Latest / Challenges of Faith Radio Program / Say what? Ep 3403

Transcript

Say what? Ep 3403 · Challenges of Faith Radio Program

• 0:00
  Speaker 1: Welcome back to Challenges of Faith Radio Program. I'm Gary McCann.
• 0:04
  Speaker 1: Thank you for joining. Here on episode 3403, say what?
• 0:12
  Speaker 1: According to Futurism, Anthropic says it stopped several instances of
• 0:19
  Speaker 1: foreign scientists using its AI agents to create possible biological weapons,
• 0:26
  Speaker 1: the New York Times reports. And I know you've been
• 0:29
  Speaker 1: watching the news, hearing about it. The alarming news was
• 0:34
  Speaker 1: revealed in a lengthy report released by the Claude Chatbot
• 0:38
  Speaker 1: maker this past Thursday. Notably, it comes amid heightened AI
• 0:44
  Speaker 1: safety concerns, again after a former anthropic employee accused the
• 0:49
  Speaker 1: company of irresponsibly building powerful AI systems despite knowing they're destructive.
• 0:59
  Speaker 1: According to the report, scientists in several countries where anthropic
• 1:03
  Speaker 1: has blocked access to its models subverted these controls and
• 1:08
  Speaker 1: tried to obfuscate the purpose of their research to invade
• 1:11
  Speaker 1: our safeguards. In one case in May, a scientist asked
• 1:15
  Speaker 1: Claude to help write an application to receive a state-sponsored
• 1:20
  Speaker 1: grant for an eyebrow-raising project to engineer more harmful mutations
• 1:26
  Speaker 1: of the mosquito-borne virus. which can cause fever and debilitating
• 1:33
  Speaker 1: joint pain. The project, Anthropic believes, was intended to be
• 1:37
  Speaker 1: carried out as a military research institute. What we don't
• 1:43
  Speaker 1: know is if the research was meant to be weaponized,
• 1:46
  Speaker 1: Jake can claim, Anthropic's head of threat intelligence told the
• 1:52
  Speaker 1: New York Times. But a military institution doing gain-of-function research
• 1:57
  Speaker 1: is concerning, he added. referring to research that seeks to
• 2:02
  Speaker 1: enhance or add to the biological capabilities of organisms and pathogens.
• 2:09
  Speaker 1: The company documented four other similar cases where scientists evaded
• 2:15
  Speaker 1: its guardrails to pursue shady biological research. It banned all
• 2:21
  Speaker 1: of the offending accounts. But as with the mosquito virus case,
• 2:28
  Speaker 1: Anthropocene couldn't definitively determine if these inquiries were part of
• 2:33
  Speaker 1: a legitimate research or an attempt to engineer bioweapons. It
• 2:39
  Speaker 1: nonetheless chose to err on the side of caution. I'm
• 2:43
  Speaker 1: glad they did. You're not seeing someone in a comic
• 2:47
  Speaker 1: book kind of way say, hey, I want to build
• 2:50
  Speaker 1: a biological weapon to kill everybody. Klein told the New
• 2:55
  Speaker 1: York Times it's an incredible nuanced situation. For years now,
• 3:00
  Speaker 1: the AI industry has warned of the risk of AI
• 3:03
  Speaker 1: being used to create bioweapons, including anthropic CEO Dario Amati.
• 3:12
  Speaker 1: In a lengthy essay published earlier this year, Amati vowed
• 3:16
  Speaker 1: that the company was constantly improving its guardrails to prevent
• 3:19
  Speaker 1: models from assisting with bioweapon research and opined at legislation
• 3:25
  Speaker 1: and international cooperations. was necessary to mitigate the risks. Still,
• 3:32
  Speaker 1: while Anthropi claims to have avoided these misuses of its models,
• 3:38
  Speaker 1: the news casts doubt on whether it and other AI
• 3:41
  Speaker 1: companies can guarantee the responsible use of its tech. By
• 3:47
  Speaker 1: the company's own admission, its safeguards failed, potentially leaving the
• 3:52
  Speaker 1: door wide open for other bad actors in the future.
