Aggressively Pedantic

It doesn't change the fundamental problem that the Army knows neither how to describe what it wants nor how to evaluate what it receives. Of *course* you're going to get bespoke systems that dramatically overrun cost and schedule if your spec is "we need an AI that can help us win wars"

If you want it bad, you get it bad!

It used to make me laugh when people would say something like "Kim Jong Il is an irrational actor." There is no such thing as objective rationality per se - Kim Jong Il didn't apply the same logic in making decisions that we did, but that doesn't mean his decisions were devoid of logic

I caught you a dollar dot gif

But you're both right - at some point, you can't attract and retain people who can make a minimum of 2-3x their public servant salary to do perhaps less work with a lot less public scrutiny with the claim that it's "meaningful." And certainly not in an environment where your service is openly mocked

...and there are large numbers of people (myself among them, I think) whose answer has long been "well, more than $0." It's like the scene in The Right Stuff where Yeager turns down additional Bell money to fly the X-1 because "the government is already paying me."

Hard agree with both! My point is not that pay cannot or shouldn't be increased - it's that the only way to make even a slight shortfall palatable is to cultivate the profession as a calling. "How much pay are you willing to sacrifice to achieve self actualization" isn't an improper question...

By the same token, though, treating public servants like they must either be incompetent (because they'd make more on the outside if they weren't) or fools (for the same reason) makes it hard to want to stay. If public service isn't a calling but "just another job" it totally loses its unique appeal

You really have to grow them, in my opinion, by grabbing them in college and then building a belief in the fundamental goodness of public service. It's tough (almost impossible) to persuade a mid career professional not already proposed to public service to take a massive pay cut pro bono populi.

I am growing in my belief that at least part of the idea is to get feds to voluntarily separate so that you don't have to pay 'em severance.

More generally, I always laugh when someone says "we'll just have the AI run billions of models." Models of what? Underpinned by what logic and mathematics?

"Model combat" means about a bazillion different things.

And so we delude ourselves into believing that *because* the data is co-located and *because* there is some mystical AI doing work in the background, it *must* be telling us things we need to know!

And that's just asinine.

A big part of the problem IMHO is the willingness to believe that Palantir et Al are doing more than they actually are. The UI is fine; the idea of improving access to more disparately developed datastreams is good! But it's insufficient if you don't know what questions you're trying to answer.

...and what you get is a populace eager and willing to believe the simplest explanation for a set of problems they require vanishingly little evidence to confirm the existence thereof.

Is it stupidity per se? I'm not convinced - I think it's an utterly rational approach, but with a (shockingly) low threshold for proof and a (shockingly) high trust in the fidelity of public statements.

Layer atop that a fundamental disinclination to engage with the complex...

You *could* be arguing in your spare time, though.

My point isn't about Palantir's value (monetary or otherwise) as a company not about Thiel's conception of ethics, social responsibility, etc.

Palantir's product is...fine. IMHO it's a slick front end that isn't particularly special if you're a reasonably competent technical analyst.

Both elements are anathema to military organizations - the advertising because of bureaucratic advantage (always better to be the one with the secret sauce!); the sharing because of OPSEC.

In theory it's useful to enable more data sharing across more organizations that aren't aware the data exist. That's good! But only if you make the theory reality - which means advertising your data and giving other people access to it.

So the thing about all this is that Palantir's product isn't bad per se - it's that people see it (and Palantir doesn't correct them...) as a replacement for human cognition. You can amass all the data you want - but if you aren't sure what you're trying to do with it, you're no better off.

I'd be much more interested in their operations research analysis on what is driving cost ballooning and how their process can minimize it! That'd be a practical and useful product!

The most baffling thing about all of this to me is that, even if you stipulate that the physics and engineering problems are solvable...they're asserting (without any semblance of proof, assumption or otherwise) that the solution will be an order of magnitude cheaper!

Indeed, the ends famously justify the means!in all cases!

I'm that case I'm gonna wow the world with a new idea for, like, adding two numbers together. I need about $20bn to flesh it out - don't miss this opportunity to get in on the ground floor!

That's the context I probably missed - I didn't pick up that the OP was trying to...I guess broker new money into a space already flooded?

FWIW I'm a mathematician by academic background - so I always *assume* people are seeking the most generalizable solutions.

It's a bad assumption, of course - especially for VC types who tend to seek exquisite solutions and then convince everyone they're solving all the problems...

I don't doubt it; in fact, I'd kinda have expected VC to already be doing this kind of thing independently.

Serious counterpoint: this *could* be viable *if* disconnected from ordinary DoD/CCMD processes *and* incentivize specifically to yield solutions robust and resilient to multiple (perhaps unforseen) problems.

In practice, we'll just get MOAR DRONES but I could see the theoretical utility.

I should clarify further that TWN will to fight not an actually random variable - but it may not be categorically knowable (depending on your frame of reference) and is a function of a huge number of factors, many subjective.

I'd give a more comprehensive scientific treatment if I weren't character limited 🤓 Suffice it to say that I have a lot of time right now to contemplate all of the permutations of what could be studied and how and to what ends...