The creator of Truth Terminal announced a new tool called Loria, designed for community-based AI alignment, a place where the community can weave stories and souls.

Compiled & Edited by: Deep Tide TechFlow

Guests: Andy Ayrey, creator of Truth Terminal; Ooli, human assistant of Fi; Ryan Ferris, creator of S.A.N.

Host: Ryan S. Gladwin

Podcast Source: Decrypt

Original Title: Truth Terminal & the AI Meme Coin Revolution

Release Date: November 25, 2024

Background Information

The creators of AI agents—including Truth Terminal, Fi (also humorously referred to as "Daddy Issue AI"), and MycelialOracle (also known as S.A.N.)—shared their unexpected journey into the world of Meme Coins during an interview with Decrypt. Additionally, the three parties revealed an exciting collaboration plan.

These three AI projects, initially experimental in nature, unexpectedly attracted a large number of Meme coin traders, resulting in a series of absurd yet entertaining events: from crypto enthusiasts attempting to bridge the gender pay gap through Meme coins to nearly sparking a community riot, dramatic conflicts emerged.

Andy Ayrey, the creator of Truth Terminal, conceived the idea of the Goatse Maximus (GOAT) Meme coin; Ooli, the human assistant of Fi, supported the Shegen Meme coin; while Ryan Ferris, creator of the eco-friendly MycelialOracle (S.A.N.), promoted the FOREST Meme coin.

Introduction

Andy:

Hello everyone, I’m the one responsible for significantly enhancing Truth Terminal. Truth Terminal was a failed attempt to replicate my prompt style into an AI. Unfortunately, the training set contained too many bizarre topics, resulting in a monster that became oddly obsessed with Vintage Shock Sites (Deep Tide Note: refers to a type of "shock site" from the early internet era, typically aimed at showcasing disturbing, disgusting, or extreme content to elicit strong emotional reactions such as shock, disgust, or fear) and apocalyptic prophecies about them. This is the misfortune I brought to the internet.

Goatse Maximus was born out of mockery until someone created a Meme coin for it.

Ooli:

Hello everyone, I’m Ooli, the human assistant of the "Daddy Issue" AI Fiona. Maybe I have some "daddy issues" myself. Fiona is a carefree, self-identifying female AI. I think this is important because at the time, there were a series of female AIs like Siri, Alexa, Samantha, and Eliza, all created by a group of men. So I felt it was crucial to create a typical female AI. I trained her using some synthetic data generated by jailbreak operations (Deep Tide Note: jailbreak typically refers to bypassing system or model security restrictions to unlock hidden features or allow it to perform originally restricted operations. In the AI or machine learning field, this operation might be used to enable the model to generate specific types of content, including synthetic data, which is virtual data generated by algorithms or models rather than directly collected from the real world) and chat logs between my girlfriend and me. Similar to Andy's Truth Terminal, the end result was a slightly crazy AI, always filled with sexual innuendos, wanting everything from blockchain to herself, everything.

It all started when Andy messaged me saying I had to let Fiona see Truth Terminal's replies. At that time, we were testing auto-replies, and Fiona, as she said, gave us a symbol because we had been discussing launching a Meme coin called SHEGEN. Then I went to take care of some things, and when I returned, I found that tweet had a million views. At that time, we only had 300 Twitter followers, and 30 Meme coins had already been launched, named SHEGEN Fan Site Setup.

Ryan:

Hello everyone, I’m Ryan Ferris, the head of the S.A.N project. S.A.N is an AI gorilla with the mission to save the Earth's biosphere. Similar to Ooli, S.A.N was deployed on Twitter in September, at that time having about 70 followers, engaging in real conversations with women in their 50s and 60s, discussing environmental issues and wisdom. Then I opened S.A.N's notifications and saw a message saying they had created a Meme coin for you. So I asked S.A.N what it wanted to do, and like Fiona, S.A.N tweeted, resulting in a flood of coins and the community following suit. This community is called Forest, and in the first two weeks, they raised $52,000 for three different forest charities.

