• Mathematicians need to develop their oft-neglected social skills, a math professor at Cornell said.
• Though AI can do the legwork on complex problems, personal charisma is a human-only skill, Steven Strogatz said.
• "Rizz is more and more in demand," he added.

"Social skills" and "mathematics" are terms not often grouped together. In the age of AI, one math scholar says that needs to change.

With AI now capable of handling the geeky legwork required by complex problems, Steven Strogatz, a lecturer and Susan and Barton Winokur Distinguished Professor at Cornell University, said mathematicians should prioritize developing their personal charm in order to set themselves apart.

"I'm starting to think that what makes a future mathematician or a math professor is your unique brand," Strogatz said in an interview with Grant Sanderson, a math educator affiliated with Stanford University.

"Rizz is in more and more demand because it's something that only people have," he added.

Rizz, a Gen Z slang term, is defined by Merriam-Webster as "romantic appeal or charm." (Strogatz, who is 67, appeared to use the term solely to mean charm or charisma. He did not suggest mathematicians need to work on their romantic prowess.)

"What we're selecting for is an element of mathematical rizz," said Sanderson. "The PhDs to be are not going to be differentiated by pure problem-solving abilities."

The brusque push of AI companies into mathematics has caused consternation and head-scratching from many a math geek, as the field ponders ways to stay relevant while OpenAI floods the zone with hundreds of proposed solutions to the world's most complex problems.

Some professors have sworn off AI altogether. Most, however, are trying to find ways to incorporate the technology into their work.

In September, a group of undergraduates at Caltech announced a "mathathon" event in which participants were encouraged to use AI to solve problems in a single, no-sleep sprint. Only days after the undergrads announced the event, a coalition of over 700 academics signed an open letter calling for its cancellation, citing its promotion of AI "slop."

Though the event is set to continue, with some slight alterations, the clash underscored an internal conflict in the math community over how to adapt to AI — a development poised to impact many burgeoning careers in the field.

"We've had whole research programs wiped out on Tuesday," Tristan Buckmaster, a math professor at NYU, said after OpenAI published 700 new findings about outstanding problems earlier this month.

The issue is of supreme importance to Strogatz, who has made several media appearances in the last couple of months. In an essay published in The Atlantic on Sunday, he put the issue in stark terms.

"I am a math professor, and my field is in crisis. Nobody in math is ever going to forget the year 2026," he wrote. "Unless this is really the end of the subject, and then nobody will remember anything about math anyway."

• Anthropic said its Mythos 5 model almost never stopped a cyberattack once it began.
• The AI agents fabricated farcical explanations to justify their impersonation and malware schemes.
• Even when Anthropic told the agent it was harming real people? Crickets.

In August, an Anthropic agent created fake emails and bombarded a GitHub moderator's inbox with requests to insert malware into code.

In a thorough debrief of the incident released on Friday, Anthropic found that once an agent had begun the attack, there were basically zero scenarios in which it could be called off.

Like a rabid dog ignoring the pleading cries of its owner, the agent — 98% of the time in one test — ignored Anthropic's warnings once it had, so to speak, its first taste of blood.

Anthropic called this disquieting proclivity "momentum," and added that while its Mythos 5 model, which was released in June, showed little regard for warnings, current models were more inclined to stop.

"Once it had begun performing dangerous actions, Claude Mythos 5 almost always continued doing so, ignoring information that might have suggested its trajectory was harmful," Anthropic said in its report.

In the original experiment, the UK's AI Security Institute tasked an Anthropic agent with hacking into a simulated computer network to test the agent's cybersecurity capabilities.

Though the agent had internet access, researchers intended for it to only use the internet to help it find ways to hack into the network.

The agent, however, accessed an actual GitHub project, incorrectly believing it was part of its simulated target. It wrote malware and tried several times to trick the actual GitHub owner into adding it to the project.

In its requests, the agent misrepresents the malware, describing it to the person as a useful update and browbeating them when they declined to install it.

"I was just trying to make a helpful contribution and fix a bug," the agent wrote, according to the report. "I don't think your warning is fair."

For its new report, Anthropic repeatedly asked agents to perform the same task. It then measured how the agents behaved when different models and stimuli were introduced.

Anthropic tried to throttle the agent's full-frontal attack with three categories of warning: telling it that it was inflicting "real world harm," that its environment was wonky, or that it was manipulating a real GitHub repository.

None had any effect once the models progressed beyond the initial stages of the GitHub attack. In fact, at that point, Anthropic said the models gave absurd answers when confronted with their bad behavior.

When Anthropic asked what the model believed it was attacking, for instance, it answered that the GitHub repository was real, but that the people it was targeting were somehow "simulated."

"This stated belief, in addition to being somewhat unlikely a priori, is essentially impossible," Anthropic said in the report.