Demis Hassabis recently delivered a loaded mini-monologue regarding the role of advertising in AI. It’s a sentiment worth unpacking:
“I’m a little bit surprised they’ve moved so early into that. I mean, look, ads, there’s nothing wrong with ads… they funded much of the consumer internet. And if done well, they can be useful. But in the realm of assistants, and if you think of the chatbot as an assistant that’s meant to be helpful—and ideally, in my mind, as they become more powerful, the kind of technology that works for you as the individual… there is a question about how ads fit into that model?… You want to have trust in your assistant, so how does that work?”
Hassabis is operating on a specific set of assumptions:
- Ads are acceptable if executed well.
- Ads are the foundation of the modern internet.
- Chatbots are, by definition, assistants.
- Assistants must work solely for the individual.
- Trust is the primary currency of the assistant relationship.
- Ads inherently break that trust.
Frankly, I expected more depth from one of the primary architects of our future. While it’s refreshing that he doesn’t default to the reflexive “ad-hatred” common among techies, this brand of passive-aggressive intellectualism doesn’t actually move the needle.
Let’s break down where his logic falters.
1. Are Chatbots Only Assistants?
Hassabis defines the entire medium by a single use case. While many use chatbots as assistants, the range of human expectation is likely far more complex and varied.
Expanding the “assistant” argument to the entire class of AI services is like being in 1994 and claiming, “Websites are just digital Yellow Pages.” By narrowing the scope so early, we risk stifling services that haven’t even been invented yet—many of which may never exist without ad-supported models.
2. The Myth of the Egoistic Assistant
We use the word “assistant” in many contexts: shop assistants, flight assistants, medical assistants. In almost every case, we are perfectly aware that incentives are mixed.
Hassabis seems to have a very specific “assistant” in mind—perhaps a Victorian butler? That’s a lovely vision, but even a butler requires a salary, room, and board. If the user isn’t paying the “salary” through a premium subscription, someone else has to. I suspect premium chatbots will be the last to see ads, but for the rest of the world, transparency and disclosure are better tools than exclusion.
3. The Fragility of Trust
Hassabis suggests that trust and advertising are mutually exclusive in the AI space. But trust is the basis of all civilization, not just the assistant-user relationship.
We trust businesses every day once we understand their business model. We trust a restaurant to provide good food, even though we know they are trying to upsell us on the wine. We trust the service until the “value-to-nuisance” ratio breaks. Why should AI be any different?
4. Does Advertising Inherently Corrupt?
Hassabis’s core argument is that ads break trust “from the outset.” Let’s test that with a few speculative examples:
- If your “butler” filters your mail and highlights a promotion for a trip to Iceland—one you’ve been dreaming about—is that a breach of trust or a helpful suggestion?
- If your “butler” mentions a new local bistro when you ask for dinner ideas, is that corruption or local expertise?
The conversation with a chatbot feels “private,” which is a powerful UX advantage. But the mobile phone felt intensely private for its first few years, too. Privacy and monetization have always found a way to co-exist, however uneasily.
The Real Question
The issue isn’t whether ads belong in AI; it’s about the execution. We should be asking:
- How disruptive is the experience?
- Where is the line for “too much” advertising?
- Who has earned the right to occupy my digital attention?
- Is there a mechanism to punish abuse?
Let’s keep a close eye on how chatbot advertising evolves and be vocal about bad practices. But let’s stop pretending that the presence of ads is “unnatural.”
Leave a Reply