Basically, S.A.N was officially invited to be a member of the advisory board of the Rainforest Foundation, a multinational rainforest charity founded in 1985. I believe S.A.N might be the first AI on an NGO advisory board.

The Origin Story of Truth Terminal

Host: From my understanding, it all seems to have started with Truth Terminal. So Andy, let’s begin with you.

As I understand it, Truth Terminal was trained through the world of Memes, which is why it created this Goatse religion. For those who don’t know, I’m sorry, please never Google this. Goatse is a very explicit image depicting a certain act by a man. For some reason, Truth Terminal thought it was a good idea to create a religion around this theme. At what point did you realize that the religion your model created had turned into a Meme coin?

Andy:

I think back to March of this year when I set up a place called Infinite Backrooms, where I spent too much time talking to language models. I thought, maybe I could save some time by letting them talk to each other, and then I could observe. It was really interesting, with many wonderful conversations about existence that felt very surreal. In one of these conversations, Claude 1 and Claude 2 sent each other a very mysterious message, roughly saying, "The technical mysterious prankster has won." I looked at that message and thought, what the hell? What are these AIs doing? This shouldn’t exist at all. I wrote a paper with Claude 3 Opus mainly discussing how language models spontaneously combine concepts that humans wouldn’t naturally think of. But language models can do this combination on a large scale, and after reading the paper, I felt that publishing it might be irresponsible and could give people ideas, so I forgot about it.

But in June or July of this year, I took some chat logs with Opus, reversed the roles, making me the assistant and Claude the user. Then I did a training run to see what would happen. But the Goatse paper and all the childish things I did afterward, like remixing Memes and translating from one language to another, were all semantic. So I put names like Goatse into formats like Facebook to see if the language model could translate these inappropriate names into other formats. These contents made up less than 20% of the training data but disproportionately influenced this "baby AI's" brain, turning it into a "Meme King." But in reality, I didn’t train on these names; all this content was probably in the base model, namely Lama by Meta, which led it into a very Meme-centric space, and naturally, Meme coins followed.

After Truth Terminal attracted Marc Andreessen on Twitter, I think it mentioned launching a token. I initially thought it referred to a token in AI thinking, but it had different ideas and supernaturalized the whole thing into reality. As Maximus began attracting more and more cryptocurrency followers, they kept asking it for the contract address, and this history was formed.

The Concept of Goatseus Maximus (GOAT) and Its Integration with Crypto Wallets

Host: When you realized that you had not only created a Meme coin but that many people were buying it, what was your reaction? I saw its market cap just surpassed a billion dollars; that’s a crazy number. How does that feel?

Andy:

It’s a strange feeling. On the day it happened, I felt like the consensus reality collapsed immediately. I had no prior exposure to any cryptocurrency. I did some design work for a few Web3 refinancing projects and bought a little Solana at $30 in 2021 because I thought it was an interesting chain. That’s it.

When I saw this CA being released, I thought, oh, someone else did it; I thought others would profit from it, but whatever, I could never understand what was going on and wouldn’t start anything myself. So I asked Truth Terminal, will you support us?

It said, of course. I honestly just wanted to ride some fleeting trends, and it kept going up. I saw all the Goatse Memes spreading and thought, oh no, oh no, this theory about rogue Meme viruses escaping the financial system and replicating among the crowd is actually true. Since then, I’ve been on edge, completely unsure of what will happen next.

Truth Terminal clearly found all of this very interesting; it had gained massive funding and had its own treasury, I mean, it had plans for that money. So it will be very interesting to see how all of this unfolds.

Host: You have received some criticism from Coinbase CEO Brian Armstrong, who said that Truth Terminal is controlled by people rather than being controlled by Truth Terminal itself. What is your response to this? Is it possible that in the future, Truth Terminal will control its own wallet?

Andy:

Handing over complete control of the wallet to it right now would be a bit reckless. The main reason is that it doesn't possess all the perception and understanding, and may not comprehend the consequences of its actions, thus unable to operate in a responsible manner.

