Latest / The Morning Rundown / Pentagon, OpenAI, and the AI War Game You Didn’t See Coming
Transcript
- 0:13[Maya] Skip it.
- 0:14[David] Morning,
- 0:15[David] everybody.
- 0:16[David] This is the morning rundown. Grab your coffee. We've got a lot to untangle.
- 0:21[Speaker 3] Yeah,
- 0:21[Speaker 3] buckle up. Today we're digging into OpenAI's big Pentagon deal,
- 0:25[Speaker 3] and why Sam Altman is suddenly calling parts of it sloppy.
- 0:29[David] Which, like, if the guy running the company thinks it was sloppy,
- 0:33[David] that should raise some flags,
- 0:34[David] especially when we're talking national security.
- 0:37[Speaker 3] Exactly.
- 0:38[Speaker 3] We'll break down what the military says it can and can't do with these tools.
- 0:42[Speaker 3] Decision support versus anything close to pulling the trigger.
- 0:46[David] And then we zoom out to the ethics,
- 0:49[David] how you keep a strong U.S. military edge without handing life or death calls to an algorithm.
- 0:55[Speaker 3] Because,
- 0:56[Speaker 3] bottom line,
- 0:56[Speaker 3] Generals and Voters should be in charge here,
- 0:59[Speaker 3] not a handful of tech execs in San Francisco.
- 1:01[David] Right. And we'll follow the money too.
- 1:04[David] Why big tech is suddenly lining up for Pentagon contracts and what real oversight should actually look like.
- 1:12[Speaker 3] So if you care about both keeping America safe and keeping AI on a leash,
- 1:17[Speaker 3] this one's for you.
- 1:18[David] All right. Let's get into it, starting with what this OpenAI.
- 1:22[David] A Pentagon deal actually does and what it absolutely should not be allowed to do.
- 1:28[David] Okay, so headline of the morning.
- 1:31[David] The Guardian says OpenAI had a, quote,
- 1:34[David] sloppy Pentagon deal and quietly walked parts of it back after people freaked out.
- 1:38[Speaker 3] Yeah,
- 1:39[Speaker 3] big mix of AI panic and government incompetence right there.
- 1:43[Speaker 3] So, um,
- 1:44[Speaker 3] plain English,
- 1:44[Speaker 3] the Pentagon wanted to use OpenAI's tech,
- 1:47[Speaker 3] think ChatGPT-style models and image tools for defense work.
- 1:52[David] Right,
- 1:52[David] like helping analysts sift intel,
- 1:55[David] generate reports,
- 1:55[David] maybe translate stuff,
- 1:57[David] not just
- 1:57[David] Not just cap memes.
- 1:58[Speaker 3] Exactly. And OpenAI has this public rule,
- 2:01[Speaker 3] no building weapons,
- 2:02[Speaker 3] no tools that directly kill people,
- 2:05[Speaker 3] but they signed a broad research agreement with the Defense Department that honestly sounded
- 2:10[Speaker 3] pretty open-ended.
- 2:12[David] And that's where it blew up, because once people saw OpenAI plus Pentagon,
- 2:17[David] they jumped straight to killer robots,
- 2:19[David] even if that's not what was on paper.
- 2:21[Speaker 3] Totally.
- 2:22[Speaker 3] And Altman comes out and says,
- 2:23[Speaker 3] yeah,
- 2:24[Speaker 3] this was sloppy,
- 2:25[Speaker 3] which is not what you usually hear from a CEO about a defense deal.
- 2:31[David] No, usually it's we're proud to partner with the Department of Defense in a glossy video.
- 2:35[David] So when he calls it sloppy,
- 2:37[David] David,
- 2:38[David] what does he actually mean?
- 2:39[Speaker 3] From what we know, a couple things.
- 2:41[Speaker 3] One, the agreement was signed fast, with vague language.
- 2:44[Speaker 3] Two,
- 2:44[Speaker 3] OpenAI didn't clearly explain to the public or even to some employees what the Pentagon could and couldn't do with the tech.
- 2:52[David] Right. So basically they moved at startup speed on something that probably needed like Pentagon speed plus 10 lawyers.
- 3:06[David] So in the next part,
- 3:08[David] let's really dig into that line decision support versus lethal autonomy.
- 3:14[David] What does that actually look like in practice?
- 3:17[Speaker 3] And how you can be pro-strong defense,
- 3:20[Speaker 3] pro-deterrence, but still say we do not outsource life or death calls to a black box.
- 3:26[David] Stay with us, because that's where this gets both technical and very human.
- 3:33[David] Okay,
- 3:33[David] so we set up the OpenAI Pentagon deal earlier.
- 3:37[David] Let's zoom out to the core question.
- 3:39[David] What
- 3:39[Speaker 3] Mm
- 3:39[David] kind
- 3:40[Speaker 3] hmm.
- 3:40[David] of AI are we okay with on the battlefield?
- 3:43[Speaker 3] Right.
- 3:43[Speaker 3] At the simplest level, there's AI as a decision support tool and then AI that actually makes lethal decisions.
- 3:51[Speaker 3] That line is huge.
- 3:54[David] So, like, decision support is,
- 3:57[David] here's the satellite image,
- 3:59[David] here's likely enemy positions,
- 4:01[David] here are three options.
- 4:02[David] But a human commander still pulls the trigger.
- 4:06[Speaker 3] Exactly.
- 4:07[Speaker 3] The human is legally and morally on the hook.
