{"id":19230,"date":"2024-01-08T14:28:13","date_gmt":"2024-01-08T14:28:13","guid":{"rendered":"https:\/\/www.michaeldempsey.me\/blog\/?p=19230"},"modified":"2024-01-08T14:28:16","modified_gmt":"2024-01-08T14:28:16","slug":"ai-agents-product-velocity-user-interfaces","status":"publish","type":"post","link":"https:\/\/www.michaeldempsey.me\/blog\/2024\/01\/08\/ai-agents-product-velocity-user-interfaces\/","title":{"rendered":"AI Agents, Product Velocity, & User Interfaces"},"content":{"rendered":"\n
\n

I kind of thought of the UI as the product and then machine learning came along and the Machine learning was there to sort of help the UI and that shifted completely. Now AI is the\u00a0product and the UI is there to to help the AI to capture better signal.<\/strong><\/em><\/p>\n\u2014 Gustav S\u00f6derstr\u00f6m, Co-President, CPO & CTO at Spotify,\u00a0Source<\/a><\/em><\/cite><\/blockquote>\n\n\n\n

It\u2019s very clear that some version of the future includes a world where software actions begin to migrate from humans to AI Agents. How this manifests itself is likely a spectrum with two very different angles of approach that have implications on how you build a company moving forward.<\/p>\n\n\n\n

Each approach has its own merits and are perhaps rate limited by technology development and human adoption. This creates complexity for startups that are aiming to build in a world that has\u00a0any<\/em>\u00a0level of agent penetration and forces them to make implicit bets on how they imagine this future playing out and the pace at which it does.<\/p>\n\n\n\n

Scenario 1: One UI to Rule Them All<\/strong><\/a><\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

The first, and perhaps most obvious scenario is one in which agents take a simplistic, classical UI-driven, multimodal approach. Startups will use existing frameworks that humans use to navigate apps and integrate understanding into the agents to accomplish tasks that they believe should be automated and move to post-human.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

This approach is enabled by things like GPT-4V (though some have seen varying success<\/a>), Gemini Pro Vision (via tools like Self-Operating Computer<\/a>) and a slew of open-source models + closed-source models to come. Most recently, a version of this was shown in AppAgent by Tencent<\/a> which used some form of autonomous exploration and imitation learning to control a mobile phone.<\/p>\n\n\n\n

This approach is predicated on a shorter-term moat and slower rate of change.<\/p>\n\n\n\n

Diversity of ability and\/or skill level (% correct actions and theoretically complexity of actions able to be completed) is what will win here and there are likely usage flywheels that can drive and expand competitive advantages based on the learning methods.<\/p>\n\n\n\n

The main need to believe for Scenario 1 to win out is that the time between humans using their phone first-hand today and humans giving up a large # of actions to AI agents will be longer than many appreciate, or merely that AI adoption will be highly fragmented across use-cases.<\/p>\n\n\n\n

Put another way, will humans will always want to do things like scroll Instagram, but perhaps won’t ever want to open TurboTax\/Docusign\/Concur\/Foursquare\/etc. again?<\/p>\n\n\n\n

This will likely bring a very psychological approach to company building in software (something well-known in consumer but lesser used in enterprise), as companies will have to look at a spectrum of tasks and figure out if only<\/em> dopamine driven tasks will still be done by humans in software, or if there will be some areas with human-in-the-loop that exist not because of utility but because of social\/emotional dynamics.<\/p>\n\n\n\n

There thus becomes an existential threat for companies that build software they believe is for humans, only for it to get steamrolled by software made for AIs.<\/p>\n\n\n\n

Scenario 1 isn\u2019t too dissimilar from the argument for humanoid robots, which is that the world is already built for humans and thus we should optimize on a mechanical level the human modality.<\/p>\n\n\n\n

In theory, this sounds nice, however the overall cost of building a humanoid robot and all of the computation needed starts to look more daunting than many appreciate. In reality, when we look at the history of highly proficient \u201crobots\u201d we end up with purpose built, non-human modalities, to repeatedly accomplish a task at very high efficiency and skill level. Like a dishwasher\/washing machine\/making coffee\/etc.1<\/a><\/sup>I found this\u00a0demo that went viral this weekend<\/a>\u00a0somewhat ironic, as the robot test was of a robot using a Keurig (a literal coffee robot) to make coffee.<\/span><\/p>\n\n\n\n

Enter Scenario 2.<\/p>\n\n\n\n

Scenario 2: Agent-Specific UIs<\/h2>\n\n\n\n
\n

Today, I'm sharing AgentSearch, an open-core effort to make humanity's knowledge accessible for LLM agents.

To start, I have embedded all of Wikipedia, Arxiv, filtered common crawl, and more.

The result is over 1 billion embedding vectors, try it at https:\/\/t.co\/TfSrQTfmmb<\/a>. pic.twitter.com\/t2EIfRNx11<\/a><\/p>— Owen Colegrove (@ocolegro) December 21, 2023<\/a><\/blockquote>