Nvidia has announced a new security platform that it says can prevent AI agents from going rogue.
According to the company, this new tech leverages its OpenShell open source software that “sets boundaries” that will limit AI agents within “milliseconds.” Nvidia also uses its Sentry tech on a separate chip for continuous monitoring and enforcement.
Of course, only time will tell how well this system actually works. While Nvidia says this tech could have prevented incidents like OpenAI agents recently autonomously hacking AI company Hugging Face, it’s unclear how efficient it will be as it’s more widely used.
This platform comes amid numerous concerns related to AI safety, not only with respect to hacking incidents like this but also how chat bots such as ChatGPT can help bad actors like the orchestrator of the Tumbler Ridge, B.C. shootings. This is, of course, to say nothing of the steep environmental costs associated with AI and threats to human art and other professions.
It’s also worth noting that Nvidia CEO Jensen Huang made some decidedly questionable AI-related comments last week. According to him, it doesn’t matter if future generations forget basic skills like math as they grow increasingly reliant on AI because the technology will simply help them “discover new ones.” He justified this baffling statement by saying he doesn’t even know his own home address anymore since tech like AI can do everything for him. (But of course, an AI could surely never replace him in the all-too-important-and-totally-not-overpaid role of CEO.)
So sure, Nvidia might be trying to regulate own AI agents, but if they’re making everyone dumber in the process, then who cares, right?
Source: Nvidia
AI 
News 
News 
News 