Video thumbnail

In a first for NOW, the multitalented Ashley Llorens kicked off the 2018 conference with a killer hip-hop performance that had the crowd putting their hands in the air. While we all would have enjoyed a longer set from Llorens (who performs around the world as SoulStice), he soon moved his presentation to his “other” area of expertise: He leads the Intelligent Systems Center at the Johns Hopkins University Applied Physics Lab, where he spearheads research and development activities in machine learning, robotic and autonomous systems, and applied neuroscience.

Llorens took the audience through an exploration of artificial general intelligence, or AGI, which refers to a future state when machines are able to develop and evolve as human beings do. He believes that we are nearing an inflection point in the long path toward AGI, but we still have a long way to go in terms of functionality as well as societal trust in truly intelligent machines.

General intelligence, AI, Machine Learning and deep learning. Humans are the best example of what Llorens defined as general intelligence—nature has designed our function. From all of the inputs we take in from the world around us, the outputs of human general intelligence are actions directed toward basic goals, like survival and procreation, and more complex ones, like self-improvement and entertainment. Today, artificial intelligence (AI) is not general but specialized. With driverless cars and other AI systems, humans design the function according to a finite, anticipated set of scenarios. We see the limitations of specialized AI when the machine encounters a scenario that it is not specifically programmed to navigate.

In recent years, research at the cutting edge of AI has made some progress in breaking through these limitations. Machine learning is an emerging science in which people help machines design their own function, and machines can achieve an intelligence that is not specifically programmed. Llorens contends that we are far from machines achieving the type of general intelligence that humans possess, but he believes we are at an inflection point in the evolution of machine learning.

Basic functions of AGI. Llorens broke down the output of intelligence into four main functions: perceive, decide, act and team. He reported that we are making progress in all four areas, but that material limitations remain across all four.

In terms of perception, machines now perform speech and object recognition with lower error rates in many applications than humans. However, machines are still incapable of reasoning or intuition. In terms of decision-making, computers can now beat the best human players of complex strategy games like chess and go. These algorithms can now learn from self-play and are capable of adapting themselves to other games with different rules. Regarding action, machines still struggle to navigate the physical world, and regarding team-based or collaboration intelligence, machine-to-machine communication is still in its infancy.

Calibrating trust in intelligent systems. Even if and when we can close these gaps in AGI function, society still needs to get comfortable with trusting intelligent machines. Llorens highlighted four key areas of consideration. First, we need to gain comfort with evaluation and comparison—will we refuse to accept 15,000 fatalities per year from driverless cars, even if we know that human drivers kill 30,000 each year? Second, we need to build resiliency to adverse influences and find ways to prevent those who would seek to “fool” AI systems for their own benefit. Third, we need strong policy platforms so we can, for example, make smart decisions to use systems that we don’t fully understand if those systems work better than the systems we do understand. Finally, we need to align the goals of machines with our own goals; in other words, for widespread adoption, we need machines to do what we intend, not just what we explicitly ask for. While it may be many years before a robot can truly behave, act, sing and dance like a human, it is not too early to begin preparing ourselves for a world where we coexist with intelligent machines. 

 

 

 

The views expressed are those of Brown Advisory as of the date referenced and are subject to change at any time based on market or other conditions. These views are not intended to be and should not be relied upon as investment advice and are not intended to be a forecast of future events or a guarantee of future results. Past performance is not a guarantee of future performance and you may not get back the amount invested.

 

 

 

The information provided in this material is not intended to be and should not be considered to be a recommendation or suggestion to engage in or refrain from a particular course of action or to make or hold a particular investment or pursue a particular investment strategy, including whether or not to buy, sell, or hold any of the securities or asset classes mentioned. It should not be assumed that investments in such securities or asset classes have been or will be profitable. To the extent specific securities are mentioned, they have been selected by the author on an objective basis to illustrate views expressed in the commentary and do not represent all of the securities purchased, sold or recommended for advisory clients. The information contained herein has been prepared from sources believed reliable but is not guaranteed by us as to its timeliness or accuracy, and is not a complete summary or statement of all available data. This piece is intended solely for our clients and prospective clients, is for informational purposes only, and is not individually tailored for or directed to any particular client or prospective client.