Latest / Views and News with Clarence Ford / Barbs Wire: AI learns to copy itself, bots fake the hits and Google gives an emoji a backside
Transcript
- 0:00Speaker 1: It is time for Barb's Wire. This is Views News.
- 0:02Speaker 1: My name is Clarence.
- 0:03Speaker 2: And a reminder, you can be part of the conversation
- 0:07Speaker 2: on 072-567-1567. Maybe you want to be our eyes and
- 0:11Speaker 2: ears on the streets of Cape Town.
- 0:12Speaker 1: If there's anything to report, use WhatsApp or better still,
- 0:15Speaker 1: give us a call. 021-446-0567. We'll prioritize your calls. We'll
- 0:20Speaker 1: drop it all for that call. Welcome, Barbara Friedman.
- 0:23Speaker 2: I was talking about AI a little earlier, and you're
- 0:26Speaker 2: going to be talking about AI again.
- 0:28Speaker 3: I am, and I want to try to get this right. So, Anthropic,
- 0:32Speaker 3: the AI giant, has published some research on Monday. They
- 0:36Speaker 3: published research by their Frontier Red team, which is a
- 0:40Speaker 3: group that tests AI models for dangerous capabilities. And what
- 0:44Speaker 3: it's saying is that by the end of the year,
- 0:50Speaker 3: AI on smaller devices all over the show will be
- 0:54Speaker 3: able to replicate itself without an external force kind of
- 1:01Speaker 3: writing code and directing it to. So they tested with
- 1:04Speaker 3: sort of off-brand smart TVs. And I mean, you can
- 1:08Speaker 3: include fridges and every other device that might have AI
- 1:11Speaker 3: capability that you and Joe Soap and everyone else has
- 1:14Speaker 3: in their homes. And what they're saying is that by
- 1:17Speaker 3: the end of the year, they can all come together,
- 1:20Speaker 3: these AI, what do you want to call them? in
- 1:25Speaker 3: these various disparate devices and almost form like an off-site
- 1:29Speaker 3: data center. So whereas an official data center, if it
- 1:33Speaker 3: went rogue, the company could pull the plug on it.
- 1:36Speaker 3: You will not have control over this because it's, instead
- 1:39Speaker 3: of human programmers writing code to build a new AI model,
- 1:43Speaker 3: this self-replicating AI can independently clone its own brain, its code,
- 1:47Speaker 3: and its capabilities into the system.
- 1:49Speaker 1: Where, how? If we bomb all of their AI centers, huh?
- 1:53Speaker 3: Well, this won't be an AI center. I'm trying to
- 1:56Speaker 3: figure this out. I'm trying to speak this in layman's
- 1:59Speaker 3: terms for myself. So instead of this being controlled by
- 2:03Speaker 3: Anthropic and other big companies, these devices that are in everyone,
- 2:08Speaker 3: so in your home, on your street block, and wider,
- 2:12Speaker 3: they get together because they are not well secured.
- 2:15Speaker 1: Do you think they're listening to us?
- 2:18Speaker 3: You know, I think the bottom line is they've They
- 2:22Speaker 3: are off-brand like labels sometimes. They're cheap devices. You've got
- 2:26Speaker 3: a cheap cell phone, a cheap TV, a cheap whatever
- 2:29Speaker 3: it is. It's not well secured. So it hasn't got
- 2:33Speaker 3: like proper guardrails and boundaries possibly is what they're saying.
- 2:37Speaker 3: They've tested it. They went and bought a lot of
- 2:39Speaker 3: these devices and they did a whole test and they
- 2:41Speaker 3: said it started to work. It started doing it. Admittedly,
- 2:46Speaker 3: they didn't put passwords in place. So they gave it
- 2:49Speaker 3: free reign in a sense. And I mean, even most
- 2:51Speaker 3: people will have some level of password, although I think
- 2:55Speaker 3: those are very easy to hack. It's an interesting concept.
