The AI revolution may die aborning unless politicians can resist the temptation to apocalypsize everything about it.
I have been urging caution in the development of AI, even reluctantly coming to the conclusion that a slowdown, or even a pause in the creation of new AI models, may be necessary. Responsible tech leaders like Anthropic CEO Dario Amodei urged AI developers and industry leaders to slow the rapid pace of model releases and capabilities to allow safety evaluations, transparency measures, and government standards to catch up.
He proposed bringing in external evaluators and collaborating with government bodies to test model safety before publicly launching new frontier capabilities.
Amodei knows that if AI companies don't take similar steps on their own, the government will do it for them. Congress, led by the Luddite-in-Chief Sen. Bernie Sanders (D-Vt.), will write legislation that will strangle AI in the crib. Worse, federal agencies will take years to write a blizzard of regulations that could make AI far less of an effective tool for business and for the rest of us.
"A new Gallup poll shows the United States is in the bottom five of 37 countries in optimism that AI will mostly help their nation," reports The New York Sun. "In the US, only 36% believed AI would help the majority of people in their country."
"That places the United States on the level with nations like Afghanistan," the article adds.
As with anything else, AI will require trade-offs. We've heard a lot from those who see catastrophe with every new AI model, talking about AI in apocalyptic terms. While we should be uncompromising about safety, there is going to be an element of risk that we're going to have to accept. Not the sort of risk that OpenAI researcher Jacob Coxon wrote about when he stated flatly that many in the industry believe AI could “kill us all by the end of the decade" — it's the risk in how revolutionary we're going to allow AI to be.
Among the more tangible benefits of AI will be more products being made in the United States.
AI is “going to drive down the cost of everything: from agricultural goods to products that right now can only be affordably made overseas,” said Michael Brickman, Policy Director at the Cicero Institute.
“You’re going to start seeing a lot more made here because AI can make that more efficiently and that’s just going to need more jobs and cheaper goods for Americans. It is going to create many, many, many more jobs than it will ever replace," Brickman told the New York Sun.
“Imagine AI could prefill an application or just get you the information or the approval that you need so you don’t have to spend time with DMV,” Mr. Brickman said. “Or you don’t have to spend six months waiting for a permit so you can build a home.”
AI could revolutionize education, Mr. Brickman added, allowing every student regardless of economic background or school district to have an individualized, one-to-one AI tutor.
“If you can put these tools in the hands of parents and students you’ve given them a personalized tutor and the ability to learn anything they want in any way they want,” he said.
“It makes it a lot easier to create new schooling models that I think can actually accelerate kids’ learning, deepen it in many ways, engage them far more in problem solving,” said Michael B. Horn, a faculty member at Harvard University’s Graduate School of Education and co-founder of the Clayton Christensen Institute.
Mr. Horn cautions against extreme reactions on the potential of AI technology. “I think there’s like sort of the extreme rosy view, which I don’t really buy into and then there’s like the extreme pessimistic view, which I don’t buy into either,” he said.
We don't do nuance anymore. Only screaming, spittle-flecked rants get any attention at all. I know; I write for a living. With AI, there must be a measure of nuance, or the numerous benefits that would accrue from adopting AI, as well as realizing potential dangers, will go by the boards.
Would it be unethical to refuse to develop AI capable of making astonishing medical breakthroughs?
“One of the strengths of AI is that it can take really complex data, like rare disease data for example, and they can process it very quickly,” observes Adam Meier, who served as Director of Montana’s Department of Public Health. “And it can identify things that humans may miss.”
Just recently, AI decoded a secret communication from Napoleon to one of his generals after 217 years. The codes had been lost to time, as had the original communication. An AI engineer used OpenAI’s GPT-6 Astra to race through a complex process of organizing the cipher and then using massive amounts of data from other sources to crack it. The new systems have cracked more than 40 historical ciphers in September alone.
As the abilities of AI models continue to improve, I believe we're in far more danger from those who see AI, both promoting it and opposing it, as a path to riches and power than those who want measured, careful progress in realizing AI's full potential.
Exclusively for our VIPs: What If a Solar Flare With the Power of a Trillion H-Bombs Hit Earth? It Would Be a Very Bad Day.