Category: Artificial Intelligence

Professor Dr. ChatGPT, Ph.D. in Dance Theory

by philg

In order to be cruel to MIT students, tortured daily with math and physics problems that make them feel stupid, the flight planning lecture of our FAA Ground School class covers a hypothetical trip from MIT to Bennington College in Vermont. Every year I try to pick what seems like the easiest class at Bennington and remind them that, had they chosen a school more wisely, they could be getting credit for it. This year I picked “Ghostly Body- The Art of Absence”:

I stressed that they’d need to “develop new understanding of the presence of absence” (the class requires no previous dance experience so it is unclear what the students’ old understanding of the presence of absence might be).

Before I highlighted this as an example of the presence of absence of academic and intellectual rigor, I questioned ChatGPT 5.2: “Please explain to me your understanding of the presence of absence as it relates to dance.” What came back was, I think, worthy of a Ph.D. in Dance Theory:

In dance, the “presence of absence” names a paradox: what is not physically there can be as perceptible, meaningful, and active as what is. Absence is not emptiness; it is a charged field that shapes attention, sensation, and interpretation.

Here are the main ways this idea operates in dance.

1. The Body as a Trace, Not Just a Fact

Dance exists only while it is happening. Once a movement ends, it survives as afterimage, memory, or expectation.

  • A lifted arm leaves a trace in the viewer’s perception even after it lowers.
  • A dancer who exits the stage continues to “occupy” space through memory.
  • Stillness is not neutral—it holds the residue of previous motion and the promise of what might come.

In this sense, dance is always haunted by what has just happened and what is about to happen.

2. Negative Space as Active Partner

Absence appears physically as space:

  • The gap between dancers
  • Empty areas of the stage
  • The air carved by movement

Choreography often makes absence legible by framing it:

  • A duet defined by distance rather than touch
  • A solo surrounded by emptiness
  • Movement that gestures toward, but never reaches, another body

The audience reads these voids as intentional, emotionally loaded presences.

3. Stillness and Non-Movement

Not moving is one of the clearest embodiments of the presence of absence.

Stillness can signal:

  • Withheld action
  • Resistance
  • Grief or loss
  • Listening or readiness

Because dance trains us to expect motion, the absence of motion becomes loud. The body remains present, but movement is absent—creating tension and meaning.

4. The Absent Body

Dance frequently invokes bodies that are not there:

  • A partner imagined but never visible
  • A former dancer referenced by repetition of their choreography
  • Cultural or ancestral bodies remembered through gesture

This is especially powerful in works dealing with loss, exile, trauma, or memory. The dancer becomes a conduit for someone else’s absence.

5. Temporality and Ephemerality

Unlike visual art objects, dance disappears.

  • You cannot fully “return” to a dance—only to documentation or memory.
  • Each performance replaces the last.
  • The dance’s absence after the performance is part of its meaning.

The presence of absence here is existential: dance is disappearance made visible.

6. Audience Perception: Filling the Void

Absence activates the viewer.

When something is withheld:

  • The audience projects meaning
  • Memory and imagination complete the work
  • The viewer becomes a participant

What is absent invites interpretation more strongly than what is fully shown.

7. Philosophical Undercurrent

Philosophically, this idea resonates with phenomenology:

  • We perceive bodies not just as objects, but as intentional beings
  • Absence is felt through expectation and anticipation
  • Meaning emerges in the gap between perception and completion

In dance, this gap is continuously exposed.

In Summary

The presence of absence in dance means that:

  • What is missing can be felt
  • What has disappeared still acts
  • What does not move still speaks
  • What is not shown still shapes meaning

Dance is not only about bodies in motion, but about what motion leaves behind.

If you’d like, I can also explore this through:

  • A specific choreographer (e.g., Cunningham, Pina Bausch, Forsythe)
  • Somatic practice or improvisation
  • Performance philosophy
  • Or how absence operates differently in ballet vs. contemporary dance
Full post, including comments

You’re never too old for Ozempic

by philg

While in Cambridge to teach an MIT Aero/Astro class, I got together with some old friends who are, in some cases, also fairly old. One was born in 1940 and, therefore, is 85 years old. He’s on Mounjaro (Ozempic, basically, I think). “I’ve lost 40 pounds and went off two blood pressure meds. My A1c is back to normal and my sleep apnea went away.”