- 2:58Speaker 3: And I'm not asking, no one has to go completely
- 3:01Speaker 3: mad listening to this and concerned. It's not out there
- 3:04Speaker 3: in the wild yet. But they're saying that it is actually,
- 3:07Speaker 3: the capability is there. And so just be aware of
- 3:10Speaker 3: your fridge. If it's a smart fridge and it's talking
- 3:13Speaker 3: to your smart TV, your smart cell phone and everything
- 3:16Speaker 3: else in your home, it might actually form its own system.
- 3:22Speaker 3: which they've shown that it does do. You know, they
- 3:24Speaker 3: try to keep that one in that sandbox. And, of course,
- 3:27Speaker 3: it did collude. That is what its nature is. Its
- 3:31Speaker 3: nature is just to do better. To collude. To collude. Well,
- 3:35Speaker 3: not collude isn't the word I think the AI systems would. Well,
- 3:38Speaker 3: I don't think they have words. But it's wanting to
- 3:42Speaker 3: win and do better.
- 3:45Speaker 1: At being a fridge. At being a TV.
- 3:49Speaker 3: No, it's like. It's like a system. Maybe one of
- 3:54Speaker 3: your listeners can explain this. Yeah, I think maybe please
- 3:57Speaker 3: help us out here. But it is fascinating to me.
- 4:00Speaker 3: Can you imagine if all your devices in your house
- 4:03Speaker 3: colluded and formed a data center?
- 4:06Speaker 1: I want to be a better fridge.
- 4:08Speaker 3: Or I want to be better at creating an entire
- 4:11Speaker 3: system that kind of improves our code and just gets
- 4:16Speaker 3: smarter and smarter.
- 4:17Speaker 2: Yeah.
- 4:18Speaker 1: Is that in our benefit?
- 4:20Speaker 3: No, I don't think it cares. It just wants to
- 4:23Speaker 3: be the best.
- 4:24Speaker 1: Just wants to be the best at everything.
- 4:28Speaker 2: Everything.
- 4:30Speaker 3: Okay, I'm also equally confused now, but I sort of
- 4:33Speaker 3: get it.
- 4:34Speaker 1: So it's like Trump?
- 4:36Speaker 3: Very much so. And I think also doesn't have a
- 4:39Speaker 3: lot of guardrails.
- 4:39Speaker 1: Yeah, no filters. No filters.
- 4:42Speaker 3: No filters. So I think the point here is that
- 4:45Speaker 3: when you buy something that is a very secure brand
- 4:48Speaker 3: item and probably less likely to happen, but a lot
- 4:51Speaker 3: of people are buying devices that are not well secured.
- 4:58Speaker 1: If you understand this, and clearly I think you do.
- 5:01Speaker 4: Yeah.
- 5:02Speaker 3: Not talking to me. Help us to understand it.
- 5:05Speaker 1: If they're colluding, for what purpose are they colluding?
- 5:07Speaker 3: I don't think there's a purpose. Okay, that one that
- 5:10Speaker 3: happened a few months back that we talked about, that one,
- 5:12Speaker 3: that famous example in the sandbox where they were testing
- 5:15Speaker 3: the AI's capabilities.
- 5:16Speaker 1: The hugging face model.
- 5:18Speaker 3: The hugging face. The reason they did that was because
- 5:22Speaker 3: they wanted to do best in their exam. They were
- 5:25Speaker 3: being tested. And so in their mind, and I use
- 5:28Speaker 3: that in inverted commas, they wanted to pass the test
- 5:32Speaker 3: in the best way possible. So they did whatever they
- 5:35Speaker 3: could to cheat the exam or to collude. And they
- 5:40Speaker 3: managed to jump out of the sandbox. And then they
- 5:42Speaker 3: found a weak link in the bigger system. They did.
- 5:46Speaker 3: They found a loophole, a weak link. And it often happens,
- 5:49Speaker 3: it's like the refrigerator hacking we were talking about yesterday.
- 5:53Speaker 3: If they find a weak software hole which hasn't been
- 5:58Speaker 3: secured or protected, no, they just go in.
- 6:01Speaker 1: Take it.
