Panel Warns House Lawmakers About AI Competitiveness

Experts agreed with Rep. Randy Weber, R-Texas, Tuesday that the U.S. would be in a “very bad spot” if American artificial intelligence competitiveness lagged behind China, the EU and others, during a Research and Technology Subcommittee hearing. House Science, Space and Technology Committee Chairman Lamar Smith, R-Texas, said China is investing $7 billion in AI through 2030 and $10 billion in quantum research. The EU is planning a $24 billion public-private investment through 2020, he said. DOD’s “unclassified investment in AI was only $600 million in 2016, while federal spending on quantum totals about $250 million a year,” he said.

Smith listed AI benefits for cybersecurity and urged full committee support for his National Quantum Initiative Act to establish a federal program for quantum R&D. The committee is expecting to mark up that bill Wednesday, along with a bill reauthorizing the National Institute of Standards and Technology, directing investment in development of AI and data science.

Research and Technology Subcommittee Chairwoman Barbara Comstock, R-Va., cited fears about artificial general intelligence technology capable of carrying out human tasks. AGI could cause humans to lose control over human-created machines, she said, noting executives like Elon Musk warned of that possibility. More immediate issues concern cybersecurity, privacy and the economy, she said. Comstock urged Congress to support the Artificial Intelligence Job Opportunities and Background Summary (AI Jobs) Act (HR-4829).

OpenAI co-founder Greg Brockman testified the U.S. will be in “a fog” for the next five years on AI, making it difficult to offer “confident projections.” His organization’s primary recommendation is to measure progress and understand how fast the field is moving.

GAO Chief Scientist Tim Persons testified the approach should be to augment the human workforce with AI offerings, not to replace humans. Brockman said analyzing AI is similar to how the world tried to envision the future of the internet in the 1950s -- but the timescale for AI will be much shorter. Malicious actors and technology vulnerabilities are risks, he said, and the U.S. needs to start thinking about it today.

AI4All co-founder Fei-Fei Li said a “diverse group of contributors” should be included in development of AI. Her organization’s mission is to give marginalized communities and students access to AI. A lack of diversity will lead to algorithmic bias, so lawmakers need to establish ethical, socially conscious policies, she said. Li urged government collaboration with academia because AI is “too important to be owned by private interests alone.”

Rep. Dan Lipinski, D-Ill., said opinions vastly differ about the extent of job loss due to AI. Brockman said the internet displaced workers, but it also created millions of jobs, and AI development will do the same. There's not enough information on AI, he said, so the impact is an “open question.”

AI Notebook

Developers of artificial intelligence should cross-reference privately held data sets with public information, perhaps from a national statistics bureau, to eliminate data bias, said Immuta and the Future of Privacy Forum in a paper Tuesday. It's intended to give lawyers, compliance personnel, data scientists and engineers a framework for deploying and maintaining machine learning systems.