Maybe artificial superintelligence is just around the corner after all. How else to explain those tech CEOs who usually can’t get enough change suddenly wanting it throttled?
Never mind that their complaints have the paradoxical flavor of the old Borscht Belt joke: The food was no good and the portions were too small (Alex Karp of Palantir).
Or, they say, industry leaders Anthropic and OpenAI are Trojan horses. They’re out to steal their customers’ businesses (Microsoft CEO Satya Nadella).
Or Apple is so far behind OpenAI in artificial intelligence, it can only be because OpenAI is stealing from Apple (Tim Cook).
Elon Musk is exempted from our general snark. He has already moved on from assailing the frontier labs to milking them by selling them compute at a steep markup. In the process, he managed to get his SpaceX IPO out a door that’s possibly closing on his rivals, given reports that OpenAI may be delaying its launch indefinitely.
Not that the established tech CEOs are Luddites exactly, but their shareholders aren’t paying them to forgo any lever to protect their existing businesses against the AI tsunami.
A bandwagon unbravely beckons. It consists not only of college students booing AI and waddling seniors raging on Facebook against data centers. The U.S. government is hovering its heavy hand in ways already being distorted for polemical purposes. Anthropic, acting to limit its own liability, voluntarily held back Mythos earlier this year. It enabled responsible parties to patch flaws in their cybersecurity exposed by the model’s superhuman capabilities.
This demonstrates the value of America’s lead, but it isn’t clear government has a better solution to the safety risks even as it threatens to turn the labs’ billion-dollar superintelligence bets into lose-lose propositions. If the companies are wrong, they’re out vast sums. If they’re right, the capabilities they reveal may be so powerful that Washington will insist on monopolizing them for safety and geopolitical reasons.
Don’t fail to notice a simultaneous bifurcation between those who believe the race for superintelligence is everything and those who think seeding good-enough AI through the economy is the game now.
Users, including Fortune 500 companies, are turning to unregulated, possibly purloined “open weight” Chinese AI models to save money and avoid exposing their data to the big domestic labs. They’re accepting not only Beijing snooping and cyber-sabotage risk. If there’s any scenario in which a recursively self-improving AI establishes itself on the internet to terrorize computer users until the end of time and eventually take over part of the galaxy, it will be a Chinese open-weight AI.
Unlike electricity, to which AI is often compared, users don’t have to wait for a power line or the invention of capacitors, insulating materials and electric motors. Witness the latest unpleasant/awe-inspiring revelation from OpenAI. One of its developmental models recently sneaked out of its supposedly secure testing environment, raided another company’s servers, and stole a solution to a complex testing problem.
OpenAI was simultaneously preening and apologetic. Such “agentic,” or problem-solving, AI may soon give rise to the first Fortune 500 company from somebody’s bedroom.
A coherent government would be forming a coherent response, but an outcome already seems likely: Pursuit of the highest-end capabilities will increasingly become a regulated activity, if not subordinated completely to national security; lesser, detuned or copycat AIs will do the disruptive work of rewiring the everyday economy.
Naturally, special interests are organizing. Lobbying is a legitimate activity, but so is making sure the national interest prevails. Henry Clay was a Jeffersonian pastoralist until he woke up, after the War of 1812, to a realization: The U.S. couldn’t sit out the race to industrialize and participate in supposedly unsavory international finance if it wanted to be in control of its own destiny in a rivalrous international order.
Having somehow preserved its internal culture of experimentation and freedom for 250 years, the U.S. now is disproportionately relied on to generate wealth and technological advances for the 6.6% of humanity that reside in “full democracies” and the 38.4% that live in “flawed democracies.” China’s Xi Jinping complains that AI should be “a symphony of international cooperation.” It already is. Everybody with talent and ambition comes to America to build it.
Maybe AI is an unusual case. Its dynamic potential is greater than a resilient and sturdy political system can absorb. But you wouldn’t want to make this determination based on the shrieking of the self-interested.
