The Evolving Limits of AI

From Data to Wisdom

1. The Foundation

Jeff Crume explains that AI isn't just about processing numbers; it's about moving up the value chain. Intelligence starts with Data and ends with Wisdom.

Decision

Select Age-Appropriate Game

2. The Language Barrier

Early AI saw words as rigid symbols. Modern AI sees meaning (vectors). See how they handle the idiom: "It's raining cats and dogs."

1965 Logic (Keyword Match)
IF input CONTAINS "cats" OR "dogs"
THEN PRINT "Tell me about your pets"
...
Modern Logic (Vector Space)
...

3. Breaking Hallucinations

Generative AI has a vivid imagination, which leads to hallucinations. RAG (Retrieval-Augmented Generation) forces it to do research before speaking.

Mode: Pure Generation (Prone to Hallucination)
User
Facts
DB
AI
User Query: "What is the capital of Mars?"
Waiting for input...
Under the hood: Prompt = Question

4. The Sustainability Limit

We can build smarter models, but intelligence requires energy. Try to achieve High Intelligence with Low Energy.
Hint: Specialized models (Small Size + Curated Data) are efficient.

Energy ≈ Size × Data × (Inefficiency)
Intelligence
CO2
Energy/Cost
Adjust sliders...

5. The Human Role

AI executes the How. Humans define the Why. You are the architect. Assign the tasks correctly to solve: "Make the city better."

Reduce Traffic
Optimize Lights
Lower Pollution
Route Trucks

Human (Why/Outcome)

AI (How/Execution)