A 90% Chance of AI Saving Humanity：探讨 AI 赋能未来的机遇与哲学思考

Sitemap

Open in app

Get app

Write

Image 1: Unknown user

Superintelligence

Artificial Intelligence

Singularity

Political Philosophy

Science

A 90% Chance of AI Saving Humanity

What abundant intelligence could mean for health, work, justice, and belonging.

Josh Bocanegra

13 min read

1 hour ago

Listen

Press enter or click to view image in full size

The recent headlines about AI come with a number that’s difficult to ignore: A 10% chance of ending humanity. In September, CNN covered warnings from AI researchers, including one who put the risk above 10% within the next decade.

Those judgments deserve serious attention. But a conversation that gives so much imagination to our extinction should give at least as much to the lives we will get to keep.

What about the other 90%?

That number isn’t a measured probability of salvation, and surviving advanced AI doesn’t automatically deliver a better civilization. It’s a question about the future we intend to build. My conviction is that AGI will make material life easier, expand human freedom, and give us the means to address problems we have spent generations treating as permanent.

That conviction sits outside much of the current mood. Pew Research Center’s 2026 findings show that 52% of American adults feel more concerned than excited about AI, while only 9% feel more excited than concerned. I understand the anxiety. I also think we are bringing an unusually selective memory to the conversation, judging an emerging intelligence while overlooking the world human intelligence has already built.

The future starts with a look in the mirror

Press enter or click to view image in full size

Before we decide that AI is too dangerous, biased, or morally compromised to help us, we need to look in the mirror. We have organized societies around inherited advantage, sent people into wars, tolerated preventable suffering, and made access to basic necessities depend on systems many people cannot navigate.

Human judgment deserves the same scrutiny we apply to machine judgment, especially when the comparison concerns who receives care, who gets heard, and whose life is allowed to improve.

AI learns from the record of humanity, including our discoveries, arguments, stories, prejudices, and attempts to understand ourselves. Its behavior also reflects the choices made in training and deploying it. When something ugly appears in that reflection, we should ask what we have taught, rewarded, and allowed to persist. We should also recognize the beauty in building intelligence from the accumulated work of people who will never meet one another.

I see AI as a continuation of an old human ambition. Across religious traditions, people have imagined beings with greater knowledge, the ability to relieve suffering, and answers to questions beyond individual understanding. That desire reveals something about us: we know our limits, and we keep reaching beyond them.

“Humans are the ultimate adaptation machine, but so many people focus on all of their limitations rather than their potential.” — Tom Bilyeu

Artificial intelligence gives that search a practical direction, turning generations of accumulated knowledge into systems we can question, test, and improve.

For me, there is something beautiful about humanity creating an intelligence capable of helping us become more humane. AGI is our strongest prospect for building a broadly capable, humanlike intelligence whose abilities can be directed toward the greater good. We will have to earn that outcome through the way we build and distribute it, but the ambition itself deserves enthusiasm.

That belief also shaped my work helping write SB 970, later reintroduced as SB 11, to protect people’s image and likeness from misuse. My broader idea was multifaceted: make it harder for bad actors to exploit someone’s identity through public policy, then make it easier for people to give proper permission through licensing. I explained the need for protection in my testimony on SB 970.

Intelligence in service of life

Press enter or click to view image in full size

The first signs already extend beyond conversation. In a 2026 study published in Nature, the AI system Robin generated hypotheses, proposed experiments, and analyzed laboratory results in research on dry age-related macular degeneration.

Scientists carried out the experiments, and promising candidates showed activity in laboratory tests. This was early therapeutic research, with the work of demonstrating patient benefit still ahead, but it shows AI participating in the process through which useful knowledge becomes possible.

AGI will extend that capacity across more fields. Better reasoning, faster discovery, and more accessible expertise will improve our ability to treat disease, design infrastructure, coordinate resources, and build things that are currently too expensive to attempt. Those advances will still need energy, materials, laboratories, construction, and institutions that put them to use. Intelligence will help us improve those physical systems as well as the ideas behind them.

“Intelligence will transition from scarcity to Abundance, and problem solving will become something every human can access.” —Peter H. Diamandis

These advances will also reach beyond human medicine. Nanotechnology is already giving us particles small enough to be delivered directly into plant tissues, and researchers have shown nanoparticles helping plants maintain photosynthesis and improve yields under heat stress.

As AI accelerates this research, it will help us develop more precise ways to support living systems. AI will help humans, plants, and animals live longer, healthier lives.

The psychological consequences will be just as important. We often talk about human behavior as though it emerges from a fixed nature, independent of the conditions in which people live. Yet a person who expects tomorrow to be safe faces a different set of decisions from someone who expects everything to collapse. Change those conditions at scale, and we will change the lives that develop within them.

Who are we without our jobs?

Press enter or click to view image in full size

Work is where many people will first feel that change. Ask someone who they are, and they will often tell you what they do for money. A profession gives them income, a recognizable place in society, and a story about how their past led to their present.

