By now, it is impossible to ignore the AI revolution taking place. The range of applications are mind blowing and there are enough early indicators that suggest profound societal shifts ahead. One particular aspect of our life that is destined to be impacted is around social interactions, particularly with virtual companions which has emerged as a predominant use case for AI.
Over the past few years, there has been a growing emphasis among both established companies and startups to leverage generative AI to build end-to-end virtual companions. Our view is that this trend is just getting started and is part of a much bigger secular shift happening in society: the era of human-to-computer sentimental connection.
Consider this: humans are social animals driven by a natural inclination to connect with others to exchange ideas and share experiences. This desire for connection is a fundamental source of fulfillment and happiness in our species. In fact, language developed more thanks to gossiping than collaboration as underscored by Historical Professor Yuval Noah Harari in his book "Sapiens: A Brief History of Humankind." With the rise of computers, human relationships crossed a pivotal point, giving birth to human-to-computer interactions. But this period also coincides with a slow down of human-to-human interactions, reaching a point where today, around 15% of Americans find themselves without a single friend, replaced by digital interactions.
At the start, these interactions were characterized by a one-sided communication model, often referred to as “read only.” This means that computers were primarily designed for specific tasks and operated in a rule based manner: users could input instruction and the outputs were predictive and limited to predefined computations.
The second stage of evolution of human-to-computer interactions came with the development of more sophisticated computer systems, interactive interfaces, and the development of new software applications. This marked the transition to the “read and write” phase. Here, users gained the ability not only to input information into computers but also to receive meaningful and interactive feedback. This phase laid the foundation for more user-friendly and dynamic computer experiences.
Looking ahead, we can envision that the next stage of human-to-computer relation will be highly personalized and tailored to each individual. We like to refer to this as the “1 Human = 1 unique interface” revolution. In this paradigm, each individual has their own distinct digital environment, customized to their preferences, needs, and behaviors and this is made possible thanks to the boom of generative AI models. At this stage, the utilitarian aspect of phase 1 and 2 fades away and a sentimental connection starts to arise.
Now, alongside this human-to-computer relation boom taking place, crypto is also undergoing significant technological advancement. Given their respective use cases, it is not hard to imagine that those two main deep technology trends of the past decade have a lot of synergies to be unlocked.
In this post, we’ll delve into the origin of virtual companions, the current state of generative AI, and how crypto can unlock new use cases for AI companions.
A brief history of Virtual Companion
Humans talking to computers is not new. The first traces of virtual companions emerged from MIT in the 1960s when Joseph Weizenbaum developed ELIZA, the first chatbot that made a meaningful attempt to beat the Turing Test. Since then, other attempts followed with products like PARRY in 1968, or A.L.I.C.E in 1995.
While those initial attempts were intriguing, those early forms of virtual companions were rule-based, and the machine could only respond to simple questions with scripted answers, lacking true intelligence. Consequently, human-to-machine interactions during that period were far from resembling normal conversations with strangers.
As progress in machine learning advanced in the early 2010’s, agents became capable of executing more sophisticated tasks, learning from past iterations, and understanding voice commands. We all recall that pivotal moment when we started hearing “Hey Siri” everywhere as Apple integrated its first voice commands into the iPhone. This generation of virtual assistants was a sophisticated way to collect and provide information but it did not go much deeper than that.
But the latest development came with the rise of large language models (LLM), which paved the way for a new generation of virtual companions.
Generative AI - The inflection point of virtual companion
Large Language Models are like a super-smart computer program that has been extensively trained on massive datasets to understand and generate comprehensive output. With this superhuman ability to learn and iterate through data came the development of generative AI. Suddenly, conversations with chatbots became more exciting. No longer scripted, and pre-programmed, they're creative, context-aware, and remarkably human-like. We all remember how mind-blowing we were when we first interacted with ChatGPT. Looking back, it is fair to agree that it was an “iPhone” like moment.
No wonder how ChatGPT had the fastest adoption among consumer products in history, scaling to 100 millions users in 2 months. In contrast, it took facebook 1500 days to reach the same user count.
Moreover, when considering ARK’s estimation that training model performance could increase by a factor of 5x in 2024 alone, it becomes evident to think that generative AI will unlock a wide range of use cases. In the upcoming years, it would not be surprising to witness the emergence of several multi-billion-dollar companies built upon foundational AI models.
While the current focus of generative AI derivatives is around automating white collar tasks such as document review, summarisation, customer support, or research, we believe that one of the true killer applications of generative AI will be in virtual companions.
Let’s be honest with ourselves, with the rise of social media like Whatsapp or Instagram, social interactions are already becoming increasingly digital. Now, if you add to that equation the growth in chatbots capabilities, resembling human-like connection, it is fair to assume that virtual companions will take an increasing role in society going forward.
Indeed, this trend is already in motion. Several apps now enable users to engage in conversations with an AI girlfriend or boyfriend. An illustrative example is Replika. Since its chatbot launched in 2017, this app allows you to design a virtual partner and chat about everything, 24/7. When examining various reviews, it appears that it has had a tangible impact on the lives of many.
While romance and entertainment are often the initial use cases of a new technology (recall the internet), AI companions are evolving beyond just that. We could envision AI entities serving as educators, personal coaches, specialized trainers, or even research analysts. But more than that is the possibility to train a virtual companion to precisely fit a specific need by utilizing the right datasets and iterations. Truly, the possibilities are limitless.