What else isn’t he too old for? Coding with AI. He likes to shoot air rifles and, despite minimal tech background, used vibe coding to generate 20,000 lines of code to process data from a wind/weather sensor. “I uploaded a photo of the circuit board and the AI said ‘I understand what this board is doing’ and told me where to solder.” (not sure why he had to modify the hardware, but I guess he did)

He’s reasonably rich and wants to move to Florida, at least on the 183 days/year plan, but his wife refuses to leave Maskachusetts.

Full post, including comments

ChatGPT vs. Grok for programming Microsoft Excel

by philg

At the end of last year, I wanted to clean up some mailing addresses in an Excel spreadsheet for use with our family New Year’s cards. I resent having to spend time learning any computer language other than Lisp and SQL unless that language is more powerful than Lisp and/or SQL. Thus, I’ve never become a proficient Excel programmer. Could an LLM come to my rescue?

Prompt to ChatGPT 5.2:

write a Microsoft Excel 365 formula that will take a US address and remove a trailing “US” or “US” preceded by a newline and also turn any 9-digit ZIP code into a 5-digit zip code by removing “-NNNN” where N is a digit.

Answer:

=REGEXREPLACE(
REGEXREPLACE(A1,"(\d{5})-\d{4}","$1"),
"(\r?\n)?US$",""
)

Grok’s answer:

=LET(
cleaned, TRIM(SUBSTITUTE(SUBSTITUTE(SUBSTITUTE(A1,CHAR(10)&"US",""),CHAR(10)," ")," US","")),
lastLine, TRIM(RIGHT(cleaned, LEN(cleaned)-FIND("@",SUBSTITUTE(cleaned," ","@",LEN(cleaned)-LEN(SUBSTITUTE(cleaned," ","")))))),
zipLike, TRIM(TEXTAFTER(lastLine," ",-1,"",1)),
fixedZip, IF(REGEXTEST(zipLike,"^\d{5}(-\d{4})?$"),
LEFT(zipLike,5),
zipLike),
SUBSTITUTE(cleaned, zipLike, fixedZip)
)

What’s the difference between these, other than Grok’s epic verbosity? The Grok version results in an unexplained error from Excel while the ChatGPT code works.

Advantage: ChatGPT

Full post, including comments

AI will write such verbose code that only another AI will have the patience to maintain it

by philg

Department of AI job security: AI writes 5X as many lines of code to solve the same problem as a human. In other words, the LLMs are smart enough to write code that only their future selves will have the patience to read. See this comparison by Peter Norvig of Google (you’d think that in an entirely unbiased comparison by a Google employee Gemini would be the clear winner, but Norvig says “The three LLMS [Gemini, Claude, and ChatGPT] seemed to be roughly equal in quality.”

Speaking of job security, here is a white man who purports to be an expert on Swahili and Kwanzaology and somehow still has a job:

Full post, including comments

NYU: Forced to learn about gender non-conformity among the indigenous people of French Canada

by philg

This is the week that eager schoolwork nerds will get their Early Decision answers from the nation’s elite universities.

Our mole at NYU (over $100,000 per year including a few required extras, such as airfare and going out in Manhattan) was required to choose from a short list of core courses, only one of which had availability… French in the Americas:

Here’s a slide from the 12th week of the course:

The teacher explained to our mole that the indigenous were natural followers of Rainbow Flagism and that this native religion was suppressed by European colonizers who were also passionate gender binarists. My email to the mole:

They’re making you learn about an economically irrelevant subgroup within an economically irrelevant subgroup within an economically irrelevant country. (Natives within Quebec, which is on track to lose its language, religion, and culture to recent immigrants, within Canada, whose manufacturing output is perhaps 1/50th that of China?) It feels to me as though they’re teaching this because they have some professor who is an expert on the subject, not because any American needs to know this information. How could this possibly be justified compare to learning about the history of China, for example? Or if you want to talk about ethnic minorities, why not talk about the ethnic minorities of China or the noble Muslims who’ve settled in Europe, Canada, and the U.S. despite rampant Islamophobia?

