Click a card to explore each era's technical depth
Symbolic Logic: The belief that intelligence is the formal manipulation of symbols and rules.
High transparency; every decision can be traced back to a specific logical rule.
Moravec's Paradox: Logic is easy, but perception/mobility is hard. Resulted in the 1st AI Winter.
Knowledge Engineering: Transferring specific human domain knowledge into "If-Then" engines.
Commercial viability in specialized fields like medicine and oil exploration.
The "Knowledge Acquisition Bottleneck"—it was too hard to manually code everything a human knows.
Data over Rules: Moving from "telling" the AI what to do to "showing" it patterns in data.
Robustness; can handle noisy, real-world data better than rigid logical rules.
Feature Engineering: Humans still had to manually tell the AI which parts of the data were important.
Hidden Representations: Deep Neural Networks that learn their own features directly from raw data.
Superhuman performance in image recognition and speech-to-text.
The "Black Box": Hard to explain *why* a deep network made a specific decision.
Autonomy & Multimodality: From AI that predicts to AI that creates and acts across all media.
Drastic reduction in the barrier to entry for complex tasks; autonomous problem solving.
Hallucinations, AI safety, and the massive compute energy required for training.
Exponential Processing: Utilizing quantum mechanics to break the limits of classical silicon.
Solving optimization problems and molecular simulations that are impossible for current supercomputers.
Error Correction (Decoherence) and the extreme cold temperatures required for hardware.