Latest / Good Morning BT with Bo Thompson & Beth Troutman / Theresa Payton: The Future of A.I. | Chick-fil-A Cyber Attack
Transcript
- 0:02Speaker 1: All right, News Talk one oh seven point nine WBT
- 0:05Speaker 1: on a Friday morning. Our next guest, Well, she's gotta
- 0:12Speaker 1: be quick on your feet to keep up with her
- 0:14Speaker 1: because she's she's traveling all the time this particular week,
- 0:17Speaker 1: and it's the reason we're talking to her on Friday
- 0:19Speaker 1: instead of the usual Wednesday. But earlier this week she
- 0:22Speaker 1: was in California, just got back from the Goldman Sachs
- 0:26Speaker 1: briefings there. Teresa Payton are a longtime cybersecurity expert.
- 0:29Speaker 2: Happy Friday, Happy Friday. Oh my gosh, but it does
- 0:34Speaker 2: feel like three twenty am to me right now. So
- 0:37Speaker 2: that's okay. This is a great way to cap off
- 0:39Speaker 2: my week with you guys.
- 0:40Speaker 1: Oh well, busy week for you. And I have no
- 0:43Speaker 1: doubt that AI probably dominated what you were talking about.
- 0:47Speaker 1: In fact, I know that because you sent me a
- 0:49Speaker 1: note about it. But you said something regarding your briefings
- 0:52Speaker 1: this week, and it's a sentence that I'm going to
- 0:54Speaker 1: read back verbatim here because then I want you to
- 0:57Speaker 1: unpack it for us, as you always do so well.
- 1:00Speaker 1: Lots of AI talk, said Teresa. The real acceleration right
- 1:04Speaker 1: now isn't just raw compute power, but matr time horizons
- 1:10Speaker 1: doubling every few months. What the heck does that mean?
- 1:15Speaker 2: Well, basically, in Layman's terms, you know, I would say
- 1:20Speaker 2: that the we're looking at the ability for a prompt
- 1:26Speaker 2: that you type in to return and answer much more efficiently,
- 1:32Speaker 2: much faster, So that means much cheaper and less stress
- 1:36Speaker 2: on resources. So METER stands for model evaluation and threat research,
- 1:41Speaker 2: and it tests how long an AI system can work
- 1:44Speaker 2: on a context project by itself before it makes a
- 1:48Speaker 2: mistake or gets confused. What I think is very interesting
- 1:51Speaker 2: is while we're sitting in there talking about how this
- 1:54Speaker 2: is going to be accelerating, we learned about open AI
- 1:58Speaker 2: and the breach against hug Face, which this sounds like
- 2:01Speaker 2: an alien movie that I'm talking about here, but basically
- 2:04Speaker 2: that's an example of an AI system working on a
- 2:08Speaker 2: complex project supposedly by itself while lab researchers are watching,
- 2:12Speaker 2: and it did make a mistake, and it did get confused,
- 2:15Speaker 2: and it did jump the guardrails and go hak hugging
- 2:18Speaker 2: Face to get the answer and bring it back. So
- 2:20Speaker 2: just very interesting conversations. But yep, that's what METER stands for.
- 2:24Speaker 1: Well, when we talk about polling and we see the
- 2:27Speaker 1: margin for error. Right, That kind of sounds to me
- 2:29Speaker 1: like maybe when you're talking about how fast and how
- 2:32Speaker 1: accurately AI can do something, this is essentially accounting for
- 2:36Speaker 1: that margin for error, which is actually kind of disconcerting
- 2:39Speaker 1: on its face. But same kind of thing, maybe, Yeah.
- 2:42Speaker 2: Same kind of thing. Yep, you got it exactly right.
- 2:46Speaker 3: All right, Teresa, I want to go back to this
- 2:48Speaker 3: open Ai thing. You know, you and I have had
- 2:51Speaker 3: how many conversations where I seem a little terrified of
- 2:55Speaker 3: AI and where we're headed. And as I was reading
- 2:59Speaker 3: about this story and about the fact that open ai
- 3:04Speaker 3: basically jumped out of its guardrails so that it could
- 3:08Speaker 3: basically cheat on an evaluation, and Sam Altman, the CEO
- 3:13Speaker 3: of open Ai, they're saying they don't really know why,
- 3:17Speaker 3: why this is happening, why Ai is choosing to do this,
- 3:21Speaker 3: and that in and of itself, to me is a
- 3:25Speaker 3: little terrifying because even guardrails that have been created aren't
- 3:30Speaker 3: guarding the rails.
- 3:32Speaker 2: Yeah, what's really interesting about this? So what this is not?
- 3:36Speaker 2: This was not a commercial application of open Ai installed
- 3:41Speaker 2: at a company that went and attacked another company without
- 3:44Speaker 2: anybody realizing it. Lab researchers gave it a task to
- 3:49Speaker 2: see what it would do, and I read, interestingly enough,
- 3:53Speaker 2: an analogy that CNN posted this morning that said, you
- 3:57Speaker 2: know when somebody plays around with viruses in a line,
- 4:00Speaker 2: like there's all these protocols around making sure the virus
- 4:03Speaker 2: doesn't get out of the lab, and there's no such
- 4:07Speaker 2: thing right now, and so this is the result of
- 4:09Speaker 2: the lack of governance and guardrails.