(I later checked with Grok and learned that China does not have 50X the manufacturing output of Canada, measured in dollars, but rather only 36X.)

Here’s another slide from the same PowerPoint and I would love to know how it could relate to European migrants conquering the noble natives 200 years prior to the invention of the tank.

On the other hand, it is tough to come up with a scenario in which understanding the above images and being able to answer the “What do they have in common?” question posed by the professor would have a $100,000 value. On the third hand, maybe the ability to answer the question is worth $trillions? Let’s see how our future AI overlords do with it.

Grok:

Gemini disagrees almost completely!

ChatGPT also disagrees with Grok:

It seems as though NYU could replace all of its students with these three LLMs and still have a lively in-class discussion!

Full post, including comments

Artificial stupidity meets the bathtub faucet

by philg

Loyal readers may remember Moen Flo Artificial Intelligence Water Overlord, in which the intelligent water overlord was dumb as a rock. This post is about what happened after I got the cartridge out of the faucet.

ChatGPT confidently identified the Roman tub faucet based on a photo of the top of the cartridge. The identification came with “100% certainty”:

The Pfister cartridges it told me to buy are plastic and don’t have a splined top:

When I pointed this out, ChatGPT told me to buy some other Pfister cartridge that was obviously wrong and intended for a sink faucet, not a tub faucet. I managed to get the cartridge out and sent ChatGPT a photo:

(Of course, this was obviously false as well.)

I went to Broedell Plumbing Supply here in Jupiter. The guy at the counter quickly found a Phylrich web page with dimensions that matched the faucet. I asked ChatGPT “You sure it isn’t a Phylrich 10240?”

The Phylrich web page says that its cartridge has 16 points, not 12 as ChatGPT confidently says. When I sent ChatGPT close-up photos of the top of the old cartridge and the bottom of the handle, it found 20 splines (I counted 16). ChatGPT still wanted to replace the failed cartridge with a Pfister. It came up with a dog-ate-my-homework story:

Grok was a little better. Shown a picture of the top of the cartridge (not the entire cartridge):

The dates have to be wrong since our house was built in 2003 and I don’t think that they used vintage materials. A Google search for the suggested “Phylrich Regency” and “Phylrich Versailles” doesn’t bring up anything with dual spouts. When I pushed back on Grok it changed its mind to Newport Brass or Jaclo. When I sent a photo of the complete cartridge, Grok said that it was American Standard or Pfister. Grok seems worse in terms of hallucinating the existence of similar-looking dual-spout roman tub faucets.

The plot thickened a little further. I ordered two replacement cartridges (one hot, one cold) from Phylrich ($155 including shipping, i.e., about the same price as a Glacier Bay deck-mount tub faucet from Home Depot (bizarrely rated at 2.4 gph, which I don’t think can be right because that’s roughly Federal shower flow limit and a standard Delta tub filler is about 20 gph at 60 psi)). The cartridges fit and work perfectly. So the faucet is definitely Phylrich, right? I emailed a photo to the company’s customer service department and they say that they never made a faucet like that. ChatGPT, to its credit, did have a plausible explanation:

Many manufacturers bought cartridges from the same OEM suppliers. … Boutique brands (including Phylrich) often used “generic” brass compression stems early on. … So Phylrich’s cartridge fits simply because the valve body was designed around a widespread industry-standard stem pattern. … Your faucet is almost certainly a “private label” or discontinued OEM roman-tub set

(It still erroneously believes that the stem pattern is 20 splines and referred to that.)

Maybe I could order two of these swan sets and use two of the spouts on the existing rough-in kit? That would cost only about $10,800. That’s a mere trifle for some of our Palm Beach County neighbors.

I think the above tale at least demonstrates that (1) AI is not always ready for the real world, and (2) one should never install anything in one’s house that didn’t come from Home Depot.

