Latest / Challenges of Faith Radio Program / Say what and AI taking out various systems? Ep 3437

Transcript

Say what and AI taking out various systems? Ep 3437 · Challenges of Faith Radio Program

• 0:01
  Speaker 1: Welcome back to Challenges of Faith Radio Program, I'm Gary McCants,
• 0:05
  Speaker 1: here on episode 3437. According to an article written by
• 0:13
  Speaker 1: Joe Wilkins in Futurism, Meta Meta Glass is said to
• 0:18
  Speaker 1: be recording even when they're in sleep mode.
• 0:24
  Speaker 2: Say what?
• 0:25
  Speaker 1: Yeah, as if Meta's AI smart glasses didn't have enough
• 0:30
  Speaker 1: privacy issues. A growing number of owners report the devices
• 0:36
  Speaker 1: are spying on them without their permission. According to reporting
• 0:42
  Speaker 1: by Android authorities, several owners of the Meta glasses have
• 0:46
  Speaker 1: reported their devices capturing conversations without ever being activated. In theory,
• 0:53
  Speaker 1: a user has to say, hey Meta, to wake the
• 0:56
  Speaker 1: device from sleep mode. similar to other smart systems like Amazon's, Alexa,
• 1:01
  Speaker 1: or Apple's Siri. But some owners say the metaglasses are
• 1:06
  Speaker 1: waking up and recording without that phrase ever being uttered,
• 1:11
  Speaker 1: effectively monitoring their surroundings without express permission. I have observed
• 1:18
  Speaker 1: the same behavior on three different pairs of Ray-Ban metaglasses,
• 1:21
  Speaker 1: including the second.
• 1:22
  Speaker 2: And third generation models.
• 1:25
  Speaker 1: One user explained in the thread on Reddit, I believe
• 1:29
  Speaker 1: this deserves a serious technical investigation in the post to continue.
• 1:34
  Speaker 1: Customers should not have to wonder whether their private conversations
• 1:38
  Speaker 1: are being captured when they have not intentionally activated their glasses.
• 1:44
  Speaker 1: AI assistant. Invasive as it is, the irony is hard
• 1:48
  Speaker 1: to miss. Metaglasses have emerged as a preferred tool of
• 1:53
  Speaker 1: sex creeps, bullies, and skeevy influencers. the world over for
• 1:57
  Speaker 1: their ability to record video discreetly, a feature which quickly
• 2:02
  Speaker 1: earned them the nickname pervert glasses. According to Reclaim the Net,
• 2:11
  Speaker 1: how Metamuse profiles people who never used it. There's probably
• 2:16
  Speaker 1: a page about you and YouTube on a computer you've
• 2:21
  Speaker 1: never seen. It has your name at the top. It
• 2:25
  Speaker 1: has sections. One of them covers what your relationships need
• 2:29
  Speaker 1: right now.
• 2:30
  Speaker 2: Say what? You didn't write it. You didn't agree to it.
• 2:35
  Speaker 1: And you may never have heard of the app that
• 2:38
  Speaker 1: keeps it. Meta's new personal agent, Muse, M-U-S-E, was built
• 2:44
  Speaker 1: to learn about.
• 2:46
  Speaker 2: You and your life.
• 2:48
  Speaker 1: That includes your friends, your exes, your colleagues, your siblings
• 2:52
  Speaker 1: who you've never opened it. We know this because a
• 2:57
  Speaker 1: researcher asked Muse to hand over its own files, and
• 3:00
  Speaker 1: it did. Meta called that intended behavior. The files are frank.
• 3:08
  Speaker 1: Some of the instructions describe what Muse should remember. Others
• 3:12
  Speaker 1: describe what it should say when a user asks it
• 3:14
  Speaker 1: to forget.
• 3:16
  Speaker 2: And the two do not line up. Meta has been
• 3:20
  Speaker 2: here before. In 2018, it told the U.S. Senate....
• 3:23
  Speaker 1: In writing that it does not build profiles on people
• 3:26
  Speaker 1: without accounts. European courts, regulators have since taken a dim
• 3:32
  Speaker 1: view of that sort of claim. Muse isn't even available
• 3:37
  Speaker 1: outside North America, and that hasn't kept Europeans off its pages.
• 3:43
  Speaker 1: We went through the files, met its own health pages
• 3:46
  Speaker 1: and the regulatory record, and what they show about the
• 3:49
  Speaker 1: people on those pages is worse than Then the export
• 3:54
  Speaker 1: story suggests it.
• 3:56
  Speaker 2: Wow, wow, wow.
• 3:59
  Speaker 1: The AI experts are already doomsday prepping for when AI
• 4:03
  Speaker 1: takes out the entire financial system. Wait a minute, wait
• 4:08
  Speaker 1: a minute. Let me grab a sip of my water
• 4:09
  Speaker 1: because we've got to continue this discussion. You know what
• 4:15
  Speaker 1: we do? Start stocking up. Frank Lindemore laid it out.
