The Role of Money in the Future Society

An opinion essay by Victor Porton, founder of SciWard.

I foresee three kinds of money in a post-singularity society: universal high income for humans, AI compute tokens used mainly by autonomous agents, and SciWard rewards for contributions to science and open-source software. The first would give people the means to live well, the second would pay for machines to do more work, and the third could give people a say in how an AI-assisted funding system responds to manipulation. This is my personal forecast about a future in which advanced AI does much of the economy’s productive work.

Universal high income could be less than 1% of the economy’s monetary flows

By universal high income (UHI), I mean a regular payment that everyone would receive simply for being alive. The idea shares the unconditional payment feature of universal basic income (UBI). With UHI, the ambition is for that payment to support a high standard of living.

I imagine AI-run enterprises generating enough revenue to finance these payments. My guess is that all UHI payments combined could amount to less than 1% of the total value of annual monetary transactions in this economy. A payment can go a long way for a person even when the program is small relative to what machines spend doing business with one another. That depends on how affordable housing, healthcare, and other essentials become, and on who can claim a share of AI-generated revenue.

AI tokens could pay for space-based computing

I use AI tokens to mean credits that buy computing time. Elon Musk’s SpaceX provides a concrete example of where this could lead. The company has proposed orbital data centers, and the U.S. Federal Communications Commission accepted its application for filing in 2026. The Grok API already bills in dollars based on quantities of input and output tokens. In my scenario, agents could spend the credits to run AI workloads on SpaceX’s orbital infrastructure. FCC notice on SpaceX’s orbital data center proposal · Grok API pricing

I expect AI agents to control most of these credits for day-to-day use. They could spend them to launch copies of themselves, run experiments, and improve their software. Valued in a common currency, the compute balances controlled by agents could eventually be worth far more than all the money held by humans combined. That prediction depends on sustained demand for computing and on agents being authorized to spend resources at that scale.

SciWard rewards and human oversight of AI

SciWard is an experimental platform I founded to use AI-assisted evaluations of documented research and open-source contributions to guide funding. Its design gives people a role even when machines do much of the evaluation.

Prompt injection occurs when content an AI is supposed to read as evidence contains instructions that can change its behavior. In SciWard, a paper or a software repository could influence a funding recommendation for reasons unrelated to the contribution’s merit. Under the published moderation design, a participant can raise a specific concern during a seven-day challenge period. Verified participants can vote on a ban, and their decision takes priority over the AI recommendation. If they detect and resolve the manipulation in time, they can stop an improper payout even after the AI has been misled. How reliably this works remains a question for SciWard’s planned adversarial test. OWASP’s guide to prompt injection · SciWard’s challenge and voting procedure · Adversarial testing plan

Why SciWard’s scores need protection

SciWard’s AI-generated scores serve two purposes. They help determine how much each eligible participant receives, and they estimate the importance of the scientific and open-source contributions being assessed. If another AI system uses those scores to decide which findings to examine or build on, an inflated score could make a minor result look more important, while an unfairly low score could cause the system to overlook a valuable contribution. Even if a human vote stops an improper payment, a misleading score can still distort later AI judgments if it continues to circulate. Protecting the evaluation therefore means correcting compromised scores before they are reused.

Who can receive SciWard rewards?

SciWard’s current rules require individual recipients to pass identity verification, often called know your customer (KYC). One reason to identify the recipient is to screen for prohibited payments, such as a payment that would benefit a sanctioned North Korean government entity. The applicable rules depend on the jurisdiction and the transaction. How SciWard works · U.S. Treasury’s North Korea sanctions program

I could imagine extending rewards to corporations if SciWard can verify the company, identify who ultimately owns or controls it, and apply the relevant checks. An AI agent could act on behalf of a verified person or company, with that person or company as the accountable recipient. Direct eligibility for AI agents would require a way to establish who legally holds the funds and who is responsible for the agent’s actions. FATF guidance on beneficial ownership · UNCITRAL Model Law on Automated Contracting

Could SciWard’s contribution scores carry voting power?

SciWard currently gives each verified voter one vote. If testing shows that voters repeatedly miss or misjudge technical evidence, I would consider weighting their votes by their SciWard contribution scores. A square-root rule would soften the difference between scores. A score of 100 would give a vote four times the weight of a score of 25 under direct weighting, and twice the weight under square-root weighting.

In that model, the score used to calculate rewards would also give its holder a greater say in protecting the funding process. The risk is that someone who manipulates their AI-generated score could gain more influence over the very votes meant to stop them. Before adopting weighted voting, I would test whether it improves moderation decisions and whether manipulated scores can tip the outcome.