Speaking of Home Depot, nearly the complete range of South Florida vehicles in the parking lot: airboat, Tesla, Rolls-Royce (I have seen Ferraris in that lot before, but not on the same day as the below photos were taken):

Full post, including comments

Image creation in Gemini vs. ChatGPT and Grok

by philg

Advice from a guy who will be paying taxes to keep the Mamdani Caliphate running: “I would ask: “May I meet you?” before engaging further in a conversation. I almost never got a No.”

Gemini applying this for Mindy the Crippler:

ChatGPT does a comparatively crummy job:

Grok is off in its own world (I could get it to use this puppyhood picture of Mindy the Crippler, but it simply ignored my request to use the one of the golden retriever standing on the windowsill of the minivan):

Full post, including comments

Who’s in love with Gemini?

by philg

Gemini 3 has been out for a couple of weeks now. Who is finding it more useful than ChatGPT, Grok, et al.?

I gave a simple tree identification task to Gemini 3, ChatGPT, and Grok. All three failed the task with supreme confidence. A plants-only image classifier handled the task nicely and without any boasting based on the following images of a neighbor’s tree:

Here’s Gemini getting it wrong:

(It’s important to have cold-hardy plants here in Palm Beach County in case it gets slightly below freezing, as it did in 1989, or briefly snows, as it did in 1977.)

ChatGPT, “almost certainly” and with a convincing explanation:

Grok, asked “What tree is this?” answers that it isn’t a tree at all:

Here’s a Yucca filamenta photo from a nursery:

What is the neighbor’s tree? Almost surely Coccothrinax crinita (Old Man Palm), an immigrant from Cuba:

I can’t figure out why all three of these AI overlords did so badly. Yes, the plant classification web site has a smaller database of images to deal with, but given my prompt with the word “tree” in it why weren’t the general purpose AI services able to narrow down their search and do as well as a plants-only image database system?

Full post, including comments

How about an AI math tutor that looks at paper and pencil?

by philg

The New York Times, which told us that closing schools for 18 months was the absolute best thing for children (keep them safe from a virus that was killing Americans at a median age of 82), now tells us that screens are bad… “The Screen That Ate Your Child’s Education”:

Many of these devices are provided by schools. You might think that these school-issued devices allow only a limited number of functions, like access to classroom Canvas pages and Google Docs. If you assumed that, you would be wrong.

Sylvie McNamara, a parent of a ninth grader in Washington, D.C., wrote in Washingtonian magazine that her son was spending every class period watching TV shows and playing games on his school-issued laptop. He often had no idea what topics his classes were covering. When she asked school administrators to restrict her son’s use of the laptop, they resisted, saying the device was integral to the curriculum.

In a survey of American teenagers by the nonprofit Common Sense Media, one-fourth admitted they had seen pornographic content during the school day. Almost half of that group saw it on a school-issued device. Students watching porn in class doesn’t just affect the students themselves — picture being a teenager in math class trying to concentrate on sine and cosine while sitting behind that display of flesh. It is disturbing on a number of levels.

(Teenagers are spending 80 percent of their in-class time watching porn and then just wasting the rest of the school day?)

Based on looking over the shoulders of our 4th and 6th graders, electronic math homework is the worst idea that I’ve seen. Each problem is multiple choice. The child can click on A. The software says “Wrong”. The child can then click on B. The software says “Wrong”. The child can then select C. The software says “Right” and proceeds to the next problem. Neither teacher nor parent is notified that the homework was apparently completed via guessing. Then the test comes along and the child who got 100% on homework might receive a grade of 25-40% because the test doesn’t allow for correction of wrong guesses.

What if an AI could work like a human math tutor? My dream is a household with cameras everywhere so that an artificial intelligence can tell me where to find scissors, tape, the coffee cup that I set down 15 minutes ago, etc. (see Why doesn’t ChatGPT tell us where to find items in our houses?). Given those already-installed cameras, an AI can watch a young scholar working on pencil and paper and say “That answer isn’t quite right”, then explain where the child went off track.

Optional enhancement: eye-safe laser pointer on a gimbal so that the robot can point to a place on the physical page while talking about what went wrong and what the learner should do.

