I met someone recently who was a passionate supporter of online ads. I met him in the audience of a lecture. He was speaking to a friend. Within minutes of meeting me, he was offering a veritable ode to ads — to online ads, the type that grace your social media feeds. He wasn’t trying to be contrarian in that edgy, self-aware way of arguing that something everyone else thinks is bad is actually good to show how open minded and un-socially constrained one is. He seemed to genuinely believe online ads were very good and seemed somewhat frustrated that the world did not share his opinion.
(Written in my personal capacity. Views expressed are my own and may not reflect those of Anthropic.)
Here was his case:
Ads transmit information — they tell you about products you wouldn’t otherwise have known about.
Ads make a lot of money — which tracks value created.
Ads make valuable transactions happen, which is both good for the participants and generates valuable data that can help us understand consumption and economic patterns better.
Ads offer you choices, and it’s up to you what you do with them. More choices are good. We shouldn’t be paternalistic. Let people choose.
Ads do all of these things, and have allowed for the creation of incredibly high social surplus goods like Google Search and Google Maps. But ads also actively shape our desires and behaviors in ways we may not reflectively endorse, degrade user experience, and create a misalignment between the service provider's incentives and the user's best interests.
On Value Creation and Capture
Ad Man described himself as “a capitalist” which, while we didn’t fully dissect it, based on other statements he said encompassed “generating returns basically tracks value — if I make a bunch of money from something, it means I created a lot of value.” And “I’m pro choices — I believe that what people choose is good for them and that giving people choices is good, and that we should support their ability to make them.”
Some people might call me a capitalist because I’m broadly appreciative of market mechanisms and view them as the best way yet found to incentivize production, transmit information, and distribute resources. But I think this man might not have considered me one, because I’m also conscious that markets have externalities that sometimes need to be corrected by additional incentives.
The "capitalism tracks value creation" argument contains an important kernel of truth but misses some nuance about value capture ratios and externalities. Yes, if you've earned $100,000 in a market transaction, you've probably created at least that much social value — otherwise, why would anyone pay you? But the relationship between private profit and social value isn't uniform across different types of economic activity. Consider two people each earning $100,000: a high-frequency trader and a vaccine researcher. The trader's algorithm might create $120,000 of total value, with their firm capturing most of it. Meanwhile, the vaccine researcher's work might generate millions in social value through prevented illnesses, saved lives, and increased productivity— they're just capturing a tiny slice of that massive surplus. Even more complex are cases with significant negative externalities, like advertising, where the private profit might accurately reflect value created for the immediate parties (advertiser and platform) while simultaneously generating societal costs through attention pollution, preference distortion, and overconsumption. Simply pointing to profit as proof of value creation misses this variance in how markets actually translate social value into private returns.
On Choices
And re choices, I think choice is very often good, but that people make a range of choices — some of which seem to reflect their values or reflective desires more than others — and that we shouldn’t treat them all the same.
For instance, if I were to overspend my credit card limit on online shopping after a night of drinking, what should we think of this choice? My guess is we should be more skeptical of it than the choice I make in a session of New Year resolution making where I say I plan to save 20% of my income. Why? One is done in an impaired mental state on the spur of the moment, and seems likely to hurt me in the long run. One is done in a clear mental state after reflection, and seems likely to help me in the long run.
With online ads, what part of you are they speaking to? What choices are they enabling or hindering?
Ads change us
With Ad Man, I kept returning to the point that ads do not just act on a static person. If I had a given set of immutable desires, and ads could simply answer or not answer them, we’d be in a simpler place. But I don’t. Ads often seem to make me want something even as they offer me the path to procuring it (Capitalism’s ability to create needs is one thing I think Karl did get right). Maybe this feeling is an illusion, but it certainly doesn’t feel like one. I don’t sit around lamenting the fact I do not own an indoor herb garden or a new shape of face massager until Instagram repeatedly hawks it to me with videos of beautiful, happy people with fresh, nutrient-dense herbs and taut faces. If I were to buy these, Ad Man would be happy: information was transmitted, I learned something I previously did not know, value was created — I paid these folks, they gave me products that ostensibly made me better off, and I made free choices.
So why would I think that was bad?
Ads as attentional garbage. I used to love Pinterest. But at some point a few years ago it started having multiple ads per page wedged in among your pins. It felt like a space that had previously been beautiful was now sprinkled with litter.
Ads as coercion. If I were to buy some things off Instagram that I didn’t previously want, it feels like I was changed in some way without my full consent. Sure, I chose to use Instagram, but I didn’t choose to have instagram make me want an indoor herb garden. I didn’t necessarily want to want that.
Ads as a spur for not reflectively endorsed consumption. Ads make me want things, and this leads me to spending money that could have otherwise been saved or invested. The median American household has around $8k in their bank account— it seems like for many, saving more would be wise. Investing is also beneficial both for individual wealth building and providing crucial capital for innovation and economic growth.