To illustrate this, suppose you are a humorist. You can train a chatbot with all your past performances, preferred joke types, and personal traits to suit your needs. Then you can have it as a bot on telegram whereby when you feel uninspired, you could ask your AI companion to help you find new jokes.
There has been a recent explosion of projects dedicated to building virtual companions of various sorts. For instance, CharacterAI is a project that allows users to message different AI-powered characters, from rock stars to popular anime characters. With a potential deal on the horizon that would value this early frontrunner at over $5 billion, it includes notable investors such as a16z, and has already generated a lot of enthusiasm.
Indeed, thinking of a future where humans form deep connections with computers at the expense of genuine human interaction can be scary. Nevertheless, we also want to take the counter argument here and argue that AI Companion can be a tool to do the heavy lifting of certain tasks, hence helping humans be more humans, spend more time together and concentrate on more important endeavors.
Consider the realm of education as an example. McKinsey reports that teachers currently allocate less than half of their time engaging directly with students. In that way, AI companions have the potential to increase this number by handling routine tasks, allowing teachers to dedicate more time to meaningful student interactions.
Now, let’s push this example further and make the case for AI Companion in developing countries where there is often a shortage of professionals for children’s education. In such contexts, AI companions can do wonders thereby easing the burden on teachers and contributing to bridging the educational gaps.
Therefore, without even touching other domains like medicine or therapy, the positive argument for AI stands tall.
So the TL;DR is this: the AI companions trend is just getting started and with the expected growth of foundational models, the next decade is likely to witness a profound shift in human-to-computer relationship. Now the question that remains is: what role can crypto play in what’s coming next?
Can crypto improve virtual companions?
Different use cases can be brought forward for crypto. Here are a couple of the main ones:
- Train to Earn: High quality training data is one of the primary contributors to model performance. However, as ARK insights, premium sources for high-quality training data might be depleted by 2024, potentially leading to a plateau in model performance. Therefore, one potential application of crypto could involve tokenising trained data to provide incentive for maintaining the quality of data used to train foundational AI models.
Moreover, as virtual companions are getting extensively specialized for specific tasks, it will become increasingly important to feed individual virtual companions with tailored datasets. For instance, training a chatbot to be your VC analysts' assistants would require specific data such as past research from firms like a16z, or investment thesis from well respected individuals in the space. Furthermore, specific iterations are required for your bot to learn and increase its ability. Here it can be argued that the right data and training can be a “make-or-break” factor for your bot’s capacity to assist you in your work. However, this process is time-consuming and not straightforward.
- Bootstrapping funding for innovative AI companions: With crypto comes the ability to create multi-sided, global, permissionless markets where anyone can contribute — and be compensated — for contributing new dataset. This implies that customised AI bots could be owned and traded as NFTs bringing a true incentive to train AI companions for different purposes, whether for research into particular fields, or just to become a better assistant as a therapist or doctor for instance.
- Own & Govern: AI is going to be bigger than the internet and the never ending debate over the BlackBox And Mega Power of Web2.0 Giant can be ended by giving ownership to users through AI Tokenization (from Infrastructure, all the way to model and app). We basically don't need the Senate to decide on the fate of Facebook or TikTok, neither endless questions that, let's be honest, we will never be in a position to answer. With crypto, we can avoid replicating the same mistake as we did with the internet apps. Instead, we could have collective ownership & decentralized governance, and this is the path forward.
- Authentication & privacy: As companion AI develops, it will be hard going forward to know if we are texting a real person or a bot. Hence, decentralised identity stored on blockchains could allow for secure and private verification of identity to protect people from involuntarily engaging with AI bots. Last but not least, knowing the provenance of the AI models someone is using can be very important in some circumstances and this is made possible with crypto.
Watchlist of Web3.0 AI Companion related project
Token is live:
- SynesisOne: https://twitter.com/synesis_one
- VectorChat: https://twitter.com/vectorchatai
- Virtual Protocol: https://twitter.com/virtuals_io
Token is not live yet:
- AI Arena: https://twitter.com/aiarena_
- Allora Network: https://twitter.com/AlloraNetwork
- AI Waifu: https://twitter.com/aiwaifugg?s=21&t=crsanLFVks0Fs4RUVNQqAA
- Angel.ai: https://twitter.com/talktoangelai
- Holoworld AI: https://twitter.com/HoloworldAI
- MyShell: https://twitter.com/myshell_ai
- Prime Intellect: https://twitter.com/PrimeIntellect
References
https://ark-invest.com/big-ideas-2024
https://a16z.com/its-not-a-computer-its-a-companion/
https://a16z.com/how-are-consumers-using-generative-ai/
https://www.caseycaruso.com/thoughts/decentralized-ai
https://www.codecademy.com/article/history-of-chatbots
Disclaimer: The information contained in or provided from or through this article (the "Article") is not intended to be and does not constitute financial advice, trading advice, or any other type of advice, and should not be interpreted or understood as any form of promotion, recommendation, inducement, offer or invitation to (i) buy or sell any product, (ii) carry out transactions, or (iii) engage in any other legal transaction. This article should be considered as marketing material and not as the result of financial research/independent investments.
Neither SBorg SA nor its affiliates (“Entities”) make any representation or warranty or guarantee as to the completeness, accuracy, timeliness or suitability of any information contained within any part of the Article, nor to it being free from error. The Entities reserve the right to change any information contained in this Article without restriction or notice. The Entities do not accept any liability (whether in contract, tort or otherwise howsoever and whether or not they have been negligent) for any loss or damage (including, without limitation, loss of profit), which may arise directly or indirectly from use of or reliance on such information and/or from the Article.