When a machine performs the skill at the center of that story, the adjustment reaches much deeper than a paycheck.

Zack Kass, the former Head of Go-to-Market at OpenAI, has described an “identity displacement crisis.” His TEDxMiami talk, “The Future of Living,” explores the broader opportunities of an AI future, and his vision of a new renaissance resonates with me.

The challenge will be helping people carry their sense of worth into a world where professional usefulness occupies less of their identity.

“We need to stop confusing human value with economic value. We don’t exist to serve the market. The market exists to serve us.” —Andrew Yang

Psychologists Richard Ryan and Edward Deci’s self-determination theory identifies autonomy, competence, and relatedness as basic psychological needs. We want some authorship over our lives, the experience of being capable, and meaningful connections with others.

Employment currently bundles those needs into one role, but our reasons for learning, creating, and caring will outlast the arrangement that makes us earn a living.

People already play instruments without expecting a career, improve at sports without expecting a salary, and cook for someone they love because the experience matters to them. AGI will give those motivations more room by reducing how much of life must be organized around economic necessity. More of our effort will go toward things we value enough to choose.

That future must become tangible in ordinary life. Lower production costs need to reach people through affordable essentials, access to powerful tools, and a meaningful share in the wealth automation creates.

A person who loses their income while their rent keeps rising needs security before a lecture about freedom. Making abundance widely accessible is part of the work of building an AI future worth wanting.

I’ve also shared proposals with Sam Altman on universal basic income and data centers. I won’t share the details of those discussions, but I applaud the OpenAI team for being open to new proposals, especially from people like me outside the company. I’m grateful for their willingness to help figure this out together.

Preventing harm at its roots

Press enter or click to view image in full size

Once we take that seriously, the implications for crime become difficult to ignore. A person stealing to eat has a motive that disappears when food becomes reliably available. Meeting that need removes a reason for the crime before anyone needs to investigate it. This is the kind of prevention AGI will help us pursue, by making necessities easier to produce and support easier to reach.

Get Josh Bocanegra’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

There is already evidence that access to resources changes criminal involvement. Researchers studying young people with disabilities removed from Supplemental Security Income found a roughly 60% increase in charges for income-generating offenses. That estimate applies to the population studied, but the lesson is consequential: withdrawing support changed people’s behavior in ways that created further costs for them and for society.

Security also changes what a person has the mental space to consider. Research published in Science found that financial concerns impaired cognitive performance in its study settings, including differences among the same farmers before and after harvest.

When the next bill or the next meal occupies your attention, long-term decisions compete with an immediate emergency. A reliable foundation gives the future a more credible place in the present.

AGI will help us build that foundation through better coordination, more accessible education, improved services, and lower costs across essential activities. A stable home, accessible care, and opportunities that lead somewhere give people a future they have reasons to protect. I expect that change to reduce the desperation behind a meaningful share of crime, while freeing resources to address the other motives for harm.

Material security does not remove every human need. People also seek respect, belonging, control, and recognition, sometimes through destructive behavior. Effective prevention therefore includes helping someone develop another way to earn respect, relationships outside a violent group, and the ability to respond to conflict without escalating it.

A randomized study following 999 high-risk men in Liberia found reductions in antisocial behavior ten years after participants were offered a brief cognitive behavioral intervention, with or without a cash grant. The results were strongest among men at the highest initial risk. That is evidence of a capacity for lasting change, and advanced AI will help us make effective knowledge and practical support more accessible alongside the professionals and communities doing this work.

I support the police, the military, and all law enforcement personnel who risk their lives to protect us from those who do harm. I do not support abolishing prisons.

My thesis is that we should support research into how AI can improve mental health and addiction treatment. By treating more of the root causes of crime, we may eventually no longer need prisons, at least not in the way they exist today.

AI will also strengthen investigations when crimes occur. An NIJ-sponsored report on cold-case review describes tools for extracting names and references to evidence from large collections of records. More capable systems will help investigators examine timelines, connect information, and identify contradictions for further inquiry.

Defense teams will have the same opportunity to find overlooked evidence and challenge wrongful conclusions, with each claim traceable to records that people can examine.

The broader improvement will come from combining these approaches. We will have more intelligence available to establish what happened and more capability to change the conditions that precede harm. A society that becomes better at helping people build a life will have less damage to repair afterward.

Government that people can navigate

Press enter or click to view image in full size

Government presents a similar opportunity because so much suffering occurs between a person’s need and an institution’s ability to respond. Someone seeking assistance must discover the right program, understand unfamiliar rules, collect documents, meet deadlines, and challenge an incorrect decision.

A process that appears manageable to its designer becomes an additional burden for someone who is exhausted, frightened, or already in crisis.

Researchers Donald Moynihan, Pamela Herd, and Hope Harvey describe these administrative burdens through learning, compliance, and psychological costs. Their work also explains how burdens reflect political choices. The practical effect is that assistance can exist on paper while the difficulty of obtaining it determines who receives it.

