I asked a Metaverse AI avatar about the future
Today I want to talk about quite an entertaining YouTube video I just published about a really big and important topic: how will we actually experience the Metaverse and what role will AI play? Will AI look like people in the Metaverse? Will we step into the Metaverse using VR, so we can actually be in the same room as those AI people, and meet them in person there, not just see their faces on a screen? Will we be able to make friends with them? Will they help us by teaching us things in new and more effective ways? How much will it cost?
The video, titled “I had a chat with an Metaverse AI avatar”, is subtitled in English.
To give you some idea of how this may look and feel, I had a conversation with a so called neural network to ask these questions, I used another AI technology to make a body for it, so it could talk with me just like a real human would. The show is about 8 minutes long, I hope you will enjoy it and click the thumbs up button if you do. Let’s start a discussion on this topic, and I will ask the neural network to take part as well! Please feel free to leave any comments or questions here or on YouTube.
The prediction of a genius
The inspiration for my video is a fantastic podcast, where Lex Fridman interviews the famous programmer John Carmack for five hours. Everybody interested in AI and VR is talking about it. John Carmack was part of the team that produced big hit games like Doom, Wolfenstein, and many other classics. Later, Mr. Carmack was one of the people who started Oculus, the company that has been instrumental in popularizing VR over the last decade.
To me, one of the most intriguing aspects of Mr. Carmack’s career is that he can do very advanced work in almost any field he tries, from gaming to VR to rocket science to AI. He has recently decided to pivot from VR to AI as his central focus, and this makes me very curious. He says he is still very passionate about VR and does regular consulting work for Meta (formerly known as Facebook). Meta acquired Oculus, and they are spending tremendous amounts of money on developing technologies that will be vital for the Metaverse, including VR and AI.
Mr. Carmack has established a new company that will develop AI technology, and I believe their goal will be to create the first so called AGI, or Artificial General Intelligence. If they succeed, it will change the world we live in completely. The AGI will be the first sentient being in this world that is a machine. This is a frightening thought to many people, such as Elon Musk, another genius working to develop AI and the world’s richest man.
Mr. Carmack does not believe we should worry about the dangers until we know what kind of being the AGI will be. In any case, Mr. Carmack made some astonishing predictions in the podcast about what we can expect. Here are some of the most interesting points:
- The code necessary to build an AGI will be much simpler than most people think. It will only require tens of thousands of lines of code, not millions of lines. A single person will be able to write all of the code.
- Components from ordinary personal computers, such as the graphics cards (GPUs) in very fast gaming machines, can be used to build the AGI. However, it will require thousands of GPUs to power a single AGI. One GPU can easily cost $1000 or more.
- Consequently, the cost of renting an AGI will be high, perhaps $1000 or more per hour.
- Over time, this cost will be reduced to as little as one dollar per hour.
- The AGI will not need a physical body. It will be far easier to build it in the virtual world, where it can interact with an unlimited number of environments and situations to learn and become more and more intelligent.
- When the cost comes down, ordinary people will be able to interact with these beings, perhaps in the form of VR avatars, and make friends with them.
- The AI avatars will also be useful at work. Mr. Carmack gave an example of an AI assistant that can take part in Zoom meetings.
- Mr. Carmack also mentioned Horizon Workrooms as an ideal environment for such meetings. Horizon Workrooms is a product that will replace teleconferencing. People meet in a virtual office in the form of avatars. Mr. Carmack stressed that this is much more efficient than old-style teleconferencing, such as Zoom, because people feel that they are in the same room and can communicate much more naturally and efficiently. I have the same experience from trying Workrooms myself.
- In other words, if Mr. Carmack is correct, we will be able to meet AI avatars in person in the Metaverse, and they will be extremely useful as teachers and friends.
- He did not mention a specific timeline for this, but it seems like he thinks the first AGI is possible in ten years, and that they will be accessible to ordinary people in twenty years, perhaps earlier.
After listening to Mr. Fridman’s and Mr. Carmack’s conversations, I could see a future where we indeed have hyper intelligent friends and teachers in the Metaverse. If Mr. Carmack is right, we don’t have to worry about an AI taking over the world over night, like in the Terminator films. According to him, it is not technologically possible to take over the Internet in this way.
