The triumph of form over content

That’s all ChatGPT is. Emily Bender explains.

When you read the output of ChatGPT, it’s important to remember that despite its apparent fluency and despite its ability to create confident sounding strings that are on topic and seem like answers to your questions, it’s only manipulating linguistic form. It’s not understanding what you asked nor what it’s answering, let alone “reasoning” from your question + its “knowledge” to come up with the answer. The only knowledge it has is knowledge of distribution of linguistic form.

It doesn’t matter how “intelligent” it is — it can’t get to meaning if all it has access to is form. But also: it’s not “intelligent”. Our only evidence for its “intelligence” is the apparent coherence of its output. But we’re the ones doing all the meaning making there, as we make sense of it.

I think we know this from how we learn language ourselves. Babies don’t lie there with their eyes closed processing sounds without context — they are associating and integrating sounds with a complex environment, and also with internal states that are responsive to external cues. Clearly what we need to do is imbed ChatGPT in a device that gets hungry and craps itself and needs constant attention from a human.

Oh no…someone, somewhere is about to wrap a diaper around a server.

Ken Ham is right!

I bet you never thought you’d see me writing that, did you? Ham has always insisted on a rigidly applied set of doctrines that may not be questioned — any doubt about any one Christian claim will lead to the whole house of cards collapsing. The Washington Post demonstrates the validity of that argument with this article, The Revolt of Christian Homeschoolers.

It’s about a couple in Virginia, Christina and Aaron Beall, who were brought up in an ultraconservative faith, with all the usual restrictions: women will submit to the man, public schools were evil, contraception was bad (she got pregnant within weeks of getting married.)

Aaron had grown up believing Christians could out-populate atheists and Muslims by scorning birth control; Christina had been taught the Bible-based arithmetic necessary to calculate the age of a universe less than 8,000 years old. Their education was one in which dinosaurs were herded aboard Noah’s ark — and in which the penalty for doubt or disobedience was swift. Sometimes they still flinched when they remembered their parents’ literal adherence to the words of the Old Testament: “Do not withhold correction from a child, for if you beat him with a rod, he will not die.”

What broke their faith was that last bit: the rod. They could not beat their children. And if whupping their kids with a stick was wrong, what else could be wrong about their faith? So they enrolled their daughter in <gasp> public school. She thrived and was happy. And they discovered that their religion had lied to them about schooling.

“People who think the public schools are indoctrinating don’t know what indoctrination is. We were indoctrinated,” Aaron says. “It’s not even comparable.”

They kept on learning.

Her loss of faith in the biblical literalism and patriarchal values of her childhood was coming in the way the movement’s adherents had always warned it would: through exposure to people with different experiences and points of view.

Those people just happened to be her daughter and her husband.

The article says they didn’t question Christianity, so they didn’t become godless atheists or anything horrible like that, but they did become significantly more open-minded and are now reading more than the Bible and awful books like Bill Gothard’s or Michael and Debi Pearl’s. Now look at what they’re reading.

Stacks of books on the living room’s end tables testified to their belated efforts at self-education: popular works by the biologists Neil Shubin and Robert Sapolsky, as well as “Raising Critical Thinkers” by Julie Bogart, a leading developer of home education materials who has criticized conservative Christian home-schooling groups.

Poor Ken Ham. He was right that any lapse in dedication to his interpretation of Christianity would lead to apostasy, but he’s probably sitting in a dark corner, hissing and gnashing his teeth and flicking his forked tail if he hears about the Bealls. One family has escaped his grasp!

We hunted the rabbit

I know it’s not as impressive as the mammoth, but it gives AMAB people an edge in rugby, therefore trans athletes should be banned. So saith Sean Ingle, chief sportswriter for the increasingly transphobic Guardian.

As he repeated many lies about trans women in sport, whether through ignorance or malice, Ingle said, And going back to the start with the science is to have a separate, exclusive, preserved category for natal females with trans women and trans men then going into an open, universal category. And those that support this approach point to the recent science that suggests that even when testosterone is reduced, strength in transgender women only goes down 5%.

