Latest / Good Morning BT with Bo Thompson & Beth Troutman / Theresa Payton: Anthropic employee resigns amid concerns of A.I. takeover
Transcript
- 0:00Speaker 1: 7.20 on WBT. We had an impromptu contest there just
- 0:08Speaker 1: a few minutes ago. And now an impromptu conversation with
- 0:14Speaker 1: our cybersecurity expert, Teresa Payton. I didn't know when the
- 0:16Speaker 1: show started today that we were going to be able
- 0:18Speaker 1: to link up with her, but I'm glad we did because...
- 0:20Speaker 1: The topic we talked about yesterday has blown up in
- 0:23Speaker 1: the last 24 hours.
- 0:24Speaker 2: Yeah, this has been a story. Teresa brought it to
- 0:28Speaker 2: our attention and probably all of your attention before most
- 0:31Speaker 2: people saw this yesterday. A guy named Jacob Coxon, who
- 0:35Speaker 2: worked for both OpenAI and Anthropic. has resigned and yesterday
- 0:41Speaker 2: posted on X about the fact that he felt that
- 0:44Speaker 2: both AI companies were acting without responsibility. They were not
- 0:50Speaker 2: approaching AI with guardrails, with guidance. Another employee actually retweeted
- 0:56Speaker 2: or re-Xed or reposted that post from X and said, look,
- 1:00Speaker 2: we believe that there is a greater than 10% chance
- 1:04Speaker 2: that AI will destroy humanity within the next 10 years.
- 1:10Speaker 2: Teresa Payton brought this story to our attention. When I
- 1:12Speaker 2: first saw this post yesterday, there were a few hundred views.
- 1:16Speaker 2: And over the course of just yesterday, this message got
- 1:20Speaker 2: more than 150 million views online and has caught the
- 1:25Speaker 2: attention of lawmakers now who I hope, I hope are
- 1:29Speaker 2: thinking maybe we need to put some guardrails in place.
- 1:32Speaker 1: Teresa, are you there?
- 1:34Speaker 3: I'm here. I'm listening. Here we are.
- 1:39Speaker 1: Well, before we listen to your thoughts on what this
- 1:41Speaker 1: has turned into, Beth was mentioning Jacob Coxon, who started
- 1:46Speaker 1: this yesterday. He was on with Brett Baier on Special
- 1:49Speaker 1: Report on Fox last night. And here is a bit
- 1:52Speaker 1: of what he said.
- 1:55Speaker 4: But because of self-improvement, this AI is making themselves smarter
- 1:59Speaker 4: could happen as soon as next year. Some people even
- 2:01Speaker 4: say six months. So this attitude is shared across the
- 2:05Speaker 4: industry where people essentially are aware that the work they
- 2:08Speaker 4: are currently doing, if we don't slow down, could lead
- 2:11Speaker 4: to AI takeover by the end.
- 2:14Speaker 1: Of the decade. And there's the sentence, by the end
- 2:16Speaker 1: of the decade. And we were talking yesterday about how
- 2:19Speaker 1: quickly this turnaround could be. Yes.
- 2:23Speaker 2: And Teresa, I saw yesterday and what really put this
- 2:27Speaker 2: back on my radar and I noticed that everything was
- 2:29Speaker 2: going viral was you. You brought it to my attention
- 2:32Speaker 2: early yesterday morning, but then yesterday afternoon, you reposted an
- 2:36Speaker 2: article from Politico, the headline being lawmakers urge Congress to
- 2:40Speaker 2: act after AI researchers dire warning. You reposted this and said, okay,
- 2:45Speaker 2: this is the way. How concerned are you now and
- 2:49Speaker 2: how surprised are you that this went viral in the
- 2:51Speaker 2: way that it did?
- 2:52Speaker 3: Yeah, I'm not surprised that it went viral, but I mean,
- 2:57Speaker 3: so there are so many quotes out there now from
- 3:01Speaker 3: all over the AI landscape. So you've got Alex Turner,
- 3:05Speaker 3: who's from Google DeepMind, and he said, I left Google
- 3:08Speaker 3: DeepMind in June. Jacob is right. Many researchers believe they
- 3:11Speaker 3: are building something that could kill everyone on the planet.
- 3:14Speaker 3: These are his words, not mine. And he said, it
- 3:16Speaker 3: was literally my day job to think about how to
- 3:18Speaker 3: stop doing that. Then you've got open AI. You've got
- 3:21Speaker 3: people from there saying, so Jason Wolf, who's a researcher,
- 3:25Speaker 3: I don't know what my probabilities are on literal extinction,
- 3:28Speaker 3: but I think there are a number of ways AI
- 3:29Speaker 3: could go poorly for humanity. And at the current, frankly,
- 3:34Speaker 3: terrifying pace, humanity will be quite lucky if we manage
- 3:37Speaker 3: to find and stay on the narrow path between all
- 3:40Speaker 3: the bad outcomes. These are their words, not mine. So AI,
- 3:46Speaker 3: if it's built the right way, has the power to
- 3:48Speaker 3: do so much good. And I actually want to take something,
- 3:51Speaker 3: one of my favorite AI robots, I'm not sure what
- 3:57Speaker 3: you would call him, is Baymax from Big Hero 6.
