🚩 There’s more to AI than just GenAI.
Let’s talk about how to navigate it with your business.
But first, some philosophy and economics (bear with me).
AI is simply a type of technology. And, at its core, technology is just… cost savings.
Consider these historical cost-saving examples:
🛞 The wheel: getting from point A to point B.
📰 The printing press: spreading knowledge.
🛜 The internet: sharing information.
🔮 Now, AI is reducing the cost of prediction. GenAI is but one type.
For example, ChatGPT takes text inputs to predict text outputs.
We call AI "intelligence" because prediction is a core step in how humans complete tasks:
1️⃣ Input data: “My leg hurts because I fell.”
2️⃣ Prediction: "Maybe it’s broken."
3️⃣ Training data: "The two options are: get surgery or take Advil."
4️⃣ Judgment: "Surgery might be the best option."
5️⃣ Action: Get surgery.
6️⃣ Feedback data: If it’s then discovered it wasn’t broken (whoops), the doctor learns to make better predictions.
Like how humans learn, this is how AI models get trained, constantly refining predictions via feedback. 📈
Why does this matter for your business?
If AI reduces the cost of prediction, then basic economic theory dictates that the demand for Prediction’s complements will rise (Data and Judgment), and the demand for its substitutes (human prediction) will fall (e.g., Actuaries).
Ask yourself:
🤔 What "manual" predictions are you or your users making?
- E.g., who are the actuaries in your business?
🧪 Does your product gather data that could train better AI models? Can you use data users generate from using your product (through a recommender model) to rank-sort the suggestions you show them?
- E.g., data that can help train AI to predict the location of gold deposits.
💡 Can your platform be where judgment and action happen in your industry?
- E.g., an app to visualize the location of those deposits and/or deploy resources. Hint: the best UX here isn’t another chatbot.
Pro-tip: You’re unlikely to beat Google or OpenAI at their own game. Pick battles you can win by competing on more niche predictions and judgments. That’s why, at Pilot, we’re not trying to beat Google Travel with an itinerary generator.
Instead of competing, we integrate and complement tools like ChatGPT by focusing on where decisions and actions happen. For us, this means being more of a productivity tool for travel planning and discovery. By designing our UX around judgment (finding, deciding, and organizing travel plans), users’ usage also gathers data to train more personalized AI models for their specific needs on top of generalized LLMs.
Netflix did this. Their recommender engine isn’t as flashy as GPT, but it predicts what you’ll watch better than any GenAI. 🍿
The takeaway here:
Look at your users' contexts and predictions, judgments and actions. Build products around those, and know where you can compete. Work backward from their problems within those to the technology—not vice versa.
Connor J. Wilson
Founder, CEO
Pilot
Originally posted on 🔗Linkedin🔗