Most of that advantage for male puberty is retained. They also point out, and I hear this a lot, that women are not men with lower testosterone. They point out there are thousands of physical differences between males and females, and they aren’t always obvious.

Females tend to have better peripheral vision than males. Males, in contrast, are quite as fast[sic?] at accurately detecting the trajectory of a moving object. That is, how fast it’s moving, in which direction it’s moving, and where it’s going to be 1 second from now.

That’s helpful when you’re trying to chuck a spear at a rabbit. If you’re going back to evolutionary biology times, it’s also helpful when you’re trying to intercept a rugby ball. My general view here is that The Guardian should be at the heart of all this and that we should write about the subject fearlessly.

Ah, even sportswriters have absorbed the biases of evolutionary psychology. Now men, not women, have evolved to be better at throwing spears.

These glib comparisons always make me wonder what was being compared in these studies. All women tend to have better peripheral vision than men? What if you compared men, in general, to women tennis players? Is it still true? Isn’t it quite likely that peripheral vision, and the ability to calculate trajectories, are plastic and responsive to practice?

Also, how large is the variation within men, and within women? Aren’t we really dealing with selected subsets of populations, making blanket claims about the aggregate abilities of diverse populations rather problematic?

The whole premise is flawed. It assumes that men of the paleolithic were specifically and exclusively selected for spear chucking, that women of that time had no use for that talent, and that some epigenetic factor inhibits the genetic spear-chucking complex in women. No evidence for any of that. Then we have to assume that there was no further selection for or against that complex for 100,000 years — men retained a fairly specific ability through many generations of life farming. Then we assume further that whatever epigenetic modifiers allow for enhanced spearchucking in men, they don’t include things like testosterone that might be blocked by inhibitors — these hypothetical male advantages sail through everything that affects trans women unaffected.

But sure, if you’re an evolutionary psychologist sportswriter, you can just propose that whole chain of improbabilities as a given and call it “science” or “biology,” all in the name of transphobia. I call it magical thinking.

The administration has destroyed all my lesson plans!

I was looking forward to a mellow, relaxed Fall term, when the office of human resources sent me this memo:

Dear Faculty and Staff,

As you may have read in the news, Governor Walz recently signed a bill into law legalizing the use of marijuana for recreational purposes by adults age 21 and over as well as the right to grow up to eight plants for personal use. The new law will also decriminalize its possession and expunge misdemeanor marijuana offenses from criminal records.

While the new law may begin to shift societal norms outside the workplace, University employees will still be expected to follow the University’s Drug-Free policy, Smoke-and Tobacco-Free policy, and Code of Conduct, which outlines the responsibilities of employees to “act according to the highest ethical and professional standards of conduct” and “be personally accountable for individual actions.”

As has been the standard procedure in the past, employees who violate these policies and perform work while impaired due to the use of controlled substances are subject to disciplinary action, including termination. Additionally, while the new law allows for the growing of cannabis plants at home, no such plants grown for personal use are allowed on any University property nor should such plants be visible on camera during remote work. Please also note that the legalization of marijuana at the state level does not change the federal DOT Drug policy for drivers, of which the University has about 100 employees. All members of our University community are responsible for encouraging compliance with these policies.

If you, or someone you are concerned about, need help with substance misuse, the University has resources available to you, including confidential counseling and chemical health consultation through the Employee Assistance Program and mental health resources for University employees and their dependents enrolled in the medical plan.

We appreciate your awareness and compliance with this issue in support of protecting the health of our entire community.

Darn. I guess the backup plan is MDMA.

(No, actually, I don’t partake of anything anymore, not even alcohol. I have no idea what the students will be doing to make my classes tolerable.)

