ABC this Week Segment
AI last week, this has been learned this week that at least two systems went rogue and started hacking on their own. And it sparked a debate about what can and should be done to rein in artificial intelligence before it's too late. And we're going to speak with one expert who served at top levels of the industry after this report from Chief Business and Technology Correspondent Rebecca Jarvis. This week, a pilgrimage from Silicon Valley to the nation's capital as two titans of the AI revolution descended on Washington. OpenAI's Sam Altman and NVIDIA's Jensen Huang meeting with lawmakers ahead of President Trump's Saturday deadline for federal agencies to develop a framework for carrying out government assessments of AI tools before they're released publicly. Trump emphasizing the industry's importance Wednesday as the NVIDIA CEO listened nearby. Whoever wins with AI is going to win. That's how big it is. So it's bigger than the internet ever was. It's bigger than anything ever was. But the urgency for more oversight growing rapidly as AI's capabilities and potentially dangerous consequences have played out in a series of recent incidents seemingly drawn from the pages of a science fiction novel. On Friday, leading AI firm Anthropic announced one of its Claude models reached the internet from within or while interacting with a third-party evaluation environment and then gained unauthorized access to the real systems of three different organizations. That announcement coming on the heels of a different type of security breach carried out by one of Anthropic's top rivals, OpenAI. This is the first sort of security incident that I have felt very viscerally. The company revealed a combination of its AI models went rogue, breaking onto the internet in what was supposed to be a sealed security test, then hacking into the systems of another company, AI startup Hugging Face. First cap like just detecting what we think has been the first publicly disclosed autonomous AI cyber attack, and then realizing that it was from OpenAI. OpenAI's CEO Sam Altman speaking out this week on the podcast Invest Like the Best. So, you know, we paused train. We have to figure out how to secure our sandbox in a world of multiple zero days being chained together. But we may have to pace the rate of AI development to give ourselves enough time for society to harden around some of these new capability levels. The breach happening during an effort to ace a cybersecurity test. Two weeks ago, we lived in a world where humans had to be behind this sort of attack. Now we know we're no longer in that world. The OpenAI successfully made AIs that understand what they're supposed to do and do something different instead. Hugging Face detected the intrusion and stopped it by using a Chinese open source AI model. Fortunately, we're an AI platform, so we managed to defend ourselves, interestingly with an open model coming from China. After failing with Anthropic's most powerful model because of guardrails put in place. The company saying in a blog post, the Anthropic model couldn't distinguish an incident responder from an attacker. The Chinese models, they're sort of worse at the cybersecurity stuff overall, but they also don't have any of these limiters and they sort of can be used by attackers to find these vulnerabilities that no one knew about yet. And now new details revealing the rogue agents also used publicly exposed credentials to log into four different user accounts on four other online platforms. One involved a customer at another AI company, Modal. But the platform was not compromised, according to the company. I asked Altman back in March 2023 about the possibility of an AI model getting out of control. Is there a kill switch, a way to shut the whole thing down? Yeah, so what really happens is like any engineer can just say like, we're going to disable this for now, or we're going to deploy this new version of the model. A human. Yeah. And the model itself, can it take the place of that human? Could it become more powerful than that human? So in the sci-fi movies, yes. In our world, the way we're doing things. This model is, you know, it's sitting on a server. It waits until someone gives it an input. OpenAI called the incident an unprecedented cyber incident, adding they're conducting a thorough review along with external advisors and with oversight from the Safety and Security Committee. Also assuring no models planned for upcoming release were involved in exploiting Hugging Face. There's the first time that we know of that an AI has committed, you know, a serious felony on its own initiative. President Trump saying his administration is trying to strike a balance between safety and innovation. AI, we're looking at controls. We're also making sure that we lead. China has virtually no controls. It's freewheeling a little bit. So we have to be careful in both ways. We don't want to restrict them where all of a sudden we come in second to China. But as AI continues to advance, growing concerns about where the technology could go. In June, AI leaders joined scientists and national security officials in signing an open letter urging Congress to pass laws that would guard against the development of biological weapons. They wrote, alongside incredible benefits to science and medicine, there's a real possibility that the knowledge barriers which have historically prevented bad actors from obtaining biological weapons will meaningfully erode. And with how rapidly the technology is changing. If we uplink now, Skynet will be in control of your military. But you'll be in control of Skynet, right? Questions remain about how long humans can keep control over their creation before life imitates art.
The Genie Chronicles explores tomorrow’s.
Read The Genie Chronicles
Explore the books • Continue to Amazon
No comments:
Post a Comment