White House Chronicle

News Analysis With a Sense of Humor

The Question That Must Be Answered: How Dangerous Is AI?

by Leave a Comment

Races have winners and losers — at least so I have always thought. At some point, the competition ends and a winner is identified, whether it is first-past-the-post as in a horse race, or first in the marketplace as in a commercial race. Think about VHS and Betamax.

Similarly, we knew who won the space race and many other international races. However, the race we are engaged in with China over artificial intelligence isn’t defined — there are no winning posts, and we don’t even know what winning will look like.

Will we wake up one day and be told we have cleared the winning post and China is in the dust, or the other way round?

We don’t even seem to be engaged in the same race as China. Is it the one President Trump talks about where we are going to out-invest, out-talent, out-research and dominate AI?

Or is it the one China appears to be running, emphasizing using AI as it is now, with the construction of many thousands of deployable robots?

Is it a purely commercial race or a purely military one for both countries?

So far, it seems, we have taken the high road, with proprietary systems, and China has taken the low road with widely deployed, open-source, ready-now systems.

China is aggressively spreading AI across its commercial operations and developing families of robots. The United States is still nibbling at universal deployment. China is mass-producing droves of humanoids that dance, leap in the air, and perform complex duties like cooking.

They entertain and threaten. They are a new workforce, presenting a new level of competition.

Although the United States has reached new heights in invention and pushed the frontiers of AI discovery, it lags in deployment. It has been slower in harvesting the technology.

Now comes the unsettling — actually terrifying — warning from AI researchers that the human race may face extinction as soon as the end of this decade. Three AI bosses — Dario Amodei of Anthropic, Sam Altman of OpenAI and Elon Musk of Grok — have called for a slowdown and regulatory oversight.

This warning deserves to be treated as truth until proven otherwise.

Could, as the doomsters fear, no bells ring, no shouts of joy echo from the bleachers at the race’s end because AI has killed us all short of the winning post.

It is a strange race run not by the government but by corporations, with the government cheering from the sidelines. Bill Gates has said the thing about AI is that — unlike other dangerous technologies, such as nuclear — the government has no depth of knowledge, no deep bench. Where AI is concerned, the government is an outsider.

When my television crew and I covered the Upscale conference in Malaga, Spain, last year, we got insight into people from around the world surging ahead in the creative uses of AI, without regard to whether they were using Chinese or American AI. For this cohort, these were just off-the-shelf tools.

This was a new generation of creatives using AI with relish to create fine art, advertisements and designs for everything from a new shape for a car or a wild-patterned shoe to a more comfy living room couch.

I had to wonder whether they weren’t the real winners in the AI race: Those who use it without scaling the heights of its technical possibilities or using it to displace truck drivers.

Only two countries are in that race, the United States and China, but all have a stake. In AI deployment, the United States is far from dominant. Indeed, it is down the list of countries where AI is already incorporated into daily commercial operations. Estonia and the UAE stand out.

The United States is challenging the future before exploiting the present.

It is easy to declare the need for regulation and guardrails, but hard to do that when the subject is in constant transmutation, changing daily and likely to change again tomorrow.

I would propose that the Trump administration appoint a small, nimble, non-industry expert panel to answer just three questions: Is AI an existential threat to the human species? What should be done about that? And how much time do we have?

The threat has been debated academically since 1958 — the heftiest warning came from British astrophysicist Stephen Hawking in December 2014.

Warnings from within should be heeded. J. Robert Oppenheimer warned about nuclear weapons. His words still govern nuclear non-proliferation policy and security settings.

The creators of an even more dangerous technology are warning, much as Oppenheimer did.

The Shady, Sometimes Wacky World of State Secrets and Security Clearances

by Leave a Comment

Beware: Classified documents don’t always hide state secrets, and security clearances are used as tools of manipulation and vengeance.

Before Xerox, if you wanted to keep a copy of something, you had to type it with a carbon sheet backing every page.

In 1969, I was commissioned by a long-gone consultancy, the Arctic Company, to write a paper on the use of hovercraft by the military, especially the infantry.

They were offering $500 for the job and, like most reporters, I was keen for the income, so I signed up.

It was a time when it was believed that hovercraft — vehicles that cover the ground on a cushion of air — would be widely deployed.

I had no great insight into the vehicles or how they might be used as chariots of war. But I did have a lively imagination and access to The Washington Post library. I gorged on newspaper clippings and wrote my commissioned piece.

After it had been accepted, and I was told by the company that the army was “very pleased” with it, I forgot about it.

Then someone unrelated asked if they could see it out of curiosity. I said I didn’t have a copy, but I had been told that it had been mimeographed and widely distributed in the Pentagon.

I asked the Arctic Company for a copy, and they referred me to the appropriate office in the Pentagon. I was rebuffed. They said that it was classified and I could only see it if I had security clearance.

The Joint Committee on Atomic Energy, which controlled the nuclear establishment, military and civilian, used classification and security clearances to keep other members of Congress and the press out of its business; it regarded itself as the only responsible custodian of the nation’s nuclear secrets.

I was told that they were so classification-obsessed they couldn’t discuss the contents of the papers they had assembled to discuss because they were marked “Eyes Only.”

When James Schlesinger became chairman of the Atomic Energy Commission in August 1971, he set about overhauling the classification of documents.

I was close to Schlesinger, and he told me that he thought more than half of the AEC documents shouldn’t be classified, and he set about declassifying them. His argument: If you classify the trivial, all classification is degraded.

Dixy Lee Ray, the last chair of the AEC, became a friend of mine. I invited her to dinner at the venerable Red Fox Inn & Tavern in Middleburg, Virginia, established in 1728. It is a pleasant place to dine and claims to be the oldest continuously operating inn in America.

Ray went everywhere with her two dogs (Ghillie, a Scottish Deerhound, and Jacques, a miniature poodle), and they were in her limousine wherever she went. The car also contained — as I am sure the secretary of energy’s car does today — the hotline that would be part of the launch procedure, in the event a nuclear attack was ordered by the president.

In her briefcase, Ray had an innocuous study she had wanted to give to me.

It was a blustery night, and her driver was waiting in the car in the parking lot with her briefcase on the back seat and both dogs on the front seat.

The moment Ray opened the rear door, two things happened: A great gust of wind arose and Ghillie leapt from the front seat to the back seat, upsetting the briefcase. Crisis!

All the papers in the briefcase, many of them marked with the big red X of classified documents, blew all over the parking lot.

The three of us, in panic mode, set about scouring the bushes for them in the dark, fearing that someone would find one of them and, so to speak, the jig would be up. We could imagine the headlines.

After an hour’s search, we figured we had gathered all the papers, and Ray did an inventory. Nonetheless, the next morning I drove out from Washington to make sure no nuclear secret was impaled on a bush branch.

From the time when J. Robert Oppenheimer lost his security clearance under murky circumstances, these have been used as a tool of manipulation and vengeance.

If a scientist or a manager loses their clearance, they can appeal in a long, difficult and expensive process. Even if the victim appeals, the damage is done; the subject is damaged goods, publicly humiliated as morally deficient and untrustworthy.