When you hear the words “artificial intelligence,” images of Skynet from the Terminator movies might pop in your mind. Although this scenario makes for good cinema, it is far from the reality of how artificial intelligence operates. Artificial intelligence is already present in our daily lives through our smartphones, Internet-connected devices, voice assistants like Alexa and Siri, medical diagnostics, and others. But what is artificial intelligence and does its potential to augment human capabilities outweigh the threat of world takeover – or the risks of a malfunctioning computer system?

What is Artificial Intelligence (AI)?

AI refers to computer systems that process data through a series of instructions (or algorithms) developed either by a human or a machine. We use the word intelligence because these systems seem to operate like a human would to solve problems in real-world situations, as long as these situations are known to the system.

AI can be narrow and highly adaptable, or general purpose, and uses specific methods like machine learning, deep learning, neural networks, image analysis, language processing, robotics, and autonomous vehicles, to name a few. This type of AI allows a machine to make a decision by following a process and determining an outcome within a known environment. For example, if you tell your voice assistant on your phone to play your favorite song, it will analyze your music tracks and play the most frequent. It can make this decision even though there is not a preset outcome paired with the input. Alternatively, when confronted with factors that are not part of a known environment, outcomes can become erratic. The reason there have been problems with autonomous vehicle testing is that autonomous vehicles have learned how to perform in a controlled environment. When anomalous actors, namely animals and humans, are encountered the vehicle does its best to classify the actor but cannot always do so accurately, resulting in an inappropriate response, like accelerating when the vehicle should brake.

General purpose AI does not yet exist in the way we conceptualize. If conceived, it will allow machines to see data and not just process it; to understand unknown data and environments without prior programming. This type of processing occurs in our brains automatically every day. What takes a human one or two examples to learn or understand takes an intelligent machine thousands of iterations to familiarize. This type of machine cognition is what many fear could lead to a robot takeover. As Dr. Joseph Sirosh, former Chief Technology Officer for AI at Microsoft has stated, machines do not have a survival instinct.

All AI is Dumb, But Some AI is Useful

Dr. Joseph Sirosh stated that all AI is dumb, but some AI is useful. Algorithms can be programmed, and machines can learn to perform tasks that augment human capabilities without a full picture to decide the outcome. Is there a risk of this technology replacing humans? Most surveyed experts agree that AI will not displace more jobs than it creates. But AI advancements bring unique concerns. The most recognized is the protection of personal privacy. Training AI algorithms has typically required huge datasets which are populated with human generated data points, although they are often anonymized. Subsets of AI applications, like facial and voice recognition, raise privacy concerns when the images and voice patterns are collected by the government. Deepfakes also pose a threat if an individual is misidentified or presented in an illicit manner that cannot otherwise be verified as false. These considerations must be balanced with the preservation of national security.

Because data is a vital component to training AI algorithms, the way in which AI systems learn and process data, bias often emerges without intent. Unintentional bias results most often because a data point only presents one characterization of a person, object, or situation at a given point in time. If the data inputs do not reflect the full environment a biased outcome may be produced, depending on the learned process.

 These challenges should not discourage the development of AI. Machines that can act without direct user input can enter situations that are unsafe for humans, such as disaster sites, environmental hazards, and even outer space. AI is faster than humans and can augment human productivity by performing processes within a shortened timeframe. AI has also been used to aid in educational institutions and can provide better and more realistic training environments, for students, scientists, and practitioners.

Robot Takeover?

While the federal government is working to advance AI in government use and the private sector, do not be worried there will be a robot takeover. In February 2019, President Trump issued an Executive Order “Maintaining American Leadership in Artificial Intelligence” to set a coordinated federal government strategy for AI development and deployment. The National Science and Technology Council established a Select Committee on AI and an AI Interagency Working Group. The Fiscal Year 2019 National Defense Authorization Act established a National Security Commission on Artificial Intelligence and the Department of Defense has a Joint Artificial Intelligence Center. The United States also plans to participate in the Global Partnership on Artificial Intelligence and adopted the Organization for Economic Cooperation and Development Principles on Artificial Intelligence.

I recently introduced a bill with Rep. McNerney to require the Consumer Product 

Safety Commission to establish a pilot program that will utilize AI to better track and identify consumer product hazards and retail trends, including the resale of recalled or harmful consumer products, and help track illegal imports. The program may also help predict fail rates and identify problems in consumer products before they can significantly impact the market. While this is an initial implementation for one federal agency, all government functions could benefit from AI capabilities.

This investment is necessary to maintain American leadership in setting international standards for AI, stay ahead of China and other global competitors, and fully harness the power of AI systems. AI is a large and complex field, and I look forward to the advancements released by American ambition and ingenuity.