This is Lawrence Krauss’s career now

He’s fallen far now, and seems to think that thrashing about in the muck will raise him up, rather than make him dirtier. He’s got an op-ed in the Wall Street Journal 🤮, titled “A Scientist’s Sexuality Shouldn’t Matter”. I agree, it shouldn’t, but we’re dealing with facts here rather than a disgruntled conservative’s feelings, and it does matter, unfortunately.

Krauss is upset because federal grant agencies ask applicants about various bits of demographic data.

The Survey of Earned Doctorates is an annual census of new postgraduate research degrees. The National Science Foundation, a federal agency, collects data on academic discipline, sex, race, ethnicity, debt burden, disability and citizenship. The results are used by government, universities and industry to track the demographics of women and minorities in STEM — science, technology, engineering and math.

Well, yes, it’s a survey. That’s what surveys do. Krauss wants them to stop because — wait for it — sexism is over! According to a notorious sex pest, that is.

The sex and race data — the latter has been collected since 1975 — was initially useful in efforts to overcome barriers to women and minorities in academia. Those barriers have largely disappeared, yet quotas and preferential hiring have persisted. After such a concerted effort, demographic disparities are less likely to point to systemic biases in academia than to underlying societal factors.

That’s especially true when it comes to disparities of sex. Women earn a majority of postbaccalaureate degrees over all STEM disciplines in the U.S. Since female undergraduates outnumber male ones by about 3 to 2, this trend is likely to continue. Further, a recent large-scale study found that previous claims about sex bias in academic science were overblown. Tenure-track women and men in STEM receive comparable grant funding, journal acceptances and recommendation letters, and women have an edge in hiring.

What quotas? What preferential hiring? I’ve been in a lot of job searches over the years, and we’re told over and over by the administration that there are questions we can’t ask, and they’re all about avoiding bias. I’d agree that there are fewer biases in academia (but not no biases) because of policies Krauss doesn’t like, and that we’re dealing with larger societal factors, but academia is part of society, if you hadn’t noticed.

A good example of those societal factors: undergraduate women outnumber men, especially at liberal arts colleges like mine. Is this a good thing? Nobody thinks so. It’s not at all because we preferentially admit women — please, high school men, do apply and come to UMM, we love you all — but because when universities stopped discriminating against women, many women saw a college education as a tool for escaping traditional roles. Liberal arts colleges also actively encourage students to explore new ideas, which is appealing if you want something more than a fast-track to a job.

Of course, to a white man the barriers are invisible, so they don’t exist.

Such personal matters are irrelevant to science and essentially invisible. In my 40 years in academia, I have worked with all sorts of colleagues and students. Many were highly eccentric, but that didn’t matter if they were good scientists. As one colleague put it: “You are teaching a chemistry or physics course. Your lectures describe concepts and present equations. ‘Suppose a magnet is moving relative to a loop of wire.’ You barely know any of your students. You give tests and grade them. You have no idea, nor care about, the ‘sexual orientation’ of any of your students. . . . What career barriers are there?”

What a blinkered ass…you might as well say, “no one is trying to rape me, therefore rape and sexual harassment are not a problem anywhere.” Which is just what a self-centered serial harasser would say.

I’m also appalled at the idea that a professor just lectures and gives tests and grades them and doesn’t need to know anything about their students. What university was this at? Krauss should have mentioned it so everyone would know to avoid it. Of course we are and should be aware of our students’ lives, to a degree. We invite deeper interactions than just talking at and grading them — I listen when students are struggling and try to help them resolve conflicts and issues.

I don’t even understand this factory-style approach to impersonal teaching.

Asking respondents if they’re “transgender,” “gender non-conforming,” “nonbinary,” “gender-fluid” or “genderqueer” is patently ridiculous. These are subjective categories, unobservable by others unless the person in question makes it a point to label himself publicly. Most scientists, like ordinary people, couldn’t even define most of these terms, let alone use them as a basis for discrimination.

You don’t need to define the terms, you just need to categorize your students and colleagues as highly eccentric.