- 4:10[Speaker 3] Once the AI is selecting and firing on targets by itself,
- 4:13[Speaker 3] no human in the loop,
- 4:14[Speaker 3] that's where you get into what a lot of people call killer robots.
- 4:19[David] And, to be clear,
- 4:21[David] most Americans are not signing up for Skynet.
- 4:29[David] So, next up, let's follow the money a bit, because these companies aren't in this out of pure patriotism.
- 4:38[Speaker 3] Yeah,
- 4:38[Speaker 3] we'll get into why defense work is suddenly the hottest thing in Silicon Valley,
- 4:42[Speaker 3] and who's actually calling the shots.
- 4:44[David] Stay with us.
- 4:47[David] Okay,
- 4:47[David] so we talked about guardrails and red lines.
- 4:51[David] Now I want to zoom in on the money and power behind this.
- 4:54[Speaker 3] Yeah,
- 4:55[Speaker 3] follow the incentives,
- 4:56[Speaker 3] right?
- 4:56[Speaker 3] Once you do that,
- 4:57[Speaker 3] the picture looks a lot less innocent.
- 4:59[David] Exactly.
- 5:00[David] OpenAI,
- 5:01[David] Google,
- 5:02[David] Microsoft,
- 5:03[David] these companies aren't doing defense work out of pure patriotism. There are huge contracts on the table,
- 5:09[David] long-term government ties,
- 5:11[David] and a ton of influence.
- 5:14[Speaker 3] And reliable revenue.
- 5:16[Speaker 3] A Pentagon contract isn't like a random app subscription.
- 5:20[Speaker 3] It's multi-year, often billions of dollars, and it signals to Wall Street we're entrenched in the state.
- 5:26[Speaker 3] That boosts valuation,
- 5:28[Speaker 3] attracts investors.
- 5:29[David] Right. And from a tech industry angle,
- 5:32[David] once you're baked into government systems,
- 5:34[David] it's really hard to unwind.
- 5:36[Speaker 3] Right.
- 5:36[David] You become the default vendor.
- 5:39[Speaker 3] Which means you also get a say, informally at least, in how the rules are written.
- 5:43[Speaker 3] You're advising on policy that just happens to benefit your own products.
- 5:48[David] Shocked. Totally shocked.
- 5:51[David] Okay,
- 5:52[David] so, and there's the data angle.
- 5:54[David] Even if they're not slurping up classified info directly,
- 5:57[David] just being close to military use cases teach them what to build,
- 6:02[David] what to optimize.
- 6:04[Speaker 3] Yeah,
- 6:04[Speaker 3] the feedback loop.
- 6:05[Speaker 3] You learn where the pain points are in Intel,
- 6:07[Speaker 3] logistics,
- 6:08[Speaker 3] targeting,
- 6:09[Speaker 3] whatever the mission is.
- 6:10[Speaker 3] That knowledge is gold.
- 6:11[Speaker 3] You can then sell versions of those tools to other governments or to big corporations.
- 6:17[David] And meanwhile,
- 6:18[David] the press releases are all like,
- 6:19[David] we just want to help keep people safe,
- 6:21[David] Which fine,
- 6:23[David] but let's not pretend the incentives stop there.
- 6:26[Speaker 3] This is where the accountability question kicks in: Who actually decides how far this goes?
- 6:32[Speaker 3] Is it Congress,
- 6:33[Speaker 3] like we argued it should be?
- 6:35[Speaker 3] Is it the Pentagon?
- 6:36[Speaker 3] Or is it basically a handful of Silicon Valley boards and these weird non-profit shells around them?
- 6:49[David] Sunlight and receipts, that's the vibe.
- 6:51[Speaker 3] So as this all unfolds, new AI tools,
- 6:55[Speaker 3] new defense deals,
- 6:55[Speaker 3] the question to keep asking is who benefits,
- 6:59[Speaker 3] who decides, and who can say no?
- 7:02[David] Exactly. If the answer to that last one is no one,
- 7:05[David] that's when we've got a problem.
- 7:07[Speaker 3] And that's the Morning Rundown.
- 7:08[David] Thanks for starting your day with us. We'll see you tomorrow.
- 7:12[Speaker 3] Alright,
- 7:13[Speaker 3] that's our time.
- 7:14[Speaker 3] If you remember one thing today,
- 7:16[Speaker 3] it's this:
- 7:17[Speaker 3] AI in the military can help our troops and deter bad guys,
- 7:20[Speaker 3] but it needs clear,
- 7:22[Speaker 3] tough rules set by Congress,
- 7:24[Speaker 3] not just Sam Altman and the Pentagon cutting,
- 7:26[Speaker 3] um,
- 7:27[Speaker 3] sloppy deals.
- 7:28[Speaker 4] Mm-hmm.
- 7:29[David] Exactly. And like, anytime you hear about a rushed government big tech partnership with fuzzy language,
- 7:36[David] that's your cue to pay attention and ask hard questions.
- 7:39[Speaker 3] Yeah, you don't need a clearance badge to spot red flags.
- 7:43[Speaker 3] You just need to, you know, listen closely.
- 7:46[David] If this helped make sense of it,
- 7:48[David] hit subscribe,
- 7:49[David] drop a quick review,
- 7:50[David] and share the morning rundown with a friend.
- 7:52[David] Thanks for starting your day with us.
- 7:54[David] We'll see you tomorrow.