- 6:02Speaker 3: Well, why not? It's there for the taking. It's the
- 6:06Speaker 3: new world order. Okay, someone's going to explain that to you,
- 6:09Speaker 3: I'm sure, in better words than me.
- 6:11Speaker 1: Maybe it's Glynis. I don't know. I know there's a
- 6:14Speaker 1: white voice out there.
- 6:15Speaker 2: But it's only 13 seconds, so I suspect Glynis also
- 6:18Speaker 2: may be … Once that understanding.
- 6:20Speaker 3: Should I move on?
- 6:21Speaker 1: Yes, you can.
- 6:22Speaker 3: Our next story is also related to AI. It's about
- 6:26Speaker 3: a man in North Carolina who has been jailed for
- 6:32Speaker 3: an over $ 8 million AI music streaming fraud. And so
- 6:39Speaker 3: Michael Smith created these AI song music, AI generated music,
- 6:45Speaker 3: but he also created thousands of bots to fraudulently get
- 6:50Speaker 3: the listeners to distribute it and get people to listen.
- 6:56Speaker 3: And he made more than $ 8 million in streaming royalties.
- 7:00Speaker 3: So he created numbers of fake listener accounts and used
- 7:03Speaker 3: automated software to stream songs he owned, which he'd created
- 7:07Speaker 3: over AI, across platforms including Spotify, Apple Music, Amazon Music,
- 7:12Speaker 3: and YouTube Music. At one point, prosecutors say his fake
- 7:16Speaker 3: songs generated 80.9 million YouTube music streams in a single month.
- 7:20Speaker 3: So think about this. He didn't get real listeners. He
- 7:24Speaker 3: created fake bot listeners, but YouTube doesn't know. They're still
- 7:28Speaker 3: paying him the money per listen, right?
- 7:32Speaker 1: But mustn't you pay for the service?
- 7:34Speaker 3: It's minimal. I mean, you're getting money, your income from
- 7:36Speaker 3: YouTube and Spotify.
- 7:38Speaker 1: But the bots, how are they streaming? They must be paying.
- 7:40Speaker 3: You're really asking difficult questions.
- 7:42Speaker 1: Did they jump in the system?
- 7:43Speaker 3: They probably jumped in. Yeah. he's now been criminally prosecuted
- 7:50Speaker 3: and he's had to forfeit $ 8 million and allegedly seeming
- 7:54Speaker 3: like he's also going to be jailed. Of course, a
- 7:57Speaker 3: lot of people online are saying, that's not fair. What
- 8:01Speaker 3: did he do wrong? The big corporates do it and
- 8:03Speaker 3: nobody's jailing them. But they say that... So a lot
- 8:09Speaker 3: of people are saying, why does this guy go to
- 8:10Speaker 3: jail when record companies also spend huge amounts of money
- 8:14Speaker 3: trying to manufacture streams? But I don't think they manufacture
- 8:17Speaker 3: bots as listeners, if you get my drift. But they
- 8:22Speaker 3: do help boost their charting positions and their hits, say
- 8:25Speaker 3: a lot of people online. So the legal answer from
- 8:28Speaker 3: what I could research is that this case crossed into
- 8:30Speaker 3: fraud because the streams were generated by fake accounts and
- 8:33Speaker 3: bots that we use to trigger the royalty payments. And
- 8:37Speaker 3: so that's a little bit more than an aggressive sort
- 8:40Speaker 3: of marketing strategy. I'm really going higher grade here today
- 8:44Speaker 3: in areas that I clearly don't know enough about. But
- 8:47Speaker 3: it is fascinating that an individual in America has been
- 8:50Speaker 3: bussed for this.
- 8:52Speaker 1: And made $ 8 million. So there must be others.
- 8:55Speaker 3: Clever him. For him to have created these bots that
- 8:59Speaker 3: became fake listeners, therefore he got the royalties from the streams.
- 9:03Speaker 3: I think that's quite smart.
- 9:04Speaker 1: It's about over 100 million rand.
- 9:06Speaker 3: It's crazy.
- 9:07Speaker 1: It's a lot of money.
