I just held my first keynote.
Riding that high, let's dissect why self-driving cars serve as a case study where machine learning is headed.
Recently, I checked the Gartner hype curve for AI, and it looks like machine learning is headed for the trough of disillusionment. I don't suspect it will be another AI winter, but more on that later.
Self-driving cars have matured through the trough and are heading towards the promised land.
Teachable Moments from Tesla and friends
Self-driving cars have been decades in the making.
The AI in self-driving cars sits at the intersection of multiple technologies. It happily uses machine learning where necessary, but the trajectory planning, for example, is classic control theory and A* for pathfinding.
Moreover, Tesla is a company that heavily invests in infrastructure and its employees. Hiring Andrej Karpathy? Probably a good move. Building one of the biggest GPU servers? Probably a good move. They have not shied away from the investment, and it shows.
This intellectual openness to take existing solutions, reach out to other fields better approaches, state-of-the-art machine learning, and persistence and targeted investment can change an industry.
Is Winter Coming?
We certainly have an interesting time ahead.
AI doesn't stand alone anymore. Machine learning is applied in all segments of our world, successfully at that! It does things in medicine, weather prediction, and obviously self-driving cars that were not possible before.
While machine learning as a discipline is approaching the trough of disillusionment, I see its applications of recent advancements fuel other scientific disciplines successfully.
Moreover, the general public may actually be happy that general AI hasn't been achieved this time due to a hefty serving of skepticism, including(!) the Tesla CEO Elon Musk.
While it may not be the hype summer anymore, where we can slap the AI label on a linear regression, it looks like machine learning is around to stay.