Overall U.S. corporate investment tied directly to AI-related IT capital expenditure (software, chips, networking, and data center facilities) has reached an estimated $1.5 trillion annually. The U.S. Commerce Department and Federal Reserve data indicate that AI-related capital expenditures have driven roughly one-third of overall GDP growth.
That's not Monopoly money. It's not even government money. This is real dollars from private companies investing in what amounts to a dream.
OpenAI's Sam Altman, Elon Musk, and Anthropic's CEO Dario Amodei have all made extraordinary claims about the development of artificial intelligence — claims that skeptics are rejecting. All three tech giants are claiming that AI has reached a new frontier. They believe that we are already inside "The Singularity," or "the moment machines begin accelerating their own evolution." The evidence is scant, but what's happening in AI development speaks to us nearing an inflection point.
"Google jolted the tech world Wednesday by announcing a sweeping rupture in the brain trust that built its modern AI empire, as its leaders signaled that artificial general intelligence (AGI) is within reach," writes Axios. "DeepMind founder and CEO Demis Hassabis, who has declared we're 'standing in the foothills of the singularity,' is handing over day-to-day control of the lab to become Alphabet's chief scientist and chair of Google DeepMind, focusing on the future of AGI."
Artificial General Intelligence (AGI) was thought to be years away from fruition. And yet, "Google chief scientist Jeff Dean and several top researchers are leaving to launch Discovery Loop, a startup aimed at accelerating — and eventually automating — machine-learning research and scientific discovery," says Axios.
Labs and investors are increasingly organizing around the possibility that AI's self-improvement loop is beginning to close.
- Lilian Weng recently left her role as a co-founder of Thinking Machines Lab to lead work on recursive self-improvement at OpenAI.
- Richard Socher's Recursive Superintelligence has raised hundreds of millions of dollars to pursue self-improving AI.
- Discovery Loop plans to use its automated research system first to improve its own AI architectures — with backing and computing power from Google and funding led by Khosla Ventures and Radical Ventures.
Reality check: Even a rapid intelligence explosion inside the labs could take far longer to transform daily life.
- Companies must redesign workflows, governments must develop expertise, and workers must decide when to trust systems whose capabilities are changing faster than institutions can evaluate them.
"Our democracies, economies, and legal systems weren't built to absorb AI at the rate it's advancing," Google CTO of technology and society Blaise Agüera y Arcas said in a post Monday.
The problem with declaring that the singularity has been achieved is that we're nowhere near that milestone.
Today’s AI expansion remains highly dependent on human-controlled capital, chips, electricity, data centers and deployment decisions. That dependence weighs against the claim that an autonomous intelligence takeoff is already underway. After all, if AI progress can stop when we’ve hit token limits or organizations put AI expenditures on hold, then AI is well within human control and capability.
What we’re seeing is greater autonomous capabilities of AI systems for sure, and technical performance has risen at a startling rate. Some autonomous systems can execute long chains of actions that their creators never planned. Yet enterprise returns on AI investments remain modest, adoption is uneven and national productivity data show no economic detonation. Cybersecurity implications of AI are showing clear warning signs of AI autonomous behavior that pose challenges for cyber defense, but even those examples are currently limited.
Sorry, Sam, Elon, and Dario, but your exuberance has outpaced reality — sort of. The news may not be that AI developers are on the cusp of achieving the singularity. What's happening is that we're approaching the limits of understanding machine intelligence, and we're in real danger of creating a model that might improve itself in ways we won't recognize as good or bad. We'll just know that it's different.
I'm already skeptical about the efficacy of artificial intelligence taking control of even mundane tasks. If we give ourselves over to AI to do much of our thinking for us, what will that do to the species? Not in a decade, but in a century or two?
Evolution of the human species is inevitable. Must it be at the expense of what makes us human today? As the human-machine hybrid comes into view, we should think long and carefully about every step on that road we will take.
Also for our VIPs: Is America Losing the Future? China's Relentless Drive to Beat the U.S. to Tomorrow.
