Venture Bytes #50: Are We There Yet?

Are We There Yet?

Fully autonomous vehicles are not yet ready for prime time and will not be ready in the forseeable future, all the hype notwithstanding. Autonomous vehicle (AV) technology has been around for almost a decade, but has not developed enough to be a viable commercial application.

The software framework of AV technology is based on deep-learning algorithms which processes the vehicle’s surroundings. This may produce erroneous results when presented with a brand new circumstance, since it lacks the ability to generalize similar, but nonidentical, past situations. The algorithms require that every single possible situation the autonomous car could encounter would have to be introduced to the algorithm multiple times. The possible inputs are just too many.

The vast amount of possible inputs needed raises an important question. Do we really need to try every possible situation a vehicle could encounter, or are some situations not frequent enough to be relevant for the development of AV software? In theory, the technology does not require the testing of every possible situation. To ensure an absolutely safe autonomous vehicle, however, the algorithm has to be stress tested for every potential road situation. That includes situations that may be considered uncommon, such as somebody walking and carrying a bike in the middle of the road. In uncommon cases such as this and many more, prior algorithm exposure would prepare a vehicle to prevent an accident. It is this dynamic - the continuous testing of countless possible events - that necessitates patience for the full rollout of AV technology. Algorithms are still adapting to the many seemingly random events that can occur on a road with human drivers.

We not only need to worry about the optimization of algorithms, but in the optimization of vehicle fueling systems. Autonomous technology significantly increases the car energy consumption. Autonomous electric vehicles (AV-EV) have been promised by technology companies and the automobile industry for years, but realizing that goal is proving to be challenging and elusive. The industry is focused on developing a fully functional AV-EV car, but such a car would require much more frequent recharging than a regular electric vehicle. One option to overcome this problem is to increase investment in battery technology or in autonomous driving software, but neither seems to be happening. As a result, many companies have settled on hybrid autonomous cars to enable vehicles to operate longer and with more range, but with higher gas consumption than a conventional car.

Then there is the regulatory angle. The industry may face regulatory headwinds due to safety concerns. As of today, many states are interested in AVs given their potential to be much safer than human-driven cars. But if incidents of collusion become noticeable, the public will demand stricter regulations. For instance, following Uber’s fatal accident last March, the company was forced to stop road testing for nine months as regulators scrutinized the technology more closely. Unfortunately, the chances of these kinds of accidents occurring are high during the algorithm-optimization period, inviting strong legislative pushback.

There is no doubt that autonomous cars will one day replace the need for human-driven cars given the safety factor. Until then, which could be anywhere from 5 to 25 years depending on whom you ask, all the pieces of the puzzle still need to come together. **

AI, ML and the Ripple Effect

Automation is at the center of current innovations. Cashierless checkouts, Amazon’s delivery robot Scout, and Boeing’s automated flying cars are all recent examples of how automation is being integrated into the real world. While these automated technologies are applied in various areas, the driver behind these technologies is all the same: the implementation of artificial intelligence (AI) and machine learning (ML). Cashiers, robots, and cars must be intelligent to some degree in order to be automated.

Artificial intelligence refers to the ability of a computer to perform complex, intellectual tasks such as driving or performing deliveries. Machine learning is a subset of artificial intelligence that deals with computers recognizing and solving problems as well as identifying patterns. As computers and machines act more and more like people, people run the risk of losing their jobs to AI and machine learning-driven technologies. According to a report by a think tank called Centre for Cities, nearly one in five jobs could be replaced by AI and machine learning driven technologies by 2030.

While this has yet to happen on a wide scale, the trajectory of the taxi industry serves as a possible case study. Uber, Lyft and other ridesharing companies have put immense downward pressure on the taxi industry’s market share in ground transportation services. According to Certify, taxis market share in ground transportation services has dropped from 37% to 6%, displacing thousands of taxi drivers from work. While these old taxi jobs are being replaced by gig economy driving opportunities, most ridesharing companies are invested in AI and machine learning through the research and development of automated cars. With the implementation of self-driving cars, the ground transportation services could potentially be entirely automated, eradicating all these drivers’ jobs thanks to AI and machine learning.

The implementation of AI and machine learning technologies have the potential to be widespread and to penetrate multiple industries. In finance, algorithmic trading, market analysis, and portfolio management could all be done by a computer. Education has the potential to be revolutionized by AI-driven tutors, student assistants, or even teachers. AI and machine learning could even displace healthcare workers through automated diagnostic and surgical processes. The total amount of jobs displaced by AI and machine learning could certainly be enormous and no industry is safe.

According to the Centre for Cities report, though, AI and machine learning will create as many, if not more jobs than they displace. However, there are less concrete examples of the new jobs these technological changes will facilitate. There will definitely need to be more computer scientists, programmers, and robotic experts, but definitely not in the same magnitude as the number of professionals potentially displaced by AI and machine learning.

While these technological changes may facilitate job growth in an unexpected, yet to be determined way, there will likely be an adjustment phase. Surely, new types of education and skillsets will be more valued with the implementation of automated technologies. It will take time for the workforce to react and adjust to these newly demanded skills.

Furthermore, the implementation of such technologies may promote economic inequality; those who own the technology, machines, and automated processes will have a greater control over the economy and potentially over the job market.

While AI and machine learning may promote technologic growth and innovation, the implementation of these technologies may have significant economic repercussions. While we should undoubtedly continue to innovate, we should continue to be cognizant and aware of the repercussions of these new, innovative technologies. **