It costs money to share the future

On AI, targeted advertising, and the subtle cruelty of paternalism

Some friends I made along the way.

The future is already here; it’s just not evenly distributed. — William Gibson

Two days ago, OpenAI announced that they’ll soon introduce ads to free and lower-tiered subscriptions to ChatGPT. Allowing advertising will give them a bigger revenue stream, which they can use to build toward a profitable business structure, improve their product, and expand global access to their platform.

Unsurprisingly, everyone’s mad. Social media apps and search engines famously rely on a similar revenue model, but the data available to advertisers through OpenAI is much more refined and personal. Several dozen articles note that data from user conversations will probably be used for targeted advertising — all the time you spend pouring your heart out to LLMs might lead to lots of ads for Better Help. Of course, hypocrisy throws fuel on the fire. In 2024, Sam Altman said the potential combination of ads and AI was “unsettling.”

Personally, I’d also prefer a world where ChatGPT doesn’t try to sell me Athletic Greens after I ask it for a workout plan. Still, I can’t help thinking of this conversation between Louis C.K. and Conan O’Brien.

I recommend watching it, because it’s very funny. But the title is self-explanatory. We live at a higher standard than medieval kings, teleport around the world in a day, fit the entire Library of Alexandria on small SD cards — and now, we have FREE access to an intelligent entity that can patiently and capably deal with mundane tasks, translate our words into any language, and explain virtually everything under the sun in whatever way we see fit.

And yet, no one is happy. We don’t just expect access to a polymathic and infinitely forbearing assistant; we apparently expect it at zero personal cost to ourselves.

Undoubtedly, part of the anger here stems from the fact that some people don’t want it at all. A quasi-religious fear of AI has swept through liberal spheres, which is to say, the spheres inhabited by people like me who write op-eds and blog posts on this topic. In these circles, admitting that you find AI useful for certain tasks — that it is, perhaps, worth something — is social poison.

Unless you’re underprivileged in some way, that is; then they’re obligated to hear you out. I have seen this happen a couple times — amidst an angry conversation about the horrors of AI, a Spanish-speaker (for instance) volunteers that ChatGPT enables him to translate his writing into English and thus share his work with the overwhelmingly English-dominated modern world. Everyone else gets quiet. “Right, yeah,” they admit. “That makes sense.”

But those use cases aren’t exceptions; the vast majority of the global population has been saddled by the womb lottery with linguistic, economic, racial and other barriers to success. For them, AI is a godsend.

Consider the graph below, which shows the degree to which people from different parts of the globe feel excited or nervous about the impact of AI.

From The Human-Centered AI Institute.

For some countries, the majority of respondents said they were not excited about products and services using AI — they fall on the left side of the vertical dotted line. Guess who the pessimists are? The United States, Canada, France, Belgium, Japan; countries that are mostly English-speaking, economically dominant, or (usually) a mix of both.

On the other side of the line fall countries that are excited about AI. Overwhelmingly, this group consists of countries in the Global South, like Peru, Malaysia, and Brazil. How strange. What do those misguided saps find so intriguing? Don’t they know about the horrors of plagiaristic undergrads?

Maybe not. Or maybe they don’t care, because AI can help them navigate the 50% of web content that is available only in English, speed up mundane and inefficient features of the informal labor that predominates in poor countries, ease the cognitive load required to do tasks like programming and writing (especially in a foreign language), and help them figure out how to apply to college.

Obviously, the individuals I have in mind won’t usually have the resources to pay for a subscription to ChatGPT. If we want them to have access to this life-changing tool — as we should, compassionate believers in equity that we are — then we should be perfectly fine with companies like OpenAI finding some way to fund the global distribution and maintenance of their gadget.

This, I think, is the biggest problem with elite perspectives on AI. It is a global technology, and we are not representative of the globe. Consequently, we tend to underestimate its benefits (since we don’t need them) to everyone other than ourselves.

Does that mean we should waive aside all concerns about targeted advertising, cheating students, AI slop and all the rest? Certainly not. But this present round of anger, like so much else in the anti-AI discourse, is not only unfounded but cruel. OpenAI needs revenue if they want people to keep accessing their technology for free. That access is indubitably good. Targeted ads are annoying and intrusive, but a world where anyone can write a resume in English is worth the price.¹

If you’re reading this, writing resumes in English probably isn’t an issue for you — the one language to rule them all was your birthright. You’re also probably well-educated, conveniently placed in the West, and comfortable with the tacit knowledge required to move through modern industry. Congratulations: you don’t need any help from ChatGPT. Perhaps I can introduce you to several billion people who do.

1. One counterargument is that underprivileged people who rely on ChatGPT are more easily influenced by targeted ads. To that point, I would suggest going to a South African township, finding some Zulu-speaking teenagers, and asking them if they’re okay giving away data in exchange for free software that can help them apply for jobs. Or save yourself the travel costs, since you already know the answer.