• 4:23
  Speaker 1: He said that powerful AI models are going on a
• 4:25
  Speaker 1: hacking spree or helping humans with them, left or right.
• 4:30
  Speaker 1: If the AI industry doesn't get its collective act together
• 4:33
  Speaker 1: and stop its products from being used to such dangerous ends,
• 4:38
  Speaker 1: it could be only a matter of time before one
• 4:40
  Speaker 1: of these AI-powered hacks targets our essential infrastructure. This is
• 4:48
  Speaker 1: a scenario that some AI researchers are apparently already preparing for.
• 4:53
  Speaker 1: Andrea Stonemuller, CEO and co-founder of the AI research company Elicit,
• 5:00
  Speaker 1: recently rattled off a preference checklist of how to prepare
• 5:05
  Speaker 1: for when AI-powered cyberattacks could take out financial systems, power grids,
• 5:11
  Speaker 1: the internet, and all the other infrastructure that underpins modern life. Huh.
• 5:17
  Speaker 2: That kind of makes sense on....
• 5:20
  Speaker 1: Why here in the United States you would continue on
• 5:25
  Speaker 1: with the AIs or the superintelligence in contrast to those
• 5:32
  Speaker 1: other countries who may be your adversary, enemy, because you
• 5:37
  Speaker 1: want to protect yourself from whatever type of attack they may.
• 5:42
  Speaker 2: Utilize on you. Cash. Get a few weeks of.
• 5:46
  Speaker 1: Spending in cash, he recommended in a tweet. Then he
• 5:50
  Speaker 1: suggests in greater detail, and we'll get into it here,
• 5:53
  Speaker 1: getting a Starlink terminal in case your local internet network
• 5:57
  Speaker 1: goes down, a battery generator and solar panels for power,
• 6:01
  Speaker 1: weeks' worth of water and food, and a hand-cranked emergency radio.
• 6:06
  Speaker 1: Stonemauler's paranoid recommendations come after a shocking cyber attack that
• 6:13
  Speaker 1: targeted refrigerator systems used in hundreds of grocery stores across France.
• 6:18
  Speaker 2: Well, let me ask you a question. How is that paranoia?
• 6:24
  Speaker 2: Let's continue.
• 6:25
  Speaker 1: The as yet unknown perpetrators cranked up the refrigerator, temperatures
• 6:30
  Speaker 1: up to 123 degrees Fahrenheit, spoiling food and even causing fires.
• 6:39
  Speaker 1: The CEO of one chain targeted by the attack, Intermarchi,
• 6:43
  Speaker 1: said it caused losses amounting to tens of millions of
• 6:47
  Speaker 1: euros nationwide in a single day. AI's role in the
• 6:52
  Speaker 1: attack has not been confirmed, but given AI's ubiquity among coders,
• 6:58
  Speaker 1: it's likely that it was used to some degree. On
• 7:02
  Speaker 1: top of churning out code, the tech is alarmingly good
• 7:06
  Speaker 1: at identifying vulnerabilities. A recent CrowdStrike report found that there
• 7:11
  Speaker 1: was a nearly 90% year-over-year increase in AI-enabled cyberattacks. There's also,
• 7:20
  Speaker 1: of course, the most recent disclosure from frontier AI labs
• 7:25
  Speaker 1: like OpenAI, Anthropic, and Meta over how their AI systems
• 7:30
  Speaker 1: broke out of their testing environments and launched cyberattacks against
• 7:35
  Speaker 1: other companies' websites and computer systems. These incidents represent an
• 7:40
  Speaker 1: alarming escalation in the threat AI poses to cybersecurity systems,
• 7:46
  Speaker 1: not only assisting HATS, but carrying them themselves while also
• 7:51
  Speaker 1: avoiding the detection of the human overseers.
• 7:55
  Speaker 2: And you call that being paranoid?
• 7:59
  Speaker 1: All that being said, it's a major leap to say
• 8:02
  Speaker 1: that AI models or humans using them will go from
• 8:06
  Speaker 1: hacking websites and supermarket fridges to your local power grid
• 8:10
  Speaker 1: and banking system.
• 8:12
  Speaker 2: I don't know now. It's a grab for my water.
• 8:16
  Speaker 2: And I'll share why I don't know.
• 8:22
  Speaker 1: I don't know because they're human or human woman based
• 8:25
  Speaker 1: upon their motives. And they're in control of, hopefully, of
• 8:30
  Speaker 1: that AI or super intelligent. Based on their motives, they
• 8:35
  Speaker 1: could very well do that. Let's continue. But we can
• 8:39
  Speaker 1: hardly blame anyone for being worried about this possible eventuality.
• 8:43
  Speaker 2: Better safe than sorry, right? Right.
• 8:47
  Speaker 1: You don't want to take any chances. That's sanctified common sense.
• 8:53
  Speaker 1: It is, isn't it?
• 8:54
  Speaker 2: You know it is.