WHEN ANTHROPIC pulled back on releasing its new AI engine, Mythos, because it was too “dangerous” to make widely available, the buzz about the powerful system spread like an out-of-control computer virus.
Nations and Companies Terrified
It appears that the widespread concern is warranted. The model can find weaknesses in software and exploit them to such a degree that Anthropic has only made it available to eleven businesses it partners with in the US and UK to find fixes and defenses for their software. Anthropic operates the LLM Claude and is generally considered the most ethical US AI company.
A New York Times story on April 22 laid bare the fears that companies and countries have about an AI that puts their computer infrastructure at risk, quoting a pro-Kremlin publication that called Mythos “worse than a nuclear bomb.” The geopolitical implications for countries that have fallen behind in the AI race have potentially devastating consequences for the future.
The fear of one country having an unrivaled online weapon has been of concern to the US government for some time, and not in a positive way. The Department of Defense wants access to Mythos, and Anthropic is again saying no to the administration, holding off officials till it helps major infrastructure companies like Microsoft repair its software flaws.
Unfortunately, even with its safeguards, a third-party Anthropic vendor reportedly gained access to the tool. Anthropic said it is investigating.
Jared Kaplan, Anthropic’s co-founder (with CEO Dario Amodei) and chief science officer, explained to The Free Press that the software was designed to identify flaws in software and fix them, but it has succeeded beyond the company's expectations, finding vulnerabilities in software that go back thirty years. Companies that are using it now to eliminate software flaws include Amazon Web Services, which hosts most websites, Apple, Broadcom, CrowdStrike, Google, JP Morgan, the Linux Foundation, Microsoft and Nvidia.
So, ironically, Mythos was created to stop bad actors and prevent hacking, but because the AI works so well it has to be restricted so it doesn’t cause the very thing it’s meant to prevent.
Gen Z Focuses on the Downside
A recent Stanford University poll of subjects identifying with Gen Z found that young people, more than any other generation, are the most concerned about the impact of AI on jobs, energy costs due to data centers, and it’s potential for other kinds of economic harm.
In the results only 10% of Americans report being more excited than worried about AI’s potential. In the science field, 84% of experts say they believe AI will improve health care, while only 44% of the public feel that way. Additionally, 73% of experts expect positive employment outcomes, as opposed to 23% of the public. As AI-adoption speeds up, people in general seem more skeptical and concerned about its impact, rather than optimistic about the future.
AI experts believe the impact of the technology will be far more positive than not because they’re looking at the bigger picture. They imagine that AI will cure most diseases, help reverse aging, create unlimited energy sources, and make other incredible breakthroughs. From that perspective, they understand that it will be worth the pain of change in the beginning to get to the final results.
Results of another poll by Gallup, the Walton Family Foundation and GSV Ventures show that the age group using AI the most is Gen Z. Some young people use ChatGPT for relationship advice, choosing a college to apply to, and for assistance with researching and writing school papers. Half of young people report using AI on a daily or weekly basis.
Their top concerns are about the dearth of entry-level jobs, AI misinformation on social media, and deepfakes. Some young people say they feel they have to know how to use it so they don’t fall behind in education or the workplace.
Their negative views on AI may evolve, along with the technology, as both continue to mature.
Journalist Toni Denis is a partner in Seeflection Inc.
