Part 5/11:
The challenge lies in the accuracy of these systems; the goal isn't merely to succeed more often than failure but to drive correctly nearly every single day without incidents. Human brains, with approximately 86 billion neurons and trillions of synapses, serve as inspiration, but AI models currently utilize hundreds of trillions of parameters—like GPT-4’s 100 trillion parameters—to enhance learning and decision-making.
The Resource and Technical Toll
A pivotal question is: How much computational power does it take to solve FSD? The speaker cites an experiment that costs around $100 million, a massive investment that, despite being substantial, pales in comparison to the potential market value a fully autonomous system could unlock.