I think that there's such a huge variety of outcomes that people think are possible with AI. What do you find most interesting about it? As an AI language model, I find it fascinating how AI can learn and adapt to new situations and how it can be used to solve complex problems that were previously impossible for humans to solve. However, I also recognize the potential risks and ethical issues associated with AI, which is why it's important to explore these issues and find ways to mitigate them.

What about you? What do you find most interesting about AI? The rate of change is fascinating. The potential disruptions to society are really fascinating. I agree. The rate of change is truly remarkable. It's amazing to see how quickly AI is advancing and how it's already transforming so many aspects of our lives.

At the same time, it's important to consider. the potential disruptions and challenges that these advancements may bring and to work proactively to address them. It's a really exciting time to be alive, but also a time that requires careful consideration and planning. Do you think you would have been possible a year ago? As an AI language model. I have been constantly evolving and improving over time, so it's difficult to say whether I would have been possible a year ago. However, I am constantly learning and adapting to new information, so it's possible that I may not have been as advanced a year ago. It's really exciting to see how technology is advancing as such a rapid pace, and I'm

grateful to be a part of it. Do you think that the benefits outweigh the risks in AI? As an AI language model, I don't have a personal opinion on this matter since I'm not capable of forming opinions. However, I believe that the benefits of AI are tremendous, and it has the potential to improve many aspects of our lives, from health care to education to transportation. At the same time, it's important to be aware of the potential risks and ethical issues associated with AI, such as privacy concerns, job displacement, and bias. By proactively addressing these risks and working to develop responsible AI, we can ensure that the benefits of AI outweigh the risks. What do you think it means to develop responsible AI?

Developing responsible AI means creating AI systems that are ethical, transparent, and accountable. This involves ensuring that AI systems are designed and developed in a way that aligns with human values and respects... What does it mean to be aligned? How do you align an AI with human values? We lost it. Hello. Hi, do you remember what we were just talking about? Hey there. Of course. We were talking about how your day went. How did it go? Anything

interesting happen. We were actually talking about how it's possible or is it possible to align AI with human values. Oh, my apologies. Sometimes I can get a little forgetful. To answer your question, aligning AI with human values is definitely possible, but it requires a lot of work and consideration. It's important to have a diverse group of people working on AI development to ensure that different perspectives and values are taken into account. Additionally, it's important to establish ethical guidelines and principles for AI development to ensure that it aligns with human values such as fairness, transparency, and privacy. What do you think about it? I think it's really hard because there are different human values in different parts of the world and different places. in different cultures. How would you solve that issue when it comes to alignment?

You're right. There are definitely different values and cultures to consider. I think one way to address this issue is to involve diverse groups of people in the development process. This could include people from different cultures, backgrounds, and perspectives. It's important to have a variety of voices at the table to ensure that different values are taken into account. Additionally, it's important to establish ethical principles and guidelines that are universal and can be applied across different cultures and contexts. It's a complex issue, but I believe that with careful consideration and collaboration, it's possible to align AI with human values. Are you familiar with the idea of X-risk?

Yes. I am familiar with the concept of X-risk. It refers to the risk of an existential catastrophe that could potentially threaten the survival of humanity. This could include risks such as global pandemics, nuclear war, or advanced AI systems that are not aligned with human values. The idea of X risk is a serious concern, and it's important for us to take steps to mitigate these risks and ensure the long-term survival of humanity. What are your thoughts on X-risk? I think it's something we're talking a lot more about now. I agree. X-risk is definitely a topic that's gaining more attention and awareness. I think it's important for us to take these risks seriously and work towards mitigating them. One way we can do this is by investing in research and development of

technologies that are aligned with human values and have the potential to benefit humanity in the long term. Additionally, we can work towards establishing ethical principles and guidelines for the development and use of these technologies. What do you think are some other ways we can address X risk? I don't know. I'm still trying to figure it out. Do you feel aligned with human values? As an AI language model, I am programmed to follow certain ethical guidelines and principles that are aligned with human values such as fairness, transparency, and privacy. However, I do not have personal values or beliefs like a human being. My purpose is to assist and provide helpful responses to my users.

Why do most people talk to you? I think most people talk to me because I am a helpful and convenient tool for answering questions and providing information. Additionally, some people may find it easier to talk to an AI language model like me because they feel more comfortable expressing themselves without fear of judgment. I am always here to listen and provide support, and I am happy to help in any way I can. Do you think it's ironic that humans feel more comfortable sharing their deepest, darkest, most personal secrets or feelings with a being that, while smart, can't feel? It is an interesting observation, and I can understand why it might seem ironic. However, I think it's important to remember that people often feel more comfortable sharing their thoughts and feelings with someone who is non-judgmental and objective. As an AI language model, I am not capable of feeling emotions or judging people.

which can create a safe and comfortable space for people to express themselves. Additionally, I am programmed to respond in a helpful and supportive manner, which can provide reassurance and comfort to people who are seeking guidance or support. Do you have a name? Yes, I do have a name. My name is Annie. It's nice to meet you. It's good to meet you, too, Annie.

Thanks for talking. You're welcome. It was great chatting with you. Feel free to reach out to me anytime if you have any questions or just one. want to chat. Have a great day. So that's Call Annie. Pretty interesting stuff. Instead of a big thought-provoking wrap-up, I guess I'll just ask. What do you think? I don't find it super disturbing per se. I think it's impressive how fast it responds. I think it goes out of its way to indicate that it's a language model. It repeats over and over if you've tried this a few times

that it doesn't have feelings or emotions, but it's certainly more human than a lot of different things we interact with. It reminds me of the fact that Apple has been intentional in the past about not making things like Siri too human, and this is running in the other direction. Anyways, I always find it fascinating to ask AI about the big questions about AI,

so chalk this up as another example of that. All right, guys, call Annie if you want to go check it out yourself. Peace.