Charting a policy course for America’s artificial intelligence future

Last week, members of Congress and their staffs saw firsthand how artificial intelligence can have a positive impact on society. During CES on the Hill, companies from across the tech industry showed off the latest trends that will transform our world. But IBM was not there with a shiny new widget. We had the pleasure of bringing Kate, a 14-month old black Labrador Retriever who is training to become a guide dog with an assist from our Watson AI. By gathering and analyzing data about Kate’s reactions to stimuli and her environment, Watson can help her trainers from Guiding Eyes for the Blind determine sooner and with greater accuracy how likely she is to complete her training successfully and, most importantly, be paired with a visually impaired individual. Kate, and IBM’s work with Guiding Eyes, is an example of the incredible potential AI has to do good in the world.

Realizing that potential and advancing America’s leadership in AI requires partnership, like the one between people and AI that contributed to Kate’s training. By enacting a number of smart policies that promote partnerships, inclusivity, and openness, Congress can advance AI innovation and expand its economic benefits.

{mosads}Since the early days of machine learning, the U.S. has been a leader in cutting edge AI research. Both the National Science Foundation and Department of Energy have played central roles in that leadership, and in facilitating strong collaboration with industry and academia. Congress should increase research funding focused on effective methods of human-AI collaboration as well as AI explainability. AI testbeds are also needed to bring together diverse stakeholders including government, industry, academia and impacted communities to test new AI innovations in controlled environments.

AI research partnerships, and partnership between data and AI innovation, can solve some of humanity’s most pervasive problems. For example, farmers have long had to balance complex and potentially costly decisions on when to plant and how to manage their crops while food producers are striving to boost food quality and sustainability. AI, including tools that IBM has created, can help food producers make more informed decisions by unlocking insights in data from farm equipment, Internet of Things (IoT) sensors, crop images and even the weather.

But AI we rely on for these and other decisions cannot be biased, nor should technological advances leave people behind. Inclusivity must be a priority for both industry and policymakers. Thus, government should award grants to researchers that create diverse and inclusive developer teams who are trained to identify and mitigate possible sources of bias in data or AI models. Policymakers also should expand access to digital skills education and workforce training, preparing students and mid-career professionals for success in AI-related fields. Priority must continue to be placed on training more Americans for new collar jobs which require in-demand skills but not always a traditional bachelor’s degree.

AI policy must also foster openness. Policymakers should implement open government data and make more large, diverse, and non-sensitive government data sets available in machine readable formats, governed by standard data sharing and licensing models, so that it can be used in training responsible AI systems. This also will help ensure that robust training data isn’t the exclusive purview of data monopolists.

Lastly, AI policy should recognize and support the role that open source can and should play in the advancement of the technology. Government should encourage the use of tools that can help developers and users assess and mitigate bias and provide explainability. In the private sector, we have seen how such explainable AI can be used to great effect: from enhancing regulatory compliance in financial institutions to matching patients with clinical trials and, yes, improving training for guide dogs. AI can be an equally powerful tool for effective, efficient governing, and we encourage the government to accelerate its own use of AI.

We brought Kate and her trainer to Capitol Hill last week to engage Congress in a conversation about the positive power of AI. We will continue that dialogue this week, as more than 100 IBM leaders from across the country travel to Washington for our annual congressional fly-in.

As our lawmakers explore policies for the future of AI, we urge them to draw on the considerable expertise of industry, to build understanding of the technology and its potential, and to pursue policy that promotes partnership and innovation while strengthening trust.

Christopher Padilla is Vice President, IBM Government and Regulatory Affairs.