- 4:00Speaker 3: He was built as autonomous AI. His entire objective function
- 4:05Speaker 3: in that movie, if you haven't watched in a while,
- 4:07Speaker 3: go watch it, is hard-coded around human well-being and safety.
- 4:12Speaker 3: I've been saying this for over three years, like we
- 4:16Speaker 3: need to be thinking about tech every time we sit
- 4:19Speaker 3: down at a whiteboard and design something new it's going
- 4:22Speaker 3: to do. The first question has to be asked, how
- 4:25Speaker 3: does this support human flourishing and how do we make
- 4:29Speaker 3: sure its primary objective is not about, you know, the
- 4:32Speaker 3: Wall Street Journal article talks about paperclips. It's not about
- 4:35Speaker 3: what the task is, but it starts first with how
- 4:39Speaker 3: do I always, no matter what I am doing, protect
- 4:42Speaker 3: humanity in this scenario?
- 4:45Speaker 2: Here's my big question, Teresa, and you can just give
- 4:49Speaker 2: perspective on this. We've talked before about the fact that
- 4:54Speaker 2: these different AI models take on the personalities of the coders,
- 4:59Speaker 2: maybe the owners of the companies who are coding and
- 5:02Speaker 2: creating these AI models and That now the folks who
- 5:05Speaker 2: are working within these industries say they don't even fully
- 5:08Speaker 2: understand how all of this is working. But on top
- 5:12Speaker 2: of that, we have some of those corporate CEOs, Elon
- 5:16Speaker 2: Musk being one of them, one of.
- 5:18Speaker 1: The creators of Grok.
- 5:20Speaker 2: Out and about in the world talking about empathy and
- 5:24Speaker 2: how empathy is potentially toxic. And if we have people
- 5:28Speaker 2: who think that empathy and being concerned about each other
- 5:32Speaker 2: and being concerned about humanity, which is basically what empathy is,
- 5:36Speaker 2: if that's a bad thing, then no wonder we're suddenly
- 5:39Speaker 2: wandering into a place where AI might kill humanity.
- 5:43Speaker 3: Yeah, I've said before that many of the AI models
- 5:47Speaker 3: take on... So for anybody listening who is a programmer,
- 5:52Speaker 3: was a programmer, or has worked on programming, you know
- 5:56Speaker 3: your programs take on your personality. You put comments in
- 5:59Speaker 3: the code so that when somebody else is trying to
- 6:01Speaker 3: maintain your program in the workplace, they can understand what
- 6:04Speaker 3: you were trying to do. And it does take on
- 6:06Speaker 3: your personality. And so we have to ask ourselves, the
- 6:10Speaker 3: leaders who are choosing the engineers, who are setting the tone,
- 6:14Speaker 3: who are setting the mission, vision, and values of the AI,
- 6:18Speaker 3: what are they like? Because if the programs take on
- 6:21Speaker 3: their personality, are any of them like Baymax or are
- 6:24Speaker 3: they like something else?
- 6:25Speaker 4: Right.
- 6:26Speaker 3: And so this is where I am not usually a
- 6:29Speaker 3: fan of like huge behemoth government regulation, and I'm not
- 6:33Speaker 3: saying that is the answer here. I am worried about
- 6:37Speaker 3: the Hill because I'm worried that a knee-jerk reaction is
- 6:40Speaker 3: actually going to get us the worst of all worlds
- 6:42Speaker 3: in regulation. And it's going to kill maybe a small
- 6:45Speaker 3: upstart who wants to go more the path of Baymax
- 6:48Speaker 3: because they'll be so burdened by regulation. We have to, one,
- 6:52Speaker 3: we need international accords. We need international agreements on this.
- 6:56Speaker 3: This is our modern-day arms race. We're worried about nuclear weapons,
- 7:01Speaker 3: and we absolutely should be. But we could have a
- 7:03Speaker 3: kinetic event with AI if we don't get this right.
- 7:07Speaker 3: We need international accords. We need USA legislation. States are
- 7:11Speaker 3: trying to do the right thing. Europe has a lot
- 7:13Speaker 3: of good things on the books that they're trying to
- 7:15Speaker 3: get passed. But we can have regulation without banning it.
- 7:21Speaker 3: We can have regulation that actually supports smart and safe
- 7:25Speaker 3: innovation that supports human flourishing. But we have to have
- 7:29Speaker 3: the grit and the will to do it. And people
- 7:31Speaker 3: on the inside are saying, please help us. We don't
- 7:34Speaker 3: know what to do.
- 7:35Speaker 1: Teresa Payton coming on with a short notice here, but
- 7:38Speaker 1: this story has really blown up in the last 24 hours.
- 7:40Speaker 1: And we first heard about it with you yesterday. So
- 7:43Speaker 1: thank you so much. And as I always say, keep
- 7:45Speaker 1: the conversation going. And if you want more on this
- 7:48Speaker 1: conversation that we don't have time to get into here,
- 7:50Speaker 1: at Tracker Payton on X is where you can see Teresa.
- 7:53Speaker 1: And she's there all day long. Thank you.
- 7:56Speaker 3: Yeah, thanks for having me on. Everybody out there listening,
- 7:58Speaker 3: please contact your elected officials. It makes a difference. And
- 8:01Speaker 3: Beth and Beau, be safe out there.