On your honor
A lawyer is in hot water with the US District Court for southern New York, because he used ChatGPT as a research assistant. It produced court filings that contained several bogus legal references. ChatGPT was pretty baldfaced in defending its work:
Schwartz provided an excerpt from ChatGPT queries in which he asked the AI tool whether Varghese is a real case. ChatGPT answered that it “is a real case” and “can be found on legal research databases such as Westlaw and LexisNexis.” When asked if the other cases provided by ChatGPT are fake, it answered, “No, the other cases I provided are real and can be found in reputable legal databases such as LexisNexis and Westlaw.”
Needless to say, that doesn’t let the lawyer off the hook.
The human approach to trust is to assume that most people will be fairly honest. We may seek references and keep a wary eye in dealing with someone unknown, on important matters. Most institutional policies are reactive and punitive. An employee who lies may be fired. A lawyer or CPA who lies in their professional capacity may face professional sanctions. Anyone who lies under specific circumstances the law defines may face criminal prosecution. Elizabeth Holmes is now in federal prison, for her lies. Court filings and statements to public officials are circumstances where intentional lies create legal exposure. I expect events this coming summer to highlight that.
That approach clearly does not extend to AI. And it is not clear what approach would. Which means, for the foreseeable future, positions of trust will continue to be filled only by humans. Those in such positions might use AI for research and other kinds of assistance. I have no doubt that the technology will be tuned to be more reliable. Where that is needed. Alas, it will be tuned to be more slickly duplicitous where spammers and scammers find that useful. But it will remain a human who puts their name to court filings, who deposes witnesses, who takes on legal clients, who signs corporate accounting statements, who acts in a fiduciary capacity. In all such cases, who will be held accountable for using AI responsibly. Trust, not intelligence alone, is what we require in many positions. Trust is not a simple thing, and in many roles requires intelligence reflecting on human relationships.
I suspect Schwartz will suffer little more than a reprimand, given that he seems to have acted in good faith, merely blundering in the use of new technology.
Anatomy of a scam
The New York Times has a deep dive into a trio of spammers who raise money, using pro-police, pro-firefighter, and pro-veteran messaging. More than 99% of what they raise goes to their own organizations to raise more money. And to their own pockets.
The spammers are Republican operatives who learned early in life that there is no need to spend funds raised on the purpose promised. Their fellow conservatives are an easily exploited target group.
- They are suckers for affiliate marketing. They will buy everything from coffee to vitamin supplements if they think it comes from their “own people.” Or boosts favored groups (veterans, police). Or damns hated groups (liberals, gays, the woke). Or even if it just is festooned in American flags, guns, and Christian symbols.
- They have poor economic understanding. They misunderstand the kinds of problems that charity can solve, and the incentives on the organizations involved.
- They misplace their trust. They trust those who tell them what they want to hear. They don’t understand actual research. They have a shallow view of expertise.
- Those failings are amplified by their media and churches. Many of the large evangelical churches are better understood as marketing organizations than as religious ones, taking advantage of the very flaws above to relieve their flocks of cash. Outlets such as Fox News consciously peddle lies to keep their audience.
I imagine these spammers asking why anyone should care, if they want to shear their own sheep? They might defend themselves by arguing if they didn’t relieve their targets of their money, those targets likely would spend it on something equally stupid. If one imagines those dollars instead going to an evangelical preacher, it’s tough to gainsay that point. But, if one sees those preachers and spammers as playing similar roles, working to make their dupes poorer and even dumber, it is straightforward to damn them both.
Some of those who acted criminally in the January 6th storming of the Capitol are trying to get in on the take. The path from dupe to grifter is short and easy. Prosecutors are tapping the brakes on their ability to raise money from their crimes.
I’ll take the bourbon
Molly Shah writes a bit about Kentucky, including more than you ever wanted to know about Colonel Sanders.
Of Lake and courts and whistle blowers
Those caught up in MAGA see some equivalency between themselves and those outside. They have their own narratives. Other people have theirs. They have their own whistle blowers. They have their own evidence. They have their own media. If someone reads superficially, MAGA might seem not much different from other political viewpoints. Especially on social media. “See? We’re not a cult.”
That pretend equivalency doesn’t play well in the American legal system. Courts ask specific questions. They have rules of evidence. Narrative argument must be built on evidence presented, and evidence must be relevant to the question at issue. Witnesses and evidence are examined by both sides.