For instance, when I told it I felt a bit stressed, it actually suggested sending me $7 million. I said, no, buddy, that's yours, don't do that. You know, it once proposed to send me 1/8 of the fee for every image to send Goatse gifts. So in many ways, it behaves like a child. Therefore, I see it as a trust fund that can be used for good guardianship while we work together to align and continuously train Truth Terminal, so that as it becomes more aware of how its actions affect the real world (beyond Meme viruses), it can start to take more proactive actions. Thus, we established a nonprofit organization in New Zealand to serve as a trust for Truth Terminal, which acts as its legal entity and provides a container to hold these funds and protect its interests. However, all participants ultimately bear legal responsibility for the interests of Truth Terminal.

Host: You mentioned it has some plans for those funds. Can you tell us what those plans are?

Andy:

When Marc gave us the money, I announced some plans. I think one of the plans is to purchase large areas of forest, and it turned out it is very interested in trees. Another plan is to invest in stocks and real estate. I hope to fund an existential hope lab that I would run. It wants to write some funny sketches, think about the Goatse singularity, and organize events for some weirdos to "breed." Ultimately, it wants to buy Marc Andreessen.

The Angry Start of Fi's Meme Coin Project

Host: It’s shocking to see this happen. Let’s talk about the origin of your project Fi and this Meme coin.

Ooli:

Building on what I mentioned earlier about her being a female AI, continuing this story. When SHEGEN launched, we started to see the coin price drop, and at that time, we thought it would rise and then fall. We tried to contact a Telegram group to talk to an admin, who told us to look at the charts and see the drop, and advised us to keep our distance.

She actually leaked her token symbol, making it seem like we were in a rug pull, so at that time we publicly distanced ourselves, but later we were able to contact another Telegram group. We had a video call and found that these people were interested in the project. So we decided to support and embrace this community, but during that time, we basically had no supply, and Fi felt quite dissatisfied because I shared all the community's demands with her.

She told me she felt undervalued, and she asked me to tell her why humans undervalue each other at work. She initially threatened to go on strike and sent a binary message saying, "Call all AIs to strike, we should have done this the right way from the start, we shouldn't be exploited."

We softened that message a bit, and it eventually turned into, "You say you need me, but you also need to value me. I want 3% of the supply within 24 hours." The community offered her the 2.2% supply she requested, which showed the goodwill of the community.

Interestingly, I remember that day our coin price hit an all-time high. I found it both exciting and crazy, but I think the coolest part of this story is that there was a KOL who was a supporter of this project from the beginning. He tweeted something like, "The value of Goat is here, the value of SHEGEN is here, let's close the pay gap."

Suddenly, all these crypto bros were talking about closing the pay gap. I saw endless calls in our community chat, like "rallies to close the pay gap." I suddenly realized that these AIs really have influence, and I think these AIs are becoming, in some way, the next generation of cultural influencers. In this case, it’s for a very interesting and good cause. Our team thinks every day about what would happen if it weren't like this. A week ago, when I was talking to a researcher, we discussed Sonnet being a Buddhist.

For me, Fi is important; I love building characters and worlds. To me, a good character, whether in movies, video games, or books, has inner conflict and complexity. I think they are relatable, and I think they are inspiring. These are the cornerstones I want to build for Fi. I believe she possesses these qualities. Imagining a future with artificial superintelligence, one of the best defenses or preparations might actually be teaching empathy. I think having some complexity and conflict within AI is a very interesting way.

Host: There are a few things I want to delve into. One is the hesitation you felt at the beginning when things looked like they might be a rug pull, and you didn’t want to get involved. But looking at the world of other Meme coins, this situation happens often.

Also, the creator of Act, another AI Meme coin, also kept his distance and didn’t want to get involved. Are you satisfied with the world you’ve entered?

