There’s a bit of a gold rush feeling to today’s AI projects. Everyone is scrambling to develop killer apps. Michael Jordan, a professor at the University of California, Berkeley, takes a step back to point out what this frantic approach is missing in Artificial Intelligence — The Revolution Hasn’t Happened Yet.
As he sees it, there are two things that artificial intelligence is lacking that are critical for it to truly be successful.
The first is that artificial intelligence today lacks an engineering approach. Systems are being designed and developed in an ad-hoc manner. The risk is that, as in other engineering fields, a lack of understanding of how components work together will lead to dangerous failures.
In civil engineering, before the discipline had a solid knowledge base, this lack of understanding led to bridge collapses. The ramifications of AI failures are likely to be more subtle, such as the danger of machine learning applications that turn their developers’ biases into algorithms. This lack of an engineering approach means there are no principles of analysis and design and that we’re blind to major flaws.
This challenge is made worse by the second thing that is lacking: an understanding of what “artificial intelligence” really means and agreement on what is artificial intelligence. AI today is often built on machine learning supplemented by data science, but that focus on methods loses sight of the original goal of AI: intelligent systems that imitated human intelligence. Today we also need to add the idea of systems that augment human intelligence and systems that provide intelligent infrastructure via the internet of things. Jordan argues that only by understanding the broad scope of AI can we truly understand how to focus on achieving societal goals.
Read more of the professor’s perspective here.