Testing¶
What does actual testing (usage) of current AI tell about its impact on Humans, including our understanding of its capabilities?
Some Cases¶
| Result Nature | Case | Usage Nature |
|---|---|---|
| Success | Tesla FSD | From user perspective - super clear usage. |
| Success | Gemini and analogues for general knowledge extraction | Highly nuanced compared to traditional Search. |
| Success (?) | Google's AlphaFold 3 for Drug Discovery and Vaccine Development | Goes through the drug testing cycle. |
| Success | Google’s flood forecasting | 7-day advanced warnings to 700 million people across 100 countries. |
| Success (?) | Controlling plasma in fusion reactor1 (absolutely huge case for the Humanity) | Being tested |
-
Inside a fusion reactor, you have to hold 100-million-degree plasma using magnets. The plasma is chaotic and "wiggles" out of place in milliseconds, touching the walls and killing the reaction. The AI Solution (2025-2026): We are now using Deep Reinforcement Learning to control those magnets. An AI can "see" the plasma wiggling and adjust the magnetic field thousands of times per second, faster than any human-coded algorithm. ↩