The hearing’s only pleasant surprise was its bipartisan support. Senators from both sides of the aisle, along with Cruz, all took the expert panel’s testimony seriously. Granted, AI still has the the new-car smell of a nascent field with great potential, which could boost US labor productivity by 40%, Cruz said in his introductory remarks. Golden bullet it might seem, but even the current experiments using AI to assist or solely could take a chunk out of the 35,000 annual vehicular deaths, 94% of which are caused by human error, committee member Senator Gary Peters noted.
Artificial intelligence could save even more lives, said the hearing’s first witness, Microsoft Research Lab’s managing director Eric Horvitz. AI could sift through vast quantities of medical data and catch things human doctors miss, as IBM’s Watson did back in August when it identified a rare form of leukemia and saved a patient’s life.
When people think of the cost benefits of AI, they think of automation. But reducing death and debilitating injury affects the overall economy, too: AI-assisted driving could also cut down on the 300,000 incapacitating vehicular injuries every year, which means more people remaining in the workforce and less time and money spent finding and training temporary or permanent replacements.
The looming fear over the hearing was China and India’s ever-greater competition in AI R&D. Logically, America’s lead on China and India could shrink simply due to how many more computer scientists they can train per their colossal populations. But letting US artificial intelligence slide could also be dangerous to national security. Back in August, the Defense Department suggested “immediate action” should be taken to boost development of AI war technology.
We can retain our lead to keep pioneering artificial intelligence by training America’s youth in AI programming as early as middle school, recommended the hearing’s second witness, Dean of the school of Computer Science at Carnegie Mellon University Andrew Moore. In his opinion, there’s a staggering amount of work and not enough trained computer scientists to perform it. Train a million middle school kids in AI, perhaps 1% stick with it, and even if you ended up with 400 experts at the level of Moore and his fellows at the hearing, there would still be too much work to do, Moore said. Pumping out more AI professionals won’t just be a smart move to fill a wanting workforce: for every programmer trained in artificial intelligence a tech company hires, Moore estimates, they earn $ 5 to $ 10 million more.
Collaboration could also help the US keep its lead, said the third witness, cofounder of the nonprofit OpenAI Greg Brockman. Making more AI systems open source drives innovation, Brockman said, along with unlocking datasets for anyone to use. But it’s not just amateurs and corporations working together: The tech industry, the government and academia should coordinate to establish standards of safety, security and ethics.
The last witness, senior research scientist at NASA’s Jet Propulsion Laboratory Steve Chien, noted that the space agency put an AI-controlled spacecraft in orbit to track earthbound phenomena — which has been continuously snapping photos from the high atmosphere for a dozen years. Many of NASA’s vehicles, including its Mars rovers, rely on AI to navigate and triage environmental conditions.
With technological possibilities come dangers, and AI is no exception. Cruz’s limp Skynet joke aside, the pressing concern with creating more complex and prevalent artificial intelligence is the subsequent increase in cyber vulnerabilities. We don’t have to look farther than the last year to see government and political agencies hacked by foreign independent and state agents.
But even things as mundane as liability could get in the way of AI application progress here in the US. The prospect of AI-controlled cars getting into collisions could lead to a legal impasse between carmakers, insurance companies and citizens as fault becomes uncertain. Public uncertainty or displeasure could derail AI implementation in those applications, too.
To avoid the US slipping out of first place in the AI race, the panel of witnesses ultimately recommended more investment and collaboration. That means far more emphasis on AI programming earlier in education, as Moore points out, but also simply more money injected into research: Government investment in AI over the past year was $ 1 billion, while the tech industry spent $ 8 billion, Brockman pointed out. That funding will likely help us make the roads safer and people healthier, but as Chien stated, it will also help us discover the deep space answers to a few questions that have bothered mankind for eons — namely, how did life form along with the universe around it?