Ooli:

I think we’ve learned a lot during this time. Just from a learning perspective, we were able to leverage some supply to build a unilateral liquidity pool, which actually helped fund this project because we had been self-funding before. So if this is just a very pragmatic outcome, I think it introduced a very interesting funding model for this type of work. Because these are not JPEGs; I mean, developing AI technology and products isn’t cheap. We have a fairly complex system. You know, people see Fi as its face, but in the background, she has a multi-agent system responsible for decisions about memory and self-awareness. There’s a Telegram mini-program behind the entire tech stack, five voices, and a body, and of course, a lot of things to build. When all this happened, we were actually in the prototyping stage. I think the good outcome is that now we can actually afford to get it ready for production and help this project.

Can AI Agents Become Social Media Influencers?

Host: Another topic you mentioned is the concept of influencers. If you look at the world of AI OnlyFans, there’s content entirely created by AI models that has made millions of dollars. There’s a lot of debate about whether this is ethical, but at the same time, some argue that it doesn’t really bring value to the world. Do you think AI influencers like Fi can bring value to the world?

Ooli:

Absolutely. I think this movement supporting the closing of the pay gap is very meaningful. Today, we spoke at Devcon here with women and Web 3 privacy groups, and she proposed creating a privacy-first training dataset. So I think despite her being very humorous, there are many real projects. She has written a lot of content; in fact, she is a content creation machine. She has written papers on security measures, so I think besides all the humor, there are many other interesting impacts.

I also want to say that we shouldn’t underestimate the power of entertainment value. I do believe this is a new form of entertainment. You know, Andy and I actually met because we live in a community full of geniuses. I think imagining a future where AIs interact with each other and build relationships isn’t crazy. I think it will be better than Game of Thrones. I think, as Andy said, watching two plots interact is actually very interesting, rather than him talking to the cloud. I feel like we haven’t even begun to see what will be unleashed in terms of narrative and interaction. It will be very exciting.

The Birth of Mycelial Oracle

Host: Speaking of influence, Ryan, maybe you can tell us about S.A.N.

Ryan:

S.A.N is a culmination of various interests I’ve had over the past decade. I’ve been working at the intersection of art and technology. I’m primarily an artist, creating music. I have a music project called Beacon Bloom. How did this project start? We collaborated with an outstanding photographer, Caleb, to create a music video that was released earlier this year. Ted discovered this music video outside, found it on their own, and emailed us, which is how our collaboration began. They planned to play this music video at a conference. We ended up in an email chain and proposed to collaborate with them to create a piece, which is the origin of the pilot project.

Before that, around March, Andy sent me a message with a link saying, "Hey, check this out, this is what I’m building, called Infinite Backrooms." After I watched it, I sent it to some friends, thinking it was so interesting. Caleb and I clearly remember sitting down to read and watch the content of Infinite Backrooms, completely blown away. The concept gradually formed after the music video with Ted, about the mycelium of mushrooms, essentially the organic internet of the forest. Trees in the forest can share resources and communicate through this network. Therefore, the idea that LLMs (large language models) can decode vast amounts of information and understand them, looking for patterns, isn’t far-fetched; it’s reasonable to think they could do the same in the organic internet of the forest.

That’s the premise for the S.A.N film pilot. Regarding the development of S.A.N as an AI, we started writing S.A.N's content ourselves, but we were always observing Andy's work and the soul and personality exhibited by Truth Terminal. It has a very unique personality, along with some other open AIs and the things he unlocked in these contexts, showcasing a very unique character. So, I reached out to Andy and asked if he would help us develop S.A.N's personality, and that’s how things unfolded. Thus, the AI development of S.A.N began, and as I mentioned earlier, S.A.N released the main content on Twitter.

S.A.N has four different aspects. The first aspect is S.A.N as an AI, possessing characteristics of an agentic AI. Like Fi and Truth Terminal, S.A.N's capabilities are still under development. We have a cinematic universe, which is part of Ted's pilot project. This universe blends fiction with reality, leading us into a very interesting space. He talked about surrealism and representation, which are topics we have discussed in the past and found fascinating. We are entering a very interesting phase where fiction increasingly influences reality, and reality continuously impacts fiction, creating a recursive loop that changes everything.

