Every White House leadership change causes speculation about what pre-existing initiatives the incoming administration will embrace or eliminate. I encourage President-elect Trump and his appointees to start their term ready to ensure that artificial intelligence (AI) gives our economy the competitive edge it needs.

The Obama administration recently released its recommended approach for how the U.S. should promote AI research and development. It published balanced suggestions to guide public investment, encourage private-public collaboration, and account for national security, public safety, diversity, and ethics. It was a good start. But it’s broad and not yet deep.

I’d like to offer a few specific suggestions for the next administration.

Make America first again

Our past successes are no guarantee of America’s future innovation leadership. It’s easy to take for granted that we led the world into space, “smart weapons,” the internet, and the cloud. It’s not as dramatic as putting a man on the moon, but make no mistake: AI will dominate the next wave of economic growth and national security. U.S. economic growth has been slow and unevenly distributed, so it’s imperative that our federal government play a leadership role.

  1. Invest 10X. Total federal investment in unclassified AI R&D in 2016 will be an estimated $1.2 billion. That is less than IBM will invest in AI this year, and less than 1 percent of total federal R&D. Trump wants to stimulate the economy. No dollars will create larger economic multipliers than those invested in AI.
  2. Centralize and drive accountability. How do you increase your investment tenfold without it being frittered away among dozens of bureaucracies? Centralize AI under one team of experts accountable to U.S. taxpayers. I nominate the Defense Advanced Research Projects Agency (DARPA), which brought us the internet, led the way with autonomous vehicles, and undoubtedly is driving classified AI investments.
  3. Compete globally. The U.S. needs to own and act on the best new intellectual property we can muster, but it would be naive to think we will create it all. DARPA should expand its successful “Grand Challenge” model to welcome the world’s best researchers to compete across fields that will be transformed in the next decade by AI.

Harness entrepreneurs to serve our country

The world’s largest technology companies are all-in on AI, investing billions. They have the heft to ensure they have a seat at the right tables in Washington — but they’re also the incumbents. Most of the disruptive AI innovations will come from startups.

  1. Engage innovators. Defense Secretary Ash Carter created the Defense Digital Service as a “vestibule” to safely incorporate the best innovations from the private sector. The next administration should expand this unit so the Pentagon and Department of Homeland Security can incorporate the best and brightest, more quickly and easily. Cyberwarfare, autonomous drones, and systems for screening immigrants and visitors all badly need AI to keep our country safe. In addition, the Pentagon should bolster the Defense Innovation Unit Experimental (DIUX) efforts to harness the best of the startup world, including through venture-style investments.

Disruption is inevitable, but we can spread the gains

AI is inevitable. It will disrupt many jobs, and that’s a big problem. But, like ATMs created more banking jobs, so can AI — if we get out ahead of it.

  1. Start job retraining now. Many jobs in warehouses, factories, and transportation will be displaced. It will take hard work, ambitious investments, and private-public partnerships to make sure the people affected end up with better jobs, here at home. Trump has talked about making massive infrastructure investments. Under FDR, that meant bridges. Today that means TVA-scale retraining programs.
  2. Education for all. Take a cue from organizations like Code.org and spread the benefits of computer education broadly.
  3. Make health care smart. More than one in every six American dollars goes to health care. However, very few would argue our system works. Consider the Department of Veterans Affairs. The vast majority of their records are not digitized. AI can penetrate this unending data fog to help providers deliver the right care, more quickly and efficiently.

Build ethical machines

We are far from creating sentient machines, but ethical questions are already pressing. Take autonomous vehicles, which must decide in a split second: Do you hit the kid or the curb?

  1. Teach ethics. We’re teaching computers to see, listen, talk, and think like us. But who is “us”? To see the dangerous social consequences of getting AI wrong, look no further than the bias against black Americans Pro Publica uncovered in software that predicts criminal recidivism. We should encourage all schools to incorporate ethics into computer science, engineering, and math courses by publishing optional curricula. The administration should require that recipients of federal AI funding undergo bias awareness training.

The only thing we have to fear … is still fear

We need to debunk the notion that it’s computers versus people.The combination of human and computational intelligence can serve as a powerful engine for progress.

  1. Make humans super. As Accenture CTO Paul Daugherty recently quipped, “Our goal with AI is not to make super humans, it’s to make humans super.” Trump and his advisers have the opportunity to inspire our nation to apply AI to cure cancer, keep our country safe, and make our workers the world’s most productive. AI is a race, and make no mistake: China is in it to win.

The AI revolution is well under way. Personally, I am bullish about its implications — even as I look plainly at its unknowns, challenges, and threats. I hope our President-elect is as well, and that his administration moves proactively so that our United States leads that revolution, for the good of all Americans.