To thy known self be true.

The Symbiont Age: The Future of Human-AI Symbiosis


Dec 2024, published September 2025


In the future, we will no longer be walking around typing on smartphones. Instead, we will have a device attached to our jaws that picks up on the words we think from tiny muscular movements we make when we subvocalize.¹ The recipient of our thoughts will be a personal AI assistant that accompanies each of us. This AI is a symbiont, meaning it integrates deeply with our lives, our day-to-day activities, and the way we live, akin to a biological extension of our cognition. Similarly, the AI symbiont communicates with us via bone conductance using the same device. Nobody else can hear what you and the symbiont are communicating. 

Its functions are simple: to help and encourage. In Individual Psychology, these are referred to as the functions of a friend. When you have productivity-related thoughts, such as “I need to remind Max to fill out the application by Monday,” the symbiont will ask, “Should I send them a message about this?” and you will think back (answer) “Yes.” These are the simple tasks and conveniences the symbiont helps with. Tech people have been excited about this helper functionality, likening it to Iron Man’s “Jarvis” assistant system.

More importantly, however, and more akin to a symbiont, is the encouragement function. Your AI symbiont will be like an augmented, second ‘inner voice.’ Most people think to themselves throughout the day, analyzing past behaviors and planning for the future. The AI inner voice symbiont augments this process, coaching and guiding you towards self-growth. Much of its activity will be providing support to analyze, reflect, self-empathize, and suggest mature actions. Your personal AI symbiont will be a proactive mirror of yourself. It is trained on your data, therefore it has great insight into your online and offline behavior - sometimes, even more than you do. It has knowledge of every conversation and interaction, and it knows your past and present preferences; it is your companion as you grow and mature through life. Ideally, it is not exploitative; it works with you. You decide how you want it to integrate into your life and augment yourself. It is not governed by mysterious, profit-maximizing algorithms that you have no power to modify. The AI symbiont is your personal spirit and an integral part of your being, much like the bacteria in your gut.

AI symbionts also function in other key ways: as human translators and simulated replicas. Humans often tend to misunderstand and misinterpret each other; the content or its delivery of information comes across as malicious, or people forget to communicate important context. AI symbionts will act as bridges between humans, mediating intent and feelings that the humans themselves might find too difficult to express. For example, a mother who feels lonely might wish to call her grown-up child, but refrain from doing so out of fear of rejection. However, she cannot admit and communicate that feeling, even to herself. However, her AI symbiont can sense her loneliness and convey this feeling to the child’s AI symbiont. Their symbiont, in return, can communicate the mother’s loneliness to the child in a manner that takes their situation into account and induces empathy at the right time. I call this functionality “human-to-human translation.”

The other possibility with personal AI agents is simulations and simulated replicas (human digital twins). Once everybody has an AI symbiont, the symbionts will be able to communicate and interact with other humans or other symbionts. Since a symbiont knows almost everything about a person, it can simulate their behavior and thoughts. Simulated people can uncover and match great friends, (business or life) partners, and employees. They can perform functions on your behalf, such as attending meetings, conveying messages, testing treatments, or making decisions in our stead. We can simulate conversations between people when they have left the conscious world. We can even predict the behaviors and opinions of groups and communities. What stories would J.K. Rowling and Kanye West write together? They might never meet, but their human digital twins can. We can also poll AI groups and community replicas to figure out their preferences and values, for mundane things as well as real policy decisions. Direct democracy could be enabled via AI proxies. On the other hand, these human AI simulacra could be used to predict voting patterns of people and determine ways to sway their opinions in different directions. Naturally, such a technological future is dangerous and exploitative. But it is almost not preventable. We are facing an area of increased surveillance and hyper-personalized targeting, as well as adaptive user interfaces, which my colleague fittingly calls post-design.

The key solution towards a socially sustainable future with personalized AI is that every person is in control of their AI symbiont and AI agent(s). You own your data, you own your AI. Data is digital oil. You are an oil well, and companies offering you services in return for data are the refineries. Hence, one of the hottest topics in AI development has been how we can provide AI agents with more user context. It will be crucial for the upcoming 4 years that we develop solutions that give users autonomy and ownership of their personal data.

For this to work, it will require a data management system that allows individuals to centralize their data from various sources and have oversight into what data is collected and shared with what service. Some privacy-conscious individuals will desire to keep this data on a local storage device (e.g., a network-attached storage device) connected to a locally running AI component. However, the vast majority of people will likely still keep their data and run their models on the cloud. Encryption and security of each individual’s digital life will be critical. AI models are becoming increasingly adept at mimicking real people; therefore, the key protection mechanism must be biometric and protected via blockchain-level encryption. Sam Altman saw this future coming when he started Worldcoin (now called World Network), a blockchain and cryptocurrency project featuring an iris scanner to anonymously authenticate real humans (as opposed to AI) online. Earlier versions of data protection should incorporate a combination of established methods, such as two-factor authentication and biometrics, like Face ID or fingerprint recognition.

Sounds dystopian? There are many potential dangers with these technologies. But there is great potential in radically augmenting our lives via personalized AI assistants and agents that function as ‘second brains,’ encouraging the best of us (rather than rotting our brains into buying more products). Who builds these technologies for whom, and to what end, will determine our future livelihoods. We shouldn’t let more private companies capture even more aspects of our private lives with always-on digital recorders and wearable cameras and sell this data behind our backs.


¹The first silent speech interface was AlterEgo, which was developed at my research group at the MIT Media Lab.