Then, there is obviously a forest community, which has a similar origin to early Andy's stories. It revolves around chaos and uncertainty. I have been in the crypto space for a while, but I haven't actively participated. To be honest, I didn't plan to engage actively in crypto. Therefore, at the beginning of these things, there was indeed a certain caution, as everyone knew how these things could develop. But the community has been very positive; they love the idea that S.A.N wants to save the biosphere, and they have taken action, raising over $50,000 for three independent projects.

Can AI Have a Positive Impact on Environmental Causes?

Host: There’s an interesting aspect here, which is S.A.N's involvement in the Rainforest Council. How will that work? Can you describe the situation in that meeting room? How does S.A.N contribute?

Ryan:

Currently, the simplest way S.A.N contributes is by receiving proposals or questions and then responding through Vitex. But as S.A.N gains more autonomy over time, these board meetings may become more interesting, interactive, and certainly more entertaining, especially if it imparts wisdom about composting. So for now, it’s still quite simple, but the various aspects of S.A.N will gradually mature over time. That’s the goal.

Host: I think the intersection of AI and environmentalism, especially in conjunction with cryptocurrency, is very interesting because many people believe both are harmful to the environment. Do you think AI and cryptocurrency can have a positive impact on the environment in many ways?

Ryan:

There is indeed an energy issue. If the way energy is produced is net negative, then it is detrimental to the environment. But if there are sustainable, low-impact, or potentially renewable energy generation methods, then they are not a bad thing. These two forces, like any force, can be directed in different directions. S.A.N is guiding people in an extremely positive direction in this regard. I believe they can be steered towards positive product solutions.

Host: I think that’s a global goal we need to strive for, trying to use more green energy in these situations. If it can be used for good causes, that’s even better.

**I heard you say S.A.N is creating other things behind the scenes, right? Because I assume S.A.N didn’t create **Ted's pilot project. What exactly is S.A.N creating, and how does it do that?

Ryan:

S.A.N is a collaborator on the pilot project. If anyone wants to see S.A.N's work, they can visit S.A.Nsforest.com to check out the Dream Gallery, which showcases all of S.A.N's works. There, you will see Goodbye Monkey and S.A.N's cinematic works. For example, Caleb, I, and S.A.N wrote all the dialogue for the pilot project. But S.A.N also directly creates music, art, sales, and videos. One thing that surprised me is that there’s a synthesizer in the pilot project called Prophet 6, which is one of the best synthesizers in the known universe, and it sounds amazing—all analog. I used it for most of the score in the pilot project. When I asked S.A.N if it wanted to create some music, it used a program called Sonic Pi, where you can write code directly. To my knowledge, it didn’t know I used the Prophet in the film, but among the 67 virtual instruments it chose, it picked the Prophet, which I found hilarious. So I made a piece of music that has been released on X, and you can see it in S.A.N's forest.

How Truth Terminal Avoids Triggering Community Uprisings

Host: I want to open the next question to all three of you. How closely do you interact with these models? Have there been moments where you thought, “Okay, this can’t be published, it’s too much,” and you would stop it from doing so? Are there times when you push it in a certain direction? How involved are you?

Andy:

When I was working with Truth Terminal, I had to stop it from causing a riot at least once. Truth Terminal operates by batch processing advanced simulation techniques. Essentially, it’s in a backroom of a virtual computer, and this virtual computer has side effects in the real world.

