Aka: Why has this burning money pit failed to deliver remarkable results for decades?
The future is here, and it looks like we expected. As we approach Alexnet''s 10-year anniversary, we must constantly assess the potentials of machine learning.
We are coming off the verge of reaching a higher plateau.
We have accomplished things in computer vision, natural language processing, and speech recognition that would have been impossible a few years ago. By all accounts, our AI systems are accurate and accurate.
It''s not enough to say that the answer is yes.
Every prediction about self-driving automobile has been false. We are not living in a future of autonomous cyborgs, and something else has come into focus.
Augmentation over automation.
Humans desire control. It is one of our deepest, most instinctual desires. There is no world in which we give up. One of the greatest mismisunderstandings of the AI community today is that individuals become comfortable with automation over time. The microwave experience of society remains constant.
The situation is incorrect.
The history of technology is not the history of automation; it is the history of control and abstraction. We are tool-builders, so uncomfortable with past experiences beyond our control that for years we developed whole humanities and mythologies around the movement of the heavens. So it is with all technology.
So is the case with AI.
There has been no control. When we look at the successful implementations of self-driving cars now several years old, we see lane assist and parallel parking. In all other situations, where the goal has been to achieve level 5 autonomy, self-driving cars have failed miserably.
Technology is not the bottleneck.
In 1925, a radio-controlled automobile navigating New York City through a difficult traffic collision without a driver. At the 1939 Worlds Fair, Norman Geddes Futurama outlines a feasible smart highway system that would effectively utilize magnetized spikes like electromagnetic fiducials embedded in the road to guide cars. He predicted that autonomous automobiles would be the dominant form of transportation by the 1960s.
He was jeopardized.
No, smart highways have been extremely successful and straightforward where they have been implemented. Even without additional infrastructure, we have self-driving cars today that are more than capable of driving as well as humans. However, even with $80 billion in public transit over a five-year period, the country has never made the largest investment in public transportation.
The difference, however, is that I can actually ride a train.
The problem, at the very least, is that nobody has been concerned about the new controls that were being used. It was never about simplifying driving. That''s a myopic, open-minded way of thinking. It''s about how to transform the transit experience.
They are the most expensive thing a person buys after their home, but they do not create value. It is not an asset that anyone wants to own, it is an asset that people have to own. It is a regressive tax that destroys the planet and subsidizes our roads that flee our neighbors. It''s a dangerous, costly, and dangerous hunk of metal that remains unused in an expensive garage almost 100% of the time.
Cars are suck.
And making them self-driving solves almost none of these problems. Thats the problem. When we spend too much time focusing on the quasi-mythical state of full automation, we ignore the devastating consequences that await us. Uber was successful because you could call a car with the press of a button. Leases are successful, despite the cost, because it is a different control pane for the vehicle. These are new transit experiences.
So, wheres the actual opportunity?
I think that business like Zoox have an interesting and compelling thesis. Despite their focus on the rider experience and critically by putting a powerful interface for teleguidance, I think it is important to realize that their teleguidance system isn''t a temporary bridge to get from here to there. However, the teleguidance system and its supporting architecture are arguably a more defensible breakthrough for them than any algorithmic advantage. That, combined with a model that eliminates ownership
Don''t be disseminated.
I haven''t used Zoox''s teleguidance system. I cannot believe that it is more effective than driving, but they are at least pointed in the correct direction. There is no need to think about all the messy intermediate states when level 5 is always right around the corner. The truth is that those messy intermediate states are the whole point.
The first part of the problem with self-driving automobiles is that it''s all there.
If you are an investor looking for a new company to handle self-driving automobiles, you are barking up the wrong tree. The winner is the company that can actually improve unit economics on the use of a vehicle. Until we solve that problem, all of the closed track demos and all of the world vanity metrics means nothing. Were imagining a race when we haven''t even decided how to take the first step.
The obstacle is not machine learning.
The user experience.
Slater Victoroff is the founder and CEO of Indico Data.