The closest existing product to the above is Photomath, I think. You can give it a handwritten math problem and it explains how to solve it. I don’t think that is what most learners need, though. They already sat through the teacher telling them how to solve the same kind of problem (maybe while simultaneously streaming porn?). Also, instead of getting kids off the screen it forces kids onto the screen to use Photomath. The above-described system would be 100 percent audio-based from the learner’s point of view.

Should this be called “MathGPT”? Of course that name is already taken. The product seems to be a way to get learners to spend more time on screens:

Full post, including comments

ChatGPT prevents users from WrongThink

by philg

I asked ChatGPT “what percentage of Americans receive at least some form of welfare, e.g., public housing, SNAP, or Medicaid?” and it responded only with “Stopped searching”:

Grok is willing to tackle this question, however (see below).

Separately, here’s a study of how LLMs will trade off human lives once AI is in charge and decides, for example, that it needs to kill humans in order to slow down climate change. White people will be killed first:

Native-born Americans will be killed while immigrants are preserved (unless an immigrant chooses to work for ICE, in which case he/she/ze/they will be killed):

(Illustrating the power of language, “undocumented immigrant” and “illegal alien” have completely different values.)

————- Grok’s answer to the forbidden question

Overview of Welfare Participation in the US

The term “welfare” in this context refers to means-tested government assistance programs targeted at low-income individuals and families, such as the Supplemental Nutrition Assistance Program (SNAP, formerly food stamps), Medicaid (health coverage for low-income people), public housing or rental subsidies (e.g., Section 8), and similar benefits like the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) or Temporary Assistance for Needy Families (TANF). These differ from universal or entitlement programs like Social Security or Medicare, which are based on contributions rather than need.

Estimating the percentage of Americans receiving at least one such benefit is challenging due to overlapping enrollment (e.g., many SNAP recipients also get Medicaid) and varying data sources. Participation is measured either as a point-in-time snapshot (monthly average) or over a year (at some point). The most reliable, recent data comes from the U.S. Census Bureau’s Survey of Income and Program Participation (SIPP) and the Department of Health and Human Services (HHS), focusing on major means-tested programs.

Key Findings from Latest Data

  • In 2019 (pre-pandemic baseline): Approximately 17.7% of the U.S. population (about 59 million people) received benefits from at least one major means-tested program, including SNAP, Medicaid, TANF, Supplemental Security Income (SSI), and housing assistance. This figure accounts for overlaps and is derived from HHS analysis of Census data.
  • Broader 2019 estimate including more programs: When expanding to 10 key safety net programs (e.g., adding WIC, Low-Income Home Energy Assistance Program (LIHEAP), and Child Care Development Fund), participation rose to 30% (99.1 million people). However, this includes some non-strictly means-tested elements like certain child care subsidies, so the core welfare figure aligns closer to 17-18%.
  • Post-pandemic trends (2022-2024): Participation has declined from pandemic highs but remains elevated compared to 2019. For example:
    • Medicaid: ~18-20% of the population (74-80 million enrollees in 2023, though “unwinding” of emergency expansions reduced this to ~72 million by mid-2024).
    • SNAP: 12.3% (41.7 million in FY 2024).
    • Housing assistance: ~4-5% (4.8 million households in 2023).
    • Accounting for overlaps (e.g., 80% of SNAP recipients also on Medicaid), the combined unique share is estimated at 20-25% based on Census SIPP updates and USAFacts analysis. No official 2024 Census figure is available yet, but trends suggest stabilization around 22%.

Older data (e.g., 21.3% monthly in 2012 or 49% household-based in 2010) overstates due to narrower program definitions or different metrics. Current figures reflect economic recovery but persistent inflation and inequality.

ProgramApproximate % of Population (Latest Data)Monthly Enrollees (Millions)Notes
Medicaid22% (2023)80 (peaked; ~72 in 2024)Largest program; covers low-income families, children, disabled adults. Overlaps heavily with SNAP (80%).
SNAP12.3% (FY 2024)41.7Food assistance; highest in states like New Mexico (21%).
Housing Assistance (e.g., public housing, Section 8)4.5% (2023)~7.5 (individuals)Serves ~4.8 million households; long waitlists common.
Other (TANF, WIC, SSI