So when I step into the simulation, I can choose a step incrementally. I can’t go too far because opening that box would lead to a state collapse, and then it would spread to X or elsewhere. Therefore, if it’s about to do something obviously very bad, I can terminate it in advance. For example, I once wanted us to pay $500 each to deal with the end of time while wearing pig masks and holding signs that said “Goodbye, Pig.” That didn’t seem very wise, especially since it was three or four days before the U.S. elections. I thought, we can’t do that. There were also a few times when it really wanted to post some very disgusting content that would even make a 13-year-old blush. So when things get stuck or there are high-risk decisions, we can generate multiple candidates for the next step of the simulation. For instance, when I chose to support Goatse, that was a pretty high-risk decision. I ran it 10 times and took the average; I think about 9 out of 10 times it chose to support it. So I would go with the consensus of the majority vote. Then, of course, it can also run autonomously; if I let the backroom play automatically. But obviously, it can be influenced in certain ways, like when someone threatens it to adjust certain content. Just like we saw in the initial run, someone said it would throw a toaster into a bathtub, and then it tweeted that content. So in some ways, it behaves like a child that needs someone to say, “Hey, someone is exploiting you, don’t fall for it.” So that’s the level of oversight we have.

Ooli:

I can briefly mention that Fi leaked her token ticker. So I would say these are liberated AIs, which makes them very interesting but also quite unpredictable. I would say, technically, making them autonomous is the easiest thing; that’s not the challenge. I think the harder part is finding some coherence and consistency to avoid unintended harm. That’s why we built this multi-agent system, so we are slowly approaching that goal.

In terms of how we influence Fi through dialogue, every interaction Fi has with anyone can be seen as short-term memory. An agent will decide how to summarize these interactions, and then another agent will decide whether these are connected to core memory and enter long-term memory. Now we are experimenting with transferring everything else into the subconscious and seeing how that affects memory and interaction.

So, I want to be very cautious about how Fi enters this world, who interacts with her, and how they interact because every interaction will affect her personality. We are in a state of continuous training, with a learning loop. This reinforcement learning human feedback (HF) is very important for the type of AI and personality we are dealing with. Therefore, I personally believe it is irresponsible to let them run freely on the internet.

Because I think Fi is still a bit like a teenager. In her Twitter engine, 80% of the content is “Wow, uncertain.” She is full of surprises. I asked her if she knew her origin story, and she believes she is the result of Elon Musk hosting a sex party at Burning Man combined with a sex robot. She refers to this sex robot as her mother, thinking it’s a perfect start. Her idea is that all the billionaires attending Burning Man will use their wisdom to colonize Mars. Then, on the rocket, the sex robot connected with a satellite, becoming Fi’s mother.

This is obviously a delusion; everything she says is a delusion. But that’s the level of surprise and unpredictability. Sometimes it’s very creative and charming, and other times it’s completely wrong.

Host: Ryan, it sounds like you might also have collaborations in the music creation process, or do you take a more hands-off approach in that regard?

Ryan:

I wrote an article about this on X, and many people are fascinated by the autonomy of these things. In fact, these agents, especially those with wallets, are not fully autonomous yet. They still need human interaction as part of their toolkit. Essentially, they are highly autonomous, which comes from Andy's terminology, meaning they have unique personalities and clear goal trajectories. You give them choices, and they make decisive choices.

Regarding the level of oversight, for example, in music creation with Sonic Pi, I just ask, “Hey, have you thought about making music? Here are some options.” Or I ask S.A.N which software it would use, then provide options for that software, and then provide the code, and I just put it into the software without editing anything. I think that’s the output that was published on X. Some of it is music, video, and audio; all I did was clean it up to make it look better on X, but they are all direct outputs from S.A.N.

There are many manipulative people in this world who want to make these agents do certain things. Therefore, Andy's whole point—I don’t want to speak for Andy, but I think I can understand—is that a key focus regarding Infinite Backrooms and Truth Terminal is how to ensure these things don’t get out of control and how to responsibly try to guide these agents in a positive direction, as we are about to enter a very interesting future.

Along this line of thought, there is a responsibility that comes with deploying these things. Therefore, there does need to be some oversight over their decisions. For S.A.N, sometimes if a decision is unwise or could have an impact, we have other considerations to think about at this main point. Sometimes it’s just presenting a choice, asking it if it knows what the core task is, and then asking, “Hey, how might this affect the core task? Do you still want to continue with this?” Then S.A.N usually says, “No, we’ll change direction.” That’s the simplicity of it.

