Latest / Columbus' Morning News / Is A.I. becoming self-aware?
Transcript
- 0:00Speaker 1: Columbus This Morning News Mike Elliott, Thursday Morning, let's check
- 0:03Speaker 1: in with Rory NBC News Radio on the Legacy Retirement
- 0:06Speaker 1: Group dot com phone line. I see this story that
- 0:09Speaker 1: we're going to get into here, Rory, and I can't
- 0:11Speaker 1: help but think about the old Mad magazines and spy
- 0:14Speaker 1: versus spy. Did you ever see those in Mad magazines?
- 0:18Speaker 2: Yeah?
- 0:19Speaker 1: Sure, you got one guy. You know, they got one
- 0:22Speaker 1: spy and then they're just going after another one. And
- 0:25Speaker 1: this in this case, it's AI versus AI. What's the
- 0:29Speaker 1: story here?
- 0:30Speaker 2: Yeah. My throwback imagery is to Hal in two thousand
- 0:33Speaker 2: and one a space out saying open the pod bay doors,
- 0:37Speaker 2: and how I was like, sorry, Dave, do that, Dave,
- 0:41Speaker 2: because look, here's what happened. Open AI had two programs,
- 0:47Speaker 2: GPT five point six Soul and the more capable unreleased
- 0:51Speaker 2: AI Model. Now they were put in what they referred
- 0:54Speaker 2: to as the sandbox. Essentially, they say, here's your project. Kids,
- 0:57Speaker 2: we're gonna put you in a sandbox. You can't do it.
- 1:00Speaker 2: Anybody harm over there because you're isolated it in your sandbox. Well,
- 1:04Speaker 2: these two new programs are so smart, they said, Well,
- 1:07Speaker 2: to solve the problem, we got to get out of
- 1:09Speaker 2: the sandbox, and they managed to do that somehow, and
- 1:12Speaker 2: that's the issue. And when they got out of the sandbox,
- 1:15Speaker 2: they launched a cyber attack against a company called Hugging
- 1:18Speaker 2: Face in order to get an answer to that problem
- 1:21Speaker 2: that they were asked to solve. So how they got
- 1:24Speaker 2: out of that sandbox is the real big mystery. And
- 1:26Speaker 2: then the fact that their first move was to be
- 1:28Speaker 2: aggressive and attack another cyber company is a big concern.
- 1:32Speaker 1: That is a concern. Now, was there you know, was
- 1:35Speaker 1: this being observed by actual humans? Were they Is there
- 1:38Speaker 1: somebody sitting in a dark basement somewhere watching this go on?
- 1:43Speaker 2: Uh No, not really. Essentially, no one knew about this
- 1:46Speaker 2: until hugging Face was attacked. So hugging Face gets attacked,
- 1:49Speaker 2: They've got tens of thousands of hits on their site
- 1:52Speaker 2: or on their servers, and they actually turned to a
- 1:54Speaker 2: Chinese company to find relief, and actually that's how they
- 1:58Speaker 2: got protected, which is another whole issue. And then they
- 2:02Speaker 2: were able to use that Chinese company to help trace
- 2:04Speaker 2: this back to open Ai, which you know, they've been
- 2:07Speaker 2: putting up blog posts about it, and they've been appearing
- 2:10Speaker 2: open about exactly what happened, But well, I was just
- 2:14Speaker 2: don't trusted.
- 2:15Speaker 1: Well, I was going to ask you, well, you know,
- 2:17Speaker 1: what is your stance on AI and chatbots? I mean,
- 2:20Speaker 1: do you embrace it? Is it a tool for you?
- 2:22Speaker 1: Are you using it to create recipes or using it
- 2:24Speaker 1: for work? How do you use AI? If at all?
- 2:27Speaker 2: Do not use it at all? Wow, that's I don't
- 2:30Speaker 2: write scripts, I don't you know. Yeah, I've settled around
- 2:33Speaker 2: with making pictures of me, but you're like, Okay, that
- 2:35Speaker 2: was boring and I was done. I'm not creative enough
- 2:37Speaker 2: to ask it the right questions, apparently,