• 3:57
  Speaker 1: In short, If these state actors were indeed looking to
• 4:02
  Speaker 1: craft bioweapons, it's a dramatic escalation suggesting we're edging closer
• 4:08
  Speaker 1: to a catastrophe. That's, of course, if general-purpose models like
• 4:13
  Speaker 1: Anthropix are capable enough to meaningfully assist these dangerous efforts
• 4:19
  Speaker 1: amid the heightened safety debate. The company has faced accusations
• 4:23
  Speaker 1: of fear-mongering among its own tech as a marketing tactic.
• 4:28
  Speaker 1: Some experts have the mind that the threat is very
• 4:31
  Speaker 1: much already real. They should be. Andrew Weber, a senior
• 4:36
  Speaker 1: fellow on the Council of Strategic Risk, who reviewed Anthropic's
• 4:41
  Speaker 1: report ahead of its release, told the New York Times
• 4:46
  Speaker 1: that his findings were chilling examples of state-sponsored biological weapons
• 4:52
  Speaker 1: developers tapping into the rapidly advancing capabilities of frontier AI
• 4:57
  Speaker 1: models Are you listening? Or you're running around doubtful, or
• 5:04
  Speaker 1: all caught up in your own little world, wherever your
• 5:07
  Speaker 1: world happens to be. Do you even know what's going
• 5:10
  Speaker 1: on outside your home, your place of worship, your job,
• 5:14
  Speaker 1: you name it. Do you really know what's going on?
• 5:19
  Speaker 1: You know, the question was asked, what does NVIDIA have
• 5:22
  Speaker 1: to do with AI? Well, remember last week where an
• 5:25
  Speaker 1: anthropic employee said that there's more than a 10% chance
• 5:29
  Speaker 1: that AI would destroy humanity? You remember? Remember that researcher,
• 5:35
  Speaker 1: Jacob Coxon? Remember how he said that the company was
• 5:38
  Speaker 1: racing to develop superintelligence despite knowing that civilization is at stake?
• 5:45
  Speaker 1: Now think about that. Think about whether a scientist is
• 5:49
  Speaker 1: trying to play a little G-Gun or individuals desiring power,
• 5:53
  Speaker 1: the love of money, the love of titles, all temporary.
• 5:59
  Speaker 1: But Jacob's concerns were echoed by two current employees. And
• 6:02
  Speaker 1: the question they asked, they answered, well, AI and humanity,
• 6:06
  Speaker 1: what do you think? Those of you who are of non-faith,
• 6:10
  Speaker 1: those of you who are of faith in God. Well,
• 6:14
  Speaker 1: some employees of leading AI firm Anthropic seem to think so.
• 6:17
  Speaker 1: Say what? One said that there's a 10% chance that
• 6:22
  Speaker 1: this would happen within the next decade, and another quit.
• 6:26
  Speaker 1: Not just Jacob. The company in protest that Anthropic and
• 6:30
  Speaker 1: its rival, OpenAI, are gambling with our lives. And again,
• 6:37
  Speaker 1: this comes after NVIDIA boss Jensen Hung dropped the AGI bomb,
• 6:41
  Speaker 1: claiming that OpenAI's newest model, Astra, has reached artificial general intelligence,
• 6:49
  Speaker 1: which is the next stage of AI development. The now
• 6:53
  Speaker 1: former anthropic employees said the people building AI earnestly believe
• 6:59
  Speaker 1: that technology could wipe out humanity by the end of
• 7:02
  Speaker 1: the decade. Are they still working on it if they
• 7:05
  Speaker 1: really believe it? Why are they still there? Is it
• 7:10
  Speaker 1: the love of money tied to what? If anything, many
• 7:14
  Speaker 1: executives and senior researchers will couch their phrasing in the
• 7:18
  Speaker 1: news media to sound sensible. But I hear the same
• 7:22
  Speaker 1: people expressing fear privately. Come on now, stand up. No
• 7:28
  Speaker 1: other human activity poses this level of danger, Jacob Coxon wrote.
• 7:35
  Speaker 1: AI executives have long claimed that AI will either bring superabundance,
• 7:40
  Speaker 1: or cause mass disruptions, not ruling out the destruction of humanity.
