We Made This Mistake With Social Media. Let’s Not Make It Again With AI.

I have a confession to make: I find the great debate over whether artificial intelligence is going to save humanity or bring about the apocalypse profoundly uninteresting.

Perhaps AI will cure cancer, eliminate mindless work, unlock unimaginable scientific discoveries and usher in an era of human flourishing. Perhaps it will destroy millions of jobs, hollow out our brains, flood the internet with synthetic garbage and eventually decide that humanity was an unfortunate beta test. I have no idea, and neither, I suspect, does anyone else.

I also have surprisingly little interest in figuring out how AI is going to reshape every aspect of my family’s existence. Maybe that comes from a place of protectiveness. Maybe I simply hate the idea that a handful of technology companies and their executives could have such an enormous role in determining how our children learn, how we work, how we consume information and how we interact with one another, while the rest of us have remarkably little say in the matter.

Or maybe it comes from having spent considerable time in parts of the world where the conversation about transformative technology feels almost absurdly detached from daily life. I have lived and worked in communities where access to clean water, reliable electricity and quality education are not the norm. Against that backdrop, listening to wealthy people debate whether artificial intelligence is going to transform humanity can sometimes feel like a conversation about which humanity we are actually talking about.

Or perhaps I just find the whole thing a little boring. Call me crazy.

What interests me far more is the considerably less cinematic question of what happens between now and the arrival of either utopia or Armageddon (or neither), because whatever AI eventually becomes, today it is a collection of extraordinarily powerful products being built by extraordinarily powerful companies competing for market share, capital, talent and dominance.

We should dispense with some of the romanticism surrounding this industry and recognize these companies for what they are: corporations seeking to make enormous amounts of money. That is not an indictment of capitalism. It is how capitalism works. Companies identify opportunities, exploit advantages, capture markets and maximize profits, and when protecting the public conflicts with those objectives, history gives us very little reason to assume that the public interest will prevail voluntarily.

There is therefore no reason to grant AI companies the benefit of the doubt. When safety conflicts with growth, we should expect growth to frequently win; when restraint threatens market share, restraint becomes considerably less attractive; and when voluntarily delaying a product means allowing a competitor to capture millions of users, it would be foolish to expect executives to choose the public good over the commercial interests of the companies they are paid to lead.

The executives running these companies also have no democratic mandate to decide how much risk the rest of society should tolerate. They did not receive one because they founded a successful company, raised billions of dollars or built a remarkable piece of technology. If their products are in fact going to transform humanity, then the public has every right to establish the terms under which that transformation occurs.

We have already made a version of this mistake with social media, where the early promise of a great democratizing force that would connect humanity, expand access to information and give ordinary people a voice eventually collided with a business model that turned human attention and rage bait into a commodity. The platforms discovered that understanding and exploiting our psychological vulnerabilities could be extraordinarily profitable, and they built increasingly sophisticated products designed to keep us scrolling, clicking, reacting and returning.

It’s not up for debate – social media platforms have contributed to an information environment rife with harassment, hate, conspiracy theories, extremist content and mis-/disinformation, while features designed to maximize engagement have created particularly serious concerns for children and adolescents.

The companies behind these platforms also did not simply stand by as the evidence of these harms mounted. For years, major platforms fought regulation, deployed armies of lobbyists and lawyers, challenged restrictions in court and resisted efforts to impose greater responsibility for what happened on products they designed, controlled and monetized, while Washington repeatedly failed to summon the political will necessary to meaningfully constrain them.

Consider Section 230, a provision of the Communications Decency Act passed in 1996, when AOL was mailing CDs to American households and Mark Zuckerberg wasn’t bar mitzvahed. The basic principle behind this type of protection is easy enough to understand through an analogy. If I picked up the telephone and told a friend I was going to kill someone, I could be held responsible for that threat, but the phone company would not be liable simply because I used its network to make it. The phone company provided the means of communication; it did not decide what I said, whom I said it to, encourage me to say it, or reward me for saying it.