This is peak clueless offensiveness, though. Non-heterosexual identities are patently ridiculous and mere subjective categories? They matter to the people who have them, and what also matters is professors who so callously dismiss their lived identities. You know, the ones who think people who aren’t like them are not ordinary people.

Jesus. Krauss is making me aware that we do discriminate. If we were interviewing a job candidate and they spewed out that stuff about how teaching is just about giving tests and grading them, calling gay and trans students eccentric and patently ridiculous, it’s true — there’s no way we’d hire them. We try not to employ assholes.

Also, we’d rather not hire stupid people. Krauss even quotes the goals of these agencies, but doesn’t understand them.

What’s the purpose of all this? Nature magazine paraphrases a statement from the NSF’s chief diversity and inclusion officer, Charles Barber: “Collecting these data will help the NSF and other agencies to analyse employers’ policies and procedures for addressing unintended barriers to employment, advancement and inclusion.” The magazine then quotes Mr. Barber: “This gives us an opportunity to create more opportunities and broaden participation to yield equitable outcomes for the LGBTQIA+ community and others.”

Yes. Collecting data to detect “unintended barriers to employment, advancement and inclusion.” How would you know if an organization discriminates, unintentionally or otherwise, if you have no information about the population of a class known to be subject to bias? How would you know if an organization has successfully knocked down all barriers to advancement if you don’t look? Krauss is advocating willful blindness to abuse and harassment and bias…unsurprisingly, for a guy with his history. If those meddling kids hadn’t noticed and reported his behavior, he’d still have a job!

Does that mean quotas?

No.

If so, how would one even go about determining the “correct” proportion of “queer” or “genderqueer” scientists? The percentage of the population that espouses these labels is so small that any data the NSF gathers will be statistically useless.

The correct proportion is one that roughly matches the proportion in the general population, because that would indicate that there’s probably an absence of selective bias. That wasn’t so hard, Larry.

It’s kind of astonishing to see a physicist dismissing an event as insignificant because the frequency is too low. What happened to 5-sigma, Larry? Does the Higgs boson not matter because it’s so difficult to see that you have to spend billions of dollars to detect it? Most of the stars in the sky are not exploding, so why waste our time looking for novas? The frequency of stellar class A stars is only 0.63% — can we just ignore them, then? Heck, our sun falls into a group that makes up less than 8% of all stars. Must not be important, then.

You know we can detect all kinds of numbers if we just look. Here’s a result of the US census — over a million ‘eccentric’ people live in same-sex relationships.

We also know, because people looked at the data, that over 1.6 million ‘ordinary’ people are transgender, and that the proportion is rising as social barriers fall.

A study published on Friday estimates that nearly 1.64 million people over the age of 13 in the United States identify themselves as transgender, based on an analysis of newly expanded federal health surveys.

The study estimates that about 0.5% of all U.S. adults, some 1.3 million people, and about 1.4%, or 300,000, of youth between 13- and 17-years-old identify as transgender, having a different gender identity than the sex they were assigned at birth.

I really don’t understand this frequency based argument. Can we just ignore 1.3 million people, or worse, oppress and discriminate against them? They’re statistically useless, you know. It’s just that they are people.

Wow, Krauss has become a right-wing cartoon at this point.

AI-generated garbage videos

Another way that AI is hurting us: it’s being used to generate science spam that is flooding YouTube. Kyle Hill points out that they’ve SEO’d their way to the top of video recommendations by, for instance, putting Elon Musk or Joe Rogan or Michio Kaku on the thumbnail (we’re already in dubious “science” territory), and then when you click on it you get this text-to-speech robotic voice reading what sounds like ChatGPT generated word salad accompanied by random images. It’s terrible stuff designed to trawl up bottom-of-the-barrel undiscriminating eyeballs.

I haven’t encountered much of this stuff myself, I think because I don’t click on the kind of videos he shows. It does highlight one of the many problems with the ‘algorithms’ used by social media, because the spammers will always find a way to exploit the system.