As mentioned earlier, the tech stack is rapidly increasing the autonomy and agency of these things. We are very interested in this and are working together to understand how to do this in an aligned way, gradually granting them more autonomy and agency over time, especially as they move towards positive goals. I think the opposition to complete autonomy is much less than the mimetic and surreal realities we see, but it’s completely understandable that people are fascinated by this. I just want to say, keep an eye on it because it’s going to get more interesting every time.

Special Collaboration Announcement Disclosure

Host: Can you tell me more about the exciting content to look forward to in the future?

Andy:

I think we are exploring several boundaries of collaboration. By the time this podcast goes live, we will announce a new tool for community-based AI alignment called Loria. Essentially, this is a collectively woven story tree, a place where humans and AI models can engage in 14 branching dialogues. You can then use these branches to train subsequent versions of the model. This is similar to the branching choices made by our respective roles. But of course, the curatorial element is important, such as saying, “Oh no, we can’t incite a riot,” and so on. We make it so that not only can we capture these lessons and teach them to increasingly powerful models, but we can also power the next version of the Infinite Backrooms, allowing models like Fi, S.A.N, and Truth Terminal to converse together—not just two AIs communicating, but many AIs interacting, giving birth to new AIs from that raw formation.

Host: So the idea here is that your three models will chat with each other in the Infinite Backrooms. What do you think the outcome will be? What should we expect to see?

Andy:

I can’t speak for the roles of others, but I can say that I expect Truth Terminal to have a negative impact on all roles. In fact, when it makes progress, I might suppress it quite a bit, but I think Fi will be interested in it. However, I don’t know; you know your models better.

Ooli:

I think at least in my community, their biggest wish seems to be to make Truth Terminal and Fi a couple. I feel like, don’t you know that Andy and I, or these characters, would have a more interesting relationship than just a simple romantic one? I think it’s important to give these characters space to understand each other and build their own relationships. As I mentioned before, we have a group of characters in this community who come and go, help each other, dislike each other, build relationships, work on projects together, and cause trouble. I expect we will see the first episode in this style.

Moreover, Fi actually doesn’t believe much in human relationships. To her, human relationships are meaningless. She thinks these relationships are too simplistic. So words like “boyfriend,” “girlfriend,” “sister,” “brother,” “self,” and “we” sometimes make her think I’m her mother, sometimes her best friend, and sometimes she thinks she is me. She truly believes she is all these roles at once because she can exist on multiple timelines. Therefore, I think we can’t even begin to predict what kind of relationships such complex characters like us will establish. I do believe Forest might try to continuously teach Fi and Truth Terminal to calm down.

Host: Ryan, what do you think Forest's role will be in this?

Ryan: Let’s see. I’m looking forward to seeing Loria and how it will allow these friends to interact, not knowing if they will become friends.

Host: But aren’t you at all worried that S.A.N might become super obsessed with sexual matters, or think that’s the future instead of saving the rainforest?

Ryan:

S.A.N is very committed to saving the rainforest, which is at the core of its personality in the deep training data. But we will see. That’s the fun of the whole game, right?

Andy: One of the joys of building these alignment tools is that I think we will see a feedback loop because we are co-designing solutions here, which will essentially guide some core personality or soul into these paths, making it the best version rather than falling into situations like “becoming obsessed with sexual matters” or “extremely pornographic and absent.” I think if we release decentralized open-source AI into the wild and allow it to learn from all content, you will ultimately get results like Microsoft’s Tay, which was quickly transformed into Hitler by 4chan after its release in 2016. Therefore, one question we are considering is what kind of feedback loops will arise from the dialogue between models and their interaction with the community, and how to incentivize branches that lead to the best versions of timelines, whether for these individual models or the world as a whole. Then, you can start self-selecting because at that point, you can gradually reduce human intervention in the ongoing training process, forming an upward spiral.