Section 230 gave emerging online services a powerful version of that intermediary protection. Online platforms generally could not be treated as the publisher of something simply because one of their users posted it, which made considerable sense for an early internet in which a service hosting millions of conversations could hardly be expected to assume legal responsibility for every sentence typed by every person using it.

The problem is that today’s social media platforms bear very little resemblance to a telephone line.

When you pick up the phone, Verizon does not listen to your conversation, determine which statements make you angriest, search through millions of other conversations for even more provocative material and then start piping those conversations into your ear because an algorithm has calculated that outrage will keep you on the phone longer. You choose whom to call, you choose whether to continue the conversation and, if your friend starts spouting hateful rubbish, at least the rubbish came from the person you chose to speak with.

Social media fundamentally altered that relationship. Its algorithms decide what billions of people see, recommend what they should see next and learn from every click, pause, like and share what will keep each individual engaged. Hate, conspiracy theories, radical content and outrage do not simply sit somewhere on these platforms waiting for someone to find them. Recommendation systems spoon-feed users material they never requested because the system has learned that provocative content keeps them watching.

Yet these companies benefited for years from a legal and political framework rooted in the notion that the underlying speech belonged to somebody else. Section 230 is not absolute immunity and it is not responsible for every failure of technology regulation, but the broader arrangement became increasingly difficult to justify as platforms designed the systems selecting and amplifying what people saw, profited enormously from the attention those systems generated and then distanced themselves from responsibility by emphasizing that somebody else created the underlying content.

Policymakers should have confronted that contradiction much earlier, rather than allowing the companies to scale first and asking the hard questions later. By treating regulation as something that should arrive only after harm had been conclusively demonstrated, quantified and litigated, we gave the platforms years to capture markets, normalize their products and make themselves nearly indispensable to modern life before seriously debating the rules under which they should operate.

AI companies should not receive the same luxury, and part of preventing that outcome begins with consumers because these products are useful, which is precisely what makes dependence so easy.

Parents should be deeply skeptical about allowing children to outsource intellectual development to machines. Teachers should establish strict rules governing when AI can be used, with serious consequences for students who violate them. School districts should be developing policies now that preserve the development of critical thinking, writing, research and problem-solving skills, while universities need to determine what intellectual work students must still be capable of performing without machine assistance.

Adults should apply the same scrutiny to ourselves. If AI is making you less capable of writing, thinking, remembering, navigating, researching or making decisions independently, perhaps the appropriate response is not simply to celebrate how much time you saved, but to ask whether the convenience is actually worth what you are giving up. There is something bizarre about worrying that artificial intelligence will eventually surpass human intelligence while eagerly outsourcing every task that requires us to exercise our own.

Consumers were also part of the failure of the social media era. We knew these products were manipulating our attention, watched what they were doing to our children, our politics, our information environment, complained about it endlessly and then opened the apps again. We never built a sustained consumer movement capable of seriously affecting the companies’ bottom lines, which made government and civil society even more important at precisely the moment when both proved far too slow to respond.

We cannot afford the same timidity with AI, particularly as we begin hearing familiar arguments about why meaningful regulation should wait: policymakers do not understand the technology, regulation could stifle innovation, America cannot afford to fall behind China and the technology is changing too quickly for laws to keep pace. There will never be a perfect regulatory framework for a technology that is constantly changing, and insisting upon one before acting effectively allows the companies developing these systems to set the rules themselves.

Government should instead establish meaningful standards for safety and transparency, give independent researchers access to study these systems, strengthen protections for children and subject particularly high-risk applications to independent testing and oversight. Policymakers should also be willing to overregulate and correct course where necessary. If a rule turns out to be unnecessary, burdensome or poorly designed, we can change it. It is much harder to claw back control after a technology has become deeply embedded in society and billions of people have become dependent on it.