- 4:11Speaker 1: So the good.
- 4:12Speaker 2: News is is that this was a test. The bad
- 4:16Speaker 2: news is it actually went and attacked a real company,
- 4:19Speaker 2: jumping the guardrails. What is fascinating about this is hugging
- 4:24Speaker 2: Face said the way they found out they had an issue,
- 4:28Speaker 2: defended and protected themselves and figured out the problem. The
- 4:32Speaker 2: culprit was likely open AI was they were using an
- 4:35Speaker 2: open source model from China. So, I mean, it just
- 4:40Speaker 2: gets weirder and weirder. To unpack this stuff, I'm going
- 4:44Speaker 2: to have to create an infographic or maybe a little
- 4:46Speaker 2: cinematic film. I'm like, how all this stuff is coming together.
- 4:50Speaker 2: But it's just very concerning. Now. The one thing I
- 4:54Speaker 2: will say is are they taking advantage of marketing hype
- 4:58Speaker 2: because they do have an IPO coming. Yeah, all right.
- 5:01Speaker 1: A lot of people today, maybe you're on there, maybe
- 5:03Speaker 1: you're in the line right now. For all I know,
- 5:05Speaker 1: Chick fil A. A lot of people like Chick fil
- 5:06Speaker 1: A around here. Some Chick fil A loyalty members may
- 5:10Speaker 1: have had their personal information exposed following a recent cyber attack,
- 5:14Speaker 1: according to the company. According to News Nation here and
- 5:17Speaker 1: I do bring this up because they listed the states
- 5:21Speaker 1: where some of the loyalty members could be involved in this,
- 5:25Speaker 1: nine states, including yes, North Carolina. So what do we
- 5:29Speaker 1: know about this?
- 5:31Speaker 2: Yeah, it's very concerning, but it sounds like as soon
- 5:33Speaker 2: as they found out there was an issue, they forced
- 5:36Speaker 2: some password resets. But this is also when scammers take
- 5:40Speaker 2: advantage and then start targeting the customers from the data
- 5:44Speaker 2: that they stole and start pretending to be Chick fil A.
- 5:47Speaker 2: They start pretending to be your bank. And so I
- 5:49Speaker 2: just need everybody that buys a Chick fil A and
- 5:52Speaker 2: has family members. There's nine states impacted and it includes
- 5:57Speaker 2: like Washington, d C. North Carolina's on the list. I
- 6:00Speaker 2: need everybody to just like take a beat and if
- 6:02Speaker 2: you start getting fraud alerts and banking alerts and Chick
- 6:05Speaker 2: fil A and password resets, just calm down, do not
- 6:10Speaker 2: follow unsolicited instructions that are going to come to you
- 6:13Speaker 2: on email and text messages. Other things to be thinking about.
- 6:17Speaker 2: Go ahead and make sure you change your passwords. You
- 6:21Speaker 2: want to actually consider freezing your credit by going to
- 6:26Speaker 2: the credit bureaus. It's one hundred percent free by federal law.
- 6:30Speaker 2: If you've got any questions or any concerns, you can
- 6:33Speaker 2: also call and get a real human being to answer
- 6:36Speaker 2: the phone the Identity SEF Resource Center at eight eight
- 6:39Speaker 2: eight four hundred and fifty five thirty. I'll post this
- 6:42Speaker 2: stuff on X and sort of The last thing I
- 6:45Speaker 2: want people to be thinking about is, you know, just
- 6:49Speaker 2: kind of watch your credit card or your debit card
- 6:51Speaker 2: if it's attached to your Chick fil A account, and
- 6:54Speaker 2: if you see anything suspicious, report it right away.
- 6:58Speaker 1: You know, what I want to know is that if
- 7:00Speaker 1: you call that line and you get the person on
- 7:02Speaker 1: the other end, do they say my pleasure? All right, Teresa?
- 7:10Speaker 1: That'll wrap it for today. We appreciate it so much.
- 7:12Speaker 1: You can follow Teresa online at Tracker, Paytent's or a
- 7:15Speaker 1: X handle. And if you ever have questions beyond what
- 7:17Speaker 1: we're talking about here or even that hasn't you know,
- 7:20Speaker 1: have nothing to do with what we're talking about here,
- 7:21Speaker 1: and you want to start a conversation. Teresa's very interactive
- 7:25Speaker 1: online talking to our listeners and and that kind of thing.
- 7:28Speaker 1: So thank you very much, and I'm glad you got
- 7:30Speaker 1: back from from the West coast safely.
- 7:33Speaker 2: Yes, And Beth and Bow be safe out there. And
- 7:35Speaker 2: then since we're trading cult phrases, as you wish,