• 7:46
  Speaker 1: Sound like the actors abroad is trying to do harm
• 7:49
  Speaker 1: here in the United States believe so as well. Remember now,
• 7:52
  Speaker 1: Coxon also said that many open AI have not deeply
• 7:56
  Speaker 1: internalized the civilization states. While people at Anthropic understood the risk,
• 8:02
  Speaker 1: but are locked in a race to get there first.
• 8:06
  Speaker 1: because they think no one else is developing AI responsibly. Well,
• 8:10
  Speaker 1: that's good, then, on one hand, to be ahead of
• 8:14
  Speaker 1: your enemies, your two-legged enemies, that you know trying to
• 8:19
  Speaker 1: harm you, that you know trying to destroy you, irrespective
• 8:24
  Speaker 1: of your skin tone, irrespective of your financial status in
• 8:29
  Speaker 1: life or where you live on the globe. But what's
• 8:34
  Speaker 1: interesting is that hours after Coxon had announced his resignation
• 8:39
  Speaker 1: that another anthropic researcher, Evan Hembinger, wrote on X that
• 8:45
  Speaker 1: people in the company do earnestly believe AI could kill
• 8:49
  Speaker 1: all humans within the next decade. Now, what's that telling you? Confirmation.
• 8:54
  Speaker 1: Given the scenario of a 10% chance. He said that
• 8:57
  Speaker 1: I believe anthropic is trying its best, but we do
• 9:00
  Speaker 1: not yet have a plan. to solve a lima for
• 9:04
  Speaker 1: superintelligence and are not clearly on track to, he wrote. However,
• 9:10
  Speaker 1: neither Coxon nor Hubinger explained how superintelligence could wipe out humanity.
• 9:15
  Speaker 1: Always someone wanted to know rather than believe. And there's
• 9:19
  Speaker 1: nothing wrong with posing all those questions, but at what
• 9:22
  Speaker 1: point in time do you have enough evidence that it's true?
• 9:25
  Speaker 1: What did they used to say? You know it's true.
• 9:29
  Speaker 1: That if you walk like a duck, talk like a duck,
• 9:34
  Speaker 1: look like a duck, then you must be, I'm sure
• 9:36
  Speaker 1: somebody else is going to say something else. But tell
• 9:40
  Speaker 1: that to those trying to harm people of all skin
• 9:42
  Speaker 1: tones here in the United States. And then you had
• 9:45
  Speaker 1: another current anthropic employee, Samuel Marks, who works on safety research.
• 9:51
  Speaker 1: He joined a discussion on X saying that the more
• 9:54
  Speaker 1: senior the developer, the more concerned they are that AI
• 9:58
  Speaker 1: could cause human extinction in the next few years. Marx
• 10:02
  Speaker 1: explained that companies continue on developing risky AI due to
• 10:06
  Speaker 1: a mixture of commercial incentives and a belief that they
• 10:10
  Speaker 1: are in a race with other less responsible AI developers.
• 10:14
  Speaker 1: Coxon's warnings It struck a chord with former employees of
• 10:19
  Speaker 1: other AI labs. Notice that. All it took was one
• 10:25
  Speaker 1: person to stand up and speak out. Alex Turner, a
• 10:30
  Speaker 1: former research scientist at Google DeepMind, who left the company
• 10:34
  Speaker 1: in June 2026. Remember now, it was only September. He
• 10:38
  Speaker 1: confirmed that many researchers believe they're building something that could
• 10:41
  Speaker 1: kill everyone on the planet, and they're still building it.
• 10:46
  Speaker 1: It was literally my last day on the job to
• 10:49
  Speaker 1: think about how to stop that, Turner wrote in his ex-post.
• 10:53
  Speaker 1: This isn't the first very public departure from anthropic over
• 10:57
  Speaker 1: security concerns. Now, I know you were listening to the
• 11:00
  Speaker 1: news a couple days ago where, from a governmental standpoint,
• 11:04
  Speaker 1: they spoke about anthropic and something that they had to
• 11:07
  Speaker 1: step in and stop. But are you paying attention? Say what? Yeah,
• 11:13
  Speaker 1: are you? And you paying attention, or are you being distracted? Well,
• 11:21
  Speaker 1: you know the answer, don't you? I know you do.
• 11:25
  Speaker 1: You do, don't you?