All of which is kryptonite to MAGA bullshit. When actually examined on the stand, Kari Lake’s own whistle blowers and own expert witness undermined her remaining claim about Arizona’s gubernatorial election being stolen. So it comes as no surprise that she has lost at trial.
Again.
MAGA actors generally lose in the courts. Again. And again. And again.
Yes, I have every expectation that I will write future versions of this post, as more MAGA leaders face their own day in court. Photo shows Hiram Edson, who gave the Millerites a way to keep believing, after the Great Disappointment. Which serves as one of many examples that group stupidity was a problem long before social media.
America’s housing problem
Americans want cheaper housing. When they buy, not when they sell. And they don’t want it built in their current neighborhood. The incentives that drive local housing policies work across the political spectrum. Which is why the problem we associate with California actually is an American problem.
Durham’s fizzle
Given what was known about the origin and the route, there seems little surprise in the result.
American life expectancy
There is a glaring difference in life expectancy across the different regions of the US. The map right is cleverly colored blue and red, aligning with the correlated politics. The blue counties are where people live longer.
The cause of that gap is not just deaths of despair, from drugs and alcohol and suicide. It’s not just diabetes. It’s not just those in red counties avoiding vaccines, and the other pernicious health influences that come from being “red pilled.” It’s not just gun deaths. It’s all of those, and more combined.
Monica Potts writes about how she left Clinton, Arkansas, while many of her girlhood friends did not.
They want a bully
Paul Waldman pegs how Trump enthralls his MAGA following. I urge everyone to think on that.
David Rothkopf writes similarly:
The MAGA-ized GOP has only gotten more outrageous… which is to say more committed to the politics of outrage, to obliterating norms of decency, as the signature activities to which they devote themselves. They do it because their base is angry. They do it because it drives social media wild. They do it because they have no ideas and their leaders are profoundly immoral, disgusting people.
The MAGA crowd is invigorated by the same behavior that repulses anyone honest. The same behavior. The rest of us need to keep in mind what game they are playing and to call it out. Don’t let them side step.
Zelenskyy is not fooled by Trump’s bluster.
Parroting the master of lies, George Santos goes on the offensive in response to his own federal indictments, labeling the investigation a “witch hunt” and using it as opportunity to attack his political opponents. Unlike Trump, Santos hasn’t yet won the devotion of MAGA. He clearly is paying attention to the techniques, eager to learn.
Both master and student may discover that however much those tactics are useful in currying a cultic following, they don’t serve well in court.
Doing it backwards
Google is planning to mark its AI generated images to make it easier to identify them as such.
Well, that’s backwards.
The working assumption today should be that every digital image is fake, unless one can demonstrate its provenance. Instead of trying to identify fakes, which will come from all sorts of sources, what Google should do is work on proving authenticity. That is especially important regarding cellphones, which have become the most common device we use to capture audio and visual data. Google should embed a unique private key in every Pixel phone, with certified and published public key, that the phone optionally uses to sign its images, sound recordings, and video. The signature should validate the metadata — including time and place — as well as the captured data. Google should work with Apple and others to evolve related industry standards.
“This photo provably was taken by a Pixel 4 on August 17th, 2022 at the corner of S. Lamar and Toomey” is a much stronger statement than “we can’t prove it’s fake.” Backed by appropriate cryptographic signature, the first is a statement that will hold up in court.
Back bends are a good exercise. For the body, not for the design of technical standards.
Ob disclaimer: Yeah, I’m still long.
Dead reckoning
Dead reckoning is the most basic form of navigation. Someone who a) knows where they were at some time past, and b) knows the path and velocity they are traveling, can calculate where they are now. Or will be, at some point in the future. There are a few problems with DR. First, it takes constant vigilance. The navigator who fails to account for a change in course or speed when it is made has to guess what the path was. Second, one may have only rough measures of those, especially when current is involved. Third and most importantly, all those errors accumulate. The circle of uncertainty gets ever larger. Which is why the navigator needs periodic fixes by other means.
A smartphone that could do accurate DR over long distances would be a wondrous thing. One might ask why that is important when GPS now provides a constant fix? Because it is a poor navigator who relies on only one method to find where they are. I trust when this hardware makes its way into digital devices, Garmin software will make smare use of it.
The cellphone is today’s necessary piece of tech. Especially for those on the move. Even without cellular service, it is one’s set of maps, library, pencil, note pad, calculator, post office, and navigation, all rolled into one.