Instructor: David Schuff, Section 003

What if AI Fails?

Image result for self-driving cars

Despite artificial intelligence capabilities dating back to the 1940s, we still know very little about these technologies. Of course, AI from back then is far different than what we refer to as AI now; think IBM Watson, self-driving cars, etc. AI has certainly come a long way, but our understanding of it has lapsed. In fact, we know so little that we’re not quite sure how to fix it when things go wrong. When an AI technology needs debugging, technologists aren’t quite sure what to do. After all, “debugging is based on understanding,” so to debug something one doesn’t truly understand, is fairly impossible.

To complicate the issue even further, bugs in AI technologies may not make themselves apparent until it’s already done damage. With AI infiltrating our lives at a rapid pace and in drastic ways, its important that we identify bugs and fix them quickly to avoid harm. Imagine a self-driving car that had an unidentified or misunderstood bug, which cause the car to crash. It is imperative that we work on this debugging issue before AI is allowed to make fatal mistakes. How cautious should we be with AI in the meantime? Do you agree that, as some are saying, we should halt the use of autonomous cars after Uber’s crash in Arizona?

 

Source

 

3 Responses to What if AI Fails?

  • I think that you bring some very good points since AI at the moment is still human code that just contains a bunch of different decision trees and conditional statements. However, it’s also valuable to point out that Machine learning and Deep learning are still at this early stages of development which will minimize the errors that AI make as long as we feed it enough data, which we have and will get more as Internet of Things at home, work, and other places thrives.

  • This is a very interesting point to highlight considering how popular AI has become over the past few years. One of the most popular uses of AI is being used in the car industry. Car makers today are trying to implement autonomous driving in the cars they offer. Companies like Tesla and Uber have either implemented or tested cars with autonomous driving. Yet, these companies have recently been in the news due to AI failing and causing fatal accidents. Early last month an Uber self-driving car in Arizona struck, and killed a pedestrian crossing the street. Uber announced following the accident they were discontinuing their self-driving program. A tesla driver was killed this past weekend when autopilot failed to identify lines on the road. This isn’t the first deadly accident for Tesla cars. AI is a great innovation, but with advancements like autonomous driving there is no room for errors.

  • It is indeed frightening to think about the world we are creating as we push AI solutions across industries into mainstream society. Some would even say that the world is slowly being taken over by robots! Any time new technology hits the market, technologists are stuck solving problems for longer periods of time than they typically would, given the complexity and lack of documentation about the technology’s functions. AI is no different, especially since its built to learn and grow on its own based on its understanding of human language. The perfectly programmed AI wouldn’t have any bugs, it would grow and develop like a human would, learning and correcting itself as it ages. Self-driving cars are newer AI ventures, whereas many people have incorporated different types of AI in to their every day lives. Look at Siri or Alexa, for example. I can’t tell you the last time I manually set a reminder for myself in my phone. The minute one becomes cautious innovating with AI, the quicker that somebody else will sprint ahead of them to develop the technology of tomorrow. With regards to the Uber crash in Arizona – we have more to gain from falling and getting back on our feet than to have never fell at all.

Leave a Reply to Tymofiy Rabchuk Cancel reply

Your email address will not be published. Required fields are marked *

Where and when do we meet?
Alter Hall 603
12:30-1:50 Tuesdays and Thursdays
Office Hours
Tuesdays and Thursdays, 2:00-3:00 PM