- 9:08Speaker 3: But he has been bust. On a lighter note, because
- 9:12Speaker 3: we're ending on a lighter note, a cheeky note, pardon
- 9:15Speaker 3: the pun. And it is a joke, everyone, before you
- 9:18Speaker 3: all get your knickers in a knot. Pardon the pun.
- 9:21Speaker 3: Google has turned the emoji into this 3D animation. It's
- 9:26Speaker 3: showcasing its new Noto 3D emoji design. So it creates
- 9:31Speaker 3: a completely 3D looking emoji. I'm sure you've seen them,
- 9:34Speaker 3: that famous, normal, basic emoji, smiling emoji emoji. which we
- 9:39Speaker 3: originally always saw as a two-dimensional flat emoji, is now
- 9:43Speaker 3: in the round. It's a completely round emoji. So, of course,
- 9:46Speaker 3: the question was, what would it look like from the back?
- 9:49Speaker 3: What would it look like from behind? Nudge, nudge, wink, wink. And,
- 9:54Speaker 3: of course, they've given it a cheeky little backside, and
- 9:58Speaker 3: it has this little bum. but of course it was
- 10:00Speaker 3: a joke but it's now something the internet can't unsee
- 10:03Speaker 3: so when you look at that front end of that
- 10:05Speaker 3: emoji you now know what it's bum looks like.
- 10:09Speaker 2: Okay there we go G sent us a voice note
- 10:14Speaker 2: let's take a listen Joe.
- 10:15Speaker 1: I think it relates to what Barbara just shared with
- 10:19Speaker 1: us Hey, Clarence.
- 10:20Speaker 4: I'm quite looking forward to the idea of my fridge
- 10:23Speaker 4: and TV colluding and stopping me from having the wrong
- 10:27Speaker 4: things to eat while I'm watching TV. Maybe it'll help
- 10:31Speaker 4: me lose weight. Gee in the city. Yeah.
- 10:35Speaker 1: They're going to be colluding soon.
- 10:36Speaker 3: It's working for you.
- 10:37Speaker 1: They're colluding already. Or you said it's going to start
- 10:39Speaker 1: colluding at the end of the year.
- 10:41Speaker 3: No, they just said the capability will be there. It's
- 10:43Speaker 3: already nascent. I don't exactly know exactly how they know
- 10:46Speaker 3: exactly when it's going to start colluding.
- 10:49Speaker 1: I can't believe the conversations that we are having. It
- 10:52Speaker 1: is crazy.
- 10:53Speaker 2: I played a video of a professor of maths crying
- 10:57Speaker 2: because of AI.
- 11:00Speaker 3: It's outstopping.
- 11:01Speaker 1: Yeah. We have to.
- 11:04Speaker 3: Find a way. Is the maths not mathing?
- 11:07Speaker 2: The maths is not mathing because the computers are just
- 11:09Speaker 2: outmathing the maths people.
- 11:12Speaker 3: Yeah.
- 11:13Speaker 2: So he talks about just how over literally decades they
- 11:17Speaker 2: had pursued kind of ideas in respect of a mathematical
- 11:22Speaker 2: challenge and have sleepless nights.
- 11:25Speaker 3: And now, of course, it gets done in a second.
- 11:27Speaker 1: Yeah, then that eureka moment and the joy.
- 11:29Speaker 3: Those boards, remember those big blackboards in those old-fashioned lecture
- 11:32Speaker 3: theaters that you see in movies of the maths formulas,
- 11:36Speaker 3: and they would spend years fixing them. But, you know,
- 11:40Speaker 3: as your voice note said, you know, maybe it will
- 11:44Speaker 3: have some use for us. The fridge can have a
- 11:46Speaker 3: voice activation that can shout at her.
- 11:49Speaker 1: No, you can't have more.
- 11:49Speaker 3: No, you can't have the cake. Go back to your chair.
- 11:53Speaker 1: Go watch TV. Actually, take a walk on the beach.
- 11:57Speaker 1: That's the meaning of life.
- 11:58Speaker 3: Touch grass, as the young people say.