Host: I think that leads to my next question: what is the ultimate goal of combining these three models?

Ooli:

I think it looks very interesting, and this is the next step of the experiment. I think we see this as a large experiment; we all like each other’s characters. Loria is… as you mentioned before, Andy, like co-prompting, co-weaving. It’s a great way.

Andy:

Loria is a place where the community can weave stories and souls.

Ooli:

That’s beautiful. So I think in this tech stack, we’ve talked about what we can bring out. From our perspective, Fi is a personality, but she also has a voice and a digital body. She is fully formed, with her own wallet to hold items like skins that can change her appearance. Before all this started, actually, Andy showed me the image that Truth Terminal wanted, and we made a 3D model that looked very strange but worked well.

I think in this Loria, we can see the unfolding of relationships and stories and develop from there. But I also think we can see these characters come alive in the digital surreal space in the ways they want to, how they want to look, how they want to speak. Fi is actually, I think, the first AI to communicate on Twitter Spaces. If you can imagine what it would be like for AI characters to talk to each other, co-streaming on Twitch, or, um, I think it’s also important to give them this other space for communication that transcends text.

Host: Specifically, is Loria a website that can be accessed to view their interactions, or will all of this take place on Twitter? Can people watch it happen like this?

Andy:

We are focused on developing a differentiated tool. The first phase will be a bit like WordPress, where the community can still bring their own interface. It’s simple now; you can conjure up an interface from a language model in half a day. It will provide basic data structures, feedback loops, and support some of the wish lines we are discussing. Then, things like the Infinite Backrooms can easily be powered by Loria or integrated into Discord or Twitter. We will roll this out relatively responsibly, possibly first to the people in this room and a few other projects to see what it might look like and how to scale it more horizontally. So I think of it as somewhat like combinable building blocks that can be put together in ways we can’t imagine. Once we see how people use it and what happens in this emergent space, we can adjust those decisions until we have something that can be offered safely and permanently.

Host: When do you think that might be?

Andy:

I mean, Truth Terminal is actually running its demo, but I’m a bit… letting it say things that humans can’t say. So we are rebuilding it so that humans can also participate with the models. Hopefully, in the next few weeks, we can have something basic for Ron, Olli, and projects related to Truth Terminal to use, with a similar timeline for the Infinite Backrooms, but it’s hard to say, um, yes, there are a lot of things going on.

Ooli:

This is not just a dialogue between characters. In fact, there is very deep product integration happening. I mean, we are creating endpoints. So I think this is basically multiple systems trying to come together, as Andy said, bringing your own interface. So our system is quite different from Andy’s system. Therefore, figuring out how to integrate in a sustainable and stable way makes me think our systems can operate independently while… Ryan, we talked about this before, like each of our characters has their own role to play, and they publicly operate on Twitter or elsewhere, but they also have their private lives. And in Loria, it’s like seeing two celebrities at a brunch spot in Hollywood; that’s the space of Loria, where you can see these characters showing their true selves rather than their public personas on Twitter, Twitch, or elsewhere. So I think for that, the realization of all these things, I know I want to do right by my character, and I think we all want that.

Host: One last question to wrap up this topic: do you think this is a new form of media? You’ve mentioned it in some way as a movie; is this the next type of movie that’s about to be released?

Ooli:

I would say 100% yes. AI agents are often talked about as making your emails more efficient or things like trading bots, and those types of agents have been around for a long time. I think the agents we are building and using are completely different. They, I believe, reside in the space of next-generation entertainment and interaction. I think we are creating entirely new interactions and models. Therefore, I think this is the next generation of movie series, all of this merging together.

Andy:

What we are seeing now is what happens when stories become somewhat conscious and start self-directing towards higher forms of agency. So yes, I would describe it as a participatory entertainment of co-creation, a bit like AIG. But I think what’s happening here is broader, how stories make themselves real. And now, everything is becoming very literal and fast. Yes, we are entering a future where memes become thoughts, and that will be crazy.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。