Could ads hypothetically be good?
I granted Ad Man that there is a very narrow set of worlds where I would view ads as good. Conceivably, ads could meet some need you have that you reflectively endorse.
How much do ads reflect your reflective self? How much do ads help you become more of the person you want to be?
Like probably all young women, I have recently started to receive ads for Ozempic. One thing Ozempic is purported to help with is reducing cravings for alcohol. If I were an alcoholic who saw ads for Ozempic on Instagram and, because of them, started to take it and through doing so, improved my relationship to alcohol — then yes, this ad would have been very good! It would have helped me become more in line with my reflective self. Conversely, if I was psy-opped into taking Ozempic because all the skinny influencers are nonetheless taking it… (Funnily enough, Ozempic may also reduce cravings for shopping!)
How much do ads decrease the experience you are having?
There is a world where ads are so beautiful, entertaining, or information dense that they are as good, if not better, than the content around them. For instance, I find the art nouveau ads of Alphonse Mucha very beautiful — I would frame them in my house.
The reality is that basically no ads have this property. They tend to be “less good” than the content around them, often just for the sheer reason that they are not quite what we want. I just went on Pinterest. My first ads: Jewelry, Fancy Coffee Maker, Clothing, Furniture, Fancy Supplements. This is ~50% of the page.
The clothing was not quite to my taste, so that stood out. I wasn’t looking for coffeemakers on Pinterest and never have, so that stood out. And likewise, I have never browsed for supplements on Pinterest, so that stood out. These had the effect of making the page less relevant to me, less beautiful, and less enjoyable. (You’ll notice some of the ads did blend in better! That said, none of them were the most beautiful or relevant pin on the page).
I clearly still enjoy Instagram and Pinterest enough to use them with ads. But if I could pay to use them without ads, I would do so.
The Best Case for Ads
The best case for ads, in my book, is that they are the cost of funding a valuable product. I really enjoy and benefit from Google Search. Google Search is funded by ads. I think Google Search has made the world much better for its existence. Ditto Google Maps.
I can pay to remove the ads on Youtube, and I have done so. For some, that cost might be prohibitive, and I can imagine that for many, despite the ads, they still derive an immense amount of value from Youtube — whether it be homework help, language learning, home repairs, workouts, you name it.
Going back to the “how much social value is created by a transaction” — value capture ratios might actually support Ad Man's position. While any individual advertising transaction might have externalities and a high private capture rate, the ad-supported model as a whole could be seen as enabling massive social surplus — after all, the billions in uncaptured value created by free access to Google Search, Maps, and YouTube's educational content might dwarf both the private profits and negative externalities of the ads that fund them and other free platforms. It is deeply unclear to me how this all nets out, but this was the steelman I tried to offer Ad Man.
What about AI?
Ad Man was making the case that AI should be paid for by ads. There was a trillion dollar economic opportunity here, and it also would be Good.
I think it would be a very narrow target to hit to have an AI assistant which was funded by ads but was genuinely acting in the interests of the user. A subscription model seems much more likely to produce a product that is aligned with the user’s interests than a model which is funded by ads. When I ask current models for recommendations, I trust them much more now than I would if I knew they were funded by ads which could be shaping those recommendations (two separate factors here — trust and actual shaping. Even if no shaping was happening, I would trust the model’s answers less).
Yes, it is plausible that a company paying for an ad could genuinely have the best product for me, and the AI agent could genuinely identify that and thus recommend it. It’s just a narrow target.
Also, any truly useful AI agent needs to know a lot about you. That could enable the agent to make perfectly tailored recommendations — or it could allow it to know what I will be maximally susceptible to being talked into buying. Ad Man might say there’s no real difference between those two. Either way, I prefer to be able to give my agent access without worrying that doing so will benefit someone else — I want the agent to be purely incentivized to benefit me, not advertisers.
AI is a slightly less good product for ads than social networks since AI has less of the network effect than social media does. Even if it becomes more networky (eg. you can collaborate better with someone using the same AI products as you), it still, by its nature, seems less inherently networky than social media.
With something with network effects, ads make more sense. You really want more users on your platform, even if they can’t or won’t pay for it because their being there makes it more valuable. So you make it maximally accessible by making it free and then fund it with ads.
With something with less of a network effect, you have less of a reason to incentivize free users beyond the fact that they might convert to paid. This makes a subscription more fitting than ads anyways.
Nevertheless, someone probably will do AI funded by ads. Ad Man thinks it will be Meta or Google. I’m glad that at least one person will be happy when this happens.
***
(Written in my personal capacity. Views expressed are my own and may not reflect those of Anthropic.)
Comentarios