Nor are policymakers starting from scratch. Organizations such as the Center for AI and Digital Policy, Encode, the Center for Countering Digital Hate and Fathom are already advancing approaches to AI accountability that policymakers can draw upon. The objective should not be getting every rule right on the first attempt, but acting, studying the consequences and adjusting where necessary rather than allowing the absence of universal agreement to become an excuse for paralysis.

Civil society needs to move just as quickly. During the social media era, children’s advocates, civil rights organizations, Jewish organizations, democracy groups and others spent too long treating technology as somebody else’s policy problem, only to discover that they were confronting different manifestations of problems traveling through the same technological infrastructure. Those coalitions should be built now, while the rules around AI are still being written. AI companies are already devoting enormous resources to shaping the political environment in which they will operate, and civil society needs to approach the task with comparable seriousness.

Jewish philanthropy has a particular reason to understand this lesson because, for too many years, much of the Jewish communal world was absent from the technology accountability debate. We subsequently watched social media become an extraordinarily effective vehicle for antisemitism, conspiracy theories, extremist recruitment and harassment. Jewish organizations have since spent enormous sums responding to those downstream harms, but we had far too little skin in the game when it came to advancing the policies, coalitions and regulatory efforts that might have held the platforms accountable much earlier.

By the time much of the Jewish communal world recognized that technology accountability should be a serious priority for philanthropy and advocacy focused on countering antisemitism, the architecture and incentives of the social internet had already been built, and we have been paying for that late arrival ever since.

Jewish philanthropy cannot afford to make the same mistake with artificial intelligence. Funders should be strengthening the organizations, coalitions, researchers and policy advocates working to regulate both social media and AI, while Jewish institutions should be sitting alongside children’s advocates, civil rights organizations, democracy groups, educators and others whose interests overlap with ours.

Jewish philanthropy cannot afford to make the same mistake with artificial intelligence. Funders should be strengthening the organizations, coalitions, researchers and policy advocates working to regulate both social media and AI, while Jewish institutions should be sitting alongside children’s advocates, civil rights organizations, democracy groups, educators and others whose interests overlap with ours.

This does not require creating some separate Jewish AI agenda. It requires recognizing that technology policy is now inseparable from many of the issues Jewish philanthropy already spends enormous amounts of money addressing, and that continually funding organizations to respond to harms downstream while ignoring the systems helping to produce and distribute those harms upstream is neither sufficient nor particularly strategic.

We can spend the next decade funding organizations to combat AI-generated antisemitism, synthetic propaganda, extremist recruitment, harassment and whatever other manifestations emerge after the technology has already reshaped society, or we can invest now in the coalitions and policy infrastructure capable of influencing the rules before the architecture becomes nearly impossible to change. In reality, we will probably have to do both, but our experience with social media leaves us without an excuse for ignoring the second.

I remain agnostic about whether artificial intelligence will ultimately save civilization or destroy it, and I still find much of that debate remarkably boring. What concerns me is something far less theoretical: a relatively small collection of extraordinarily wealthy companies is racing to make its products indispensable to billions of people while governments, civil society and the public are once again struggling to keep pace.

We should view these companies with the same healthy skepticism we would apply to any other industry asking society to trust that it can police itself while pursuing enormous profits in a brutally competitive market. Their executives are not elected representatives, they have no special mandate to determine how much risk society should tolerate and assurances about responsible innovation should never substitute for enforceable rules created by institutions accountable to the public.

The failure of the social media era was not simply that we failed to predict what would happen. It was that we continued giving technology companies the benefit of the doubt long after there were compelling reasons not to, allowing products and business models with profound consequences for public life to become deeply entrenched before imposing anything approaching adequate accountability.

Having lived through that experience once, we have no excuse for handing AI companies the same freedom and hoping that this time they will somehow police themselves. 

Coby Schoffman is a Los Angeles–based serial social entrepreneur and the founder of The Nation Foundation (TNF), which operates project zones across East Africa. Schoffman holds an MSc in Transnational Security from New York University and a BA in Counterterrorism and Conflict Resolution from Reichman University. The views expressed are his own and do not reflect those of any affiliated organization.