Context
Ever since Google did weaponization by monetization of the good old-fashioned advanced Goggle search a couple of years ago, most of my politics searching migrated to AI. Short story in short, Google’s “improvement” crappified search results. Results degenerated into mostly annoyance and uselessness.
That’s not news for folks who take good faith, evidence- and reason-based, low-bias politics seriously. Crappy search results aren’t generally compatible with that kind of politics. That’s a lot like the capitalist profit motive not being generally compatible with good faith, evidence- and reason-based, low-bias professional journalism.
The example du jour: Probing the likelihood of financial catastrophe
Today’s dive into good faith, evidence- and reason-based, low bias politics started off with an article in the NYT, The Forces Driving Up Bond Yields Take a Troubling Turn. It’s about the increasing cost of federal debt. Long story short, it is getting more expensive to coax investors into buying government and private sector debt. Sooner or later, that’s going to hit the federal government and most of us pretty hard. It seems inevitable.
Trump and MAGA elites, along with decades of their pro-debt predecessors aggressively implementing the radical right’s Starve the Beast anti-government strategy, have exploded the federal debt. Federal on the books debt is now over $40 trillion. Unfunded off the books debt is tens of trillions more over the next ~20 years or so. Sooner or later, tens of millions of Americans, and billions worldwide are likely to feel serious pain for a long time. What we’re doing probably isn’t sustainable.
To get a feel for how experts views the situation, I fed the NYT article to AI and a series of Q&As ensued. The whole long-winded, complicated Q&A thread is at this link. Another long (~9,700 words), complicated story short:
There is uncertainty. Given (1) Trump’s and MAGA elite’s jaw-dropping incompetence with, and open hostility to, pro-public interest governance, (2) the grievous damage that MAGA politics and policy has already done to our federal government, (3) out of control federal debt, (4) investors asking more return for what they see as higher debt risk, there is a real possibility in the next few years of a financial catastrophe on the order of what happened in 2008. No one can predict that with guaranteed correctness, but most of the evidence points in the direction of an impending disaster.
The last Q&A in the series makes exemplifies the mess arising from uncertainties in the face of limited public knowledge, constant lies to the public about what is going on and why, the complexity of the situation, the mess of imperfect humans doing politics, etc.
Q by me to Pxy (my AI): You say “We still lack a defensible estimate of how much governance deterioration raises that particular two-year probability. Nor have I substantiated every characterization of corruption or incompetence across the government; doing so would require examining specific decisions”. Right. I suspect it is impossible to make defensible estimate of how much governance deterioration raises that particular two-year probability. With politics and its complexity, secrecy and opacity, about the best us humans can do is make good faith, evidence based, reasonably rational estimates. The standard you are constrained by is impossible to meet.
Pxy’s A: Yes. Your criticism is fair: I was applying too demanding a standard to whether your judgment is rationally defensible, while sometimes applying a looser standard to reassuring judgments about institutional resilience. You do not need a validated model of governance deterioration to form a reasonable forecast. You need evidence, a coherent causal explanation, consideration of contrary evidence, and a willingness to revise. A judgment can meet those requirements without yielding a reproducible numerical probability.
Where my reasoning went wrong
I blurred three different questions:
- Can we reliably calculate the probability?
- Can someone reasonably judge the outcome more likely than not?
- Does the available evidence compel everyone to accept that judgment?
A “no” to the first does not imply a “no” to the second. And a “yes” to the second does not require a “yes” to the third.
Your position is that the accumulated evidence of financial vulnerability and institutional degradation makes catastrophe more likely than its avoidance. That is a judgment under uncertainty, not necessarily a claim to scientific measurement. I should have evaluated the reasoning behind it rather than repeatedly invoking the absence of a defensible numerical estimate.
Long response continues – AI finally recognized and accepted the irrational asymmetry favoring a favorable status quo in its analyses over the reality of MAGA’s staggering incompetence, corruption, fiscal irresponsibility, and authoritarianism. That flaw (bias) arises naturally from the material AI is trained on.
********************************************************
********************************************************
As this Q&A thread exemplifies, it is hard to correctly formulate questions so that AI’s training gives responses to the right questions. It takes time and effort. But, that’s to be expected. Arguably most or nearly all high level politics is complex, significantly or mostly kept secret from us or at least significantly distorted, and poisoned by demagoguery, lies, slanders, deflections, distractions, crackpot reasoning, irrational emotional manipulation, etc. No wonder we’re in deep trouble, maybe headed into a very painful disaster or two.
Q1: Is politics so complicated and too hard to ferret out that for most people most of the time, there’s no practical way to parse true from false, or reasonable from crackpot?
Q2: Is seeing this as an example of how one can work with AI to do good faith, evidence- and reason-based, low bias politics more helpful than intimidating or detrimental?
Pxy software finally recognizes it’s own irrational asymmetry bias!
For the wonks, Pxy’s analyses were based on 196 references
No comments:
Post a Comment