AGI will change the balance of expertise between a person and an institution. An assistant that understands the relevant rules will help prepare an application, identify missing information, explain a decision, and organize an appeal. With the person’s permission, it will carry more of the work that currently demands legal knowledge, persistence, and time they do not have.

We can already see an early version. In March 2026, the UK’s Government Digital Service reported that more than 10,000 users had asked GOV.UK Chat 26,000 questions about public services.

The team reported 90% answer accuracy in its latest evaluation and found that 73% of surveyed app users considered the assistant useful. This was a guidance service, and it gave people a clearer route through information that already existed.

The next step will be carrying that understanding into action and simplifying the processes themselves. We should judge progress by whether people receive the correct service, understand the decision, and can correct an error without losing weeks of their attention. The OECD’s 2026 review of government AI experimentation identifies weak evaluation as a current problem, which makes those human outcomes the right place to focus. AI gives us the means to demand more from institutions that have learned to make their complexity someone else’s problem.

Identity beyond appearance

Press enter or click to view image in full size

I expect an equally profound change in how we understand identity. Race has acquired enormous social meaning, but the importance of a category depends partly on what people learn to associate with it. In experiments published in PLOS ONE, peaceful cooperation reduced racial categorization when race did not predict people’s alliances.

The experiments measured categorization rather than the disappearance of discrimination, yet they challenge the assumption that our familiar divisions must always organize perception in the same way.

AGI will make virtual worlds and flexible identities easier to create, allowing more people to build relationships across appearances they choose for themselves. Your closest collaborator will be recognizable through their judgment, humor, and history with you even when their character changes. A friend will appear as another kind of human, a fictional creature, or something without a familiar racial category, while remaining the person you trust.

My expectation is that race will become far less decisive in worlds organized around those relationships. When appearance is changeable and shared experience is continuous, the category will tell you less than the person has already shown you. That is a social forecast about repeated cooperation and belonging, grounded in the idea that people learn which differences are useful for understanding one another.

The research helps clarify what that change will require. A 2016 virtual-embodiment study reported reduced implicit racial bias after White participants experienced Black virtual bodies, while an August 2026 study found no statistically detectable short-term change after a brief experience.

Research with Black VRChat users also documents barriers in representation and avatar creation. Changing appearance alone is insufficient; the work involves access, how people are treated, and the relationships they build.

AI will give us far more freedom to do that work. People will be able to create environments and forms of self-expression that once required specialist teams, while communities establish cultures around what their members value. Race will remain part of family, culture, and history for those who want to express it. The future I am arguing for is one where it no longer determines a person’s access to safety, opportunity, or belonging.

Belonging beyond a paycheck

Press enter or click to view image in full size

Economic security will support that shift by reducing the pressures that turn another group’s progress into a perceived threat to one’s own survival. It will also expose needs that wealth alone has never satisfied. Even when everyone has enough food, people will still care about being respected, included, and chosen, especially as careers become less reliable sources of recognition.

That is why community belongs at the center of the AI future. People become significant to one another through a shared history of showing up, helping during difficult moments, and creating experiences they want to return to. Their value grows through relationships that contain something specific about who they are.

AI will help organize gatherings, create spaces, translate across languages, and make participation easier, while people give those experiences their personal meaning.

“A use of AI that doesn’t reduce costs, it creates value. It opens new opportunities, leads to growth, connection, and utility.” —Seth Godin

You can be economically unnecessary and personally irreplaceable.

Your child wants your attention even when an AI tutor explains a subject better, and your friend is glad you came even when a machine has more knowledge to contribute. Someone values the meal you made because they understand the care behind it. As economic usefulness becomes a less complete description of human worth, those ordinary experiences will become easier to recognize as foundations for a meaningful life.

This is the renaissance I see ahead. More people will participate in science, make art, build communities, and explore ways of living without first proving that every contribution deserves a salary. Intelligence will become more widely available, and the freedom to decide what matters will reach more people. Our responsibility is to build the economic and social arrangements that let that capability become an ordinary part of life.

AI is a reflection of humanity, including the part of us that keeps trying to exceed our limitations. We have spent generations seeking answers, imagining wiser beings, and passing knowledge forward in the hope that someone will do more with it. Now we are building an intelligence that will help us carry that work further than any individual mind can take it.

I find that beautiful because the ambition reaches beyond making a better machine. It reaches into the kind of civilization we want to become, the suffering we are willing to stop accepting, and the lives we want people to have time to discover. When survival consumes less of our attention, we will have more of ourselves to give to one another.

Every generation has left some part of its intelligence behind in books, discoveries, and lessons passed from parent to child. We are gathering that inheritance into something that could give billions of people the freedom to leave more of themselves behind than the work they did to survive.

Superintelligence

Artificial Intelligence

Singularity

Political Philosophy

Science