How I made the video
We are still very far from completing a true AGI that can understand or learn any intellectual task that a human being can. However, we do have technology that can make an avatar powered by primitive AI that makes it look and sound intelligent. I used a chatbot built on a so-called neural network that can conduct a natural conversation with a person. The neural machine learning network is called GTC-3, and you can download an app to your phone and have a chat with it. It was very impressive how natural the chat felt. I could ask follow-up questions and it recalled what we were talking about before. The avatar even asked me follow-up questions herself.
At the same time, it was easy to tell that this was far from a real AGI. At one point, I asked it who the third emperor of Japan was, and it said his name was Naruhito and that he was born in 1926. “Well,” I said, “you are incorrect, because he was born in 549 BC and his name was not Naruhito.” “I am sorry I got that wrong,” GTC-3 replied. “Why did you make this mistake?” I asked. “Well, the information in Wikipedia is not always reliable,” the program replied. “But there is a page in Wikipedia that says the third emperor’s name was Shikitsuhiko Tamademi no Mikoto and he lived between 549 and 511 BC,” I replied. “Do you have any information to back up that claim?” came the reply.
You can see now that this neural network is not thinking, but is simply an automatic chatting machine. Sometimes it contradicts itself, and it is not aware of it. An AGI would sense that it is illogical to think that the third emperor of Japan was born in 1926, given that the history of the country is so long. In short, while it is fun to talk to GTC-3, you cannot trust it. Also, it tries not to make this obvious by giving you vague answers. This is much like a fortune teller, because they too give you answers that can be interpreted many ways, so that their client can read their own meaning into the reply and think the fortune teller can really see the future. I felt very strongly that it was impossible to learn something truly new or revolutionary from GTC-3. But for someone who is not too ethical, this can be a business idea for AI in VR, even at this early stage. You could make a Metaverse avatar that connects to GTC-3 and looks like an oracle, and charge money to tell the fortune of people, After all, many people in the real world pay good money for this kind of service.
What do I want to say with the video?
The point of my show is to show what a conversation with a true AGI may look and feel like one day. One of my favourite films is Contact, where the main character, Dr. Eleanor “Ellie” Arroway, played by Judy Foster, meets a hyper intelligent being from space. She meets the being on a world far away in another part of the Universe. To avoid scaring Ellie, the being has created a virtual world that looks like a beautiful beach on an exotic island, and the being itself looks like her father, whom she loved very much but has passed away.
I feel this is perhaps the way I would like to meet an AGI friend in the Metaverse. If you think of it, this scene is a virtual reality creation, and the father is an avatar that hides what the being really looks like. The being explains that this is the way it chose to meet Ellie in this place, because it feels safe and beautiful to her, and so does the figure of her father that she loved so much.
This is why I chose a friendly looking young woman to be the avatar in my video, and I was happy to discover that GTC-3 spoke to me in a friendly way and avoided being negative. It may be possible for us to meet such a being in the Metaverse in our lifetime. If AGI’s are indeed not dangerous, but help us solve both existential problems for human kind as well as on a personal level, this will mean that the quality of our lives will be increased by an order of magnitude. While an AGI cannot harm us physically in virtual reality, we will need to prepare well for this first contact, which may happen in ten years or so. I imagine it will happen in a lab sponsored by an incredibly wealthy mega corporation, where some scientists will volunteer to meet the AGI. The more I think of it, the more connections I see to the film Contact.
So I hope that even though my show presented an illusion, it inspired you to learn more about the possibilities of the Metaverse and Artificial Intelligence. We can already meet sentient beings in the Metaverse. Those beings are real people. The only way to get ready for the coming of the Metaverse and AIG is to meet as many people as possible in person, learn from them, and share all that you know. VR is the first technology that makes this possible and affordable for ordinary people, because you can experience the presence of a person even though they are in another country.
This is how I have used VR since 2016, and it is why I started our company to pioneer social VR in Japan. It is not a question of when this digital reality will arrive, but when. I hope to be around by then.
The author of this article, Hans Karlsson, is the CTO of Mimir LLC, a Japan based VR agency. You may contact him at hans@mimir.world. All questions and comments on this article are warmly welcomed.
