The advent of AI agents will dismantle the traditional advertising business model, requiring content creators to be compensated for the marginal utility their information provides to agents.
The Ads business model will Die for agents
AI agents, unlike humans, do not consume or react to advertisements, making the current ad-supported business model obsolete. An Uber banner ad, for instance, becomes worthless when an agent is the primary customer.
A new 'AdSense for agents' model is proposed, where content owners receive variable payments based on the benefit an agent derives from their information.
AI model Alignment is not Adversary proof
Even with good intentions and alignment training, AI models remain vulnerable to exploitation by individuals who 'want to do bad things. ' This vulnerability stems from a lack of 'proportionate countermeasures to contain them,' rather than the AI's inherent 'world-hacking' power.
The focus should be on closing these security and alignment gaps.
Agent-Driven Web search requires new infrastructure
AI agents will interact with the web 'thousandx more' than humans, making existing web search infrastructure inadequate. Web search for agents requires a 'completely different technological approach' because inputs, outputs, time constraints, and compute allocation are vastly different.
This necessitates new technology and business models tailored for agent consumption.
The Web's shift from pull to push for agents
With 'always-on agents' working on behalf of users, the web's interaction model will shift from agents constantly querying for updates to the web itself notifying agents when specific, relevant events occur. This 'web event stream' represents a transition from 'pull to push,' enabling more efficient and timely agent operations.
So the web event stream is the web going from pull to push and I'm super excited about that because as you have more and more persistent agents you're going to see incentives to move people to move queries into push on search rather than pull on search.