The first aspect of cyber defense of AI starts with the networks, Marine Corps Lt. Gen. Michael S. Groen said today during a virtual fireside chat at the Billington CyberSecurity Summit. "The department is undergoing a little bit of a mind shift on networks and architecture. Our networks are a core piece of our warfighting architecture. Our networks are weapons, and, so, we have to treat them like weapons. We have to, we have to plan to protect them, make them resilient because everything that we're going to do in an artificial intelligence or data-driven way will depend on the security [of] those networks," he said.

In the many decades that video games have had us battling through monster-filled dungeons, we've seen hundreds of variations on the format – but Boyfriend Dungeon marks the first time that we've been able to go on dates with the weapons we wield. As you fight esoteric enemies, from flying VHS tapes to sentient cocktail glasses, you build up affinity with your chosen weapon, which has the uncanny ability to transform into a human. After a while, you unlock a date with your human/weapon hybrid companion. And after you've, say, taken a stroll around the park or gone to a concert – choosing along the way whether you want to take things further or remain just friends – you unlock powerful extra abilities for that blade. You can romance everyone within sight, or opt for platonic friendships across the board. Either way, your phone will be constantly buzzing with requests for your company.

Unemployment in the United States stands at a 50-year low. The quit rate of workers hovers near an all-time high. And the number of open jobs continues to outpace the number of unemployed individuals. Workers have reaped the benefit of this employment boom, through more job options and bigger paychecks. But it has ramped up pressure on HR departments grappling with recruiting and retaining top talent.

Reinforcement learning frameworks have introduced abstractions to implement and execute algorithms at scale. They assume standardized simulator interfaces but are not concerned with identifying suitable task representations. We present Wield, a first-of-its kind system to facilitate task design for practical reinforcement learning. Through software primitives, Wield enables practitioners to decouple system-interface and deployment-specific configuration from state and action design. To guide experimentation, Wield further introduces a novel task design protocol and classification scheme centred around staged randomization to incrementally evaluate model capabilities.

THEY don't drink, they don't get tired and they don't go on strike. To hospital managers, the idea of robots operating on patients without human intervention is an attractive one. To patients, though, the crucial question is, "are they better than human surgeons?" Surgery is messy and complicated. A routine operation can become life-threatening in minutes. Such considerations have meant that the role of robots in operating theatres has been limited until now to being little more than motorised, precision tools for surgeons to deploy--a far cry from the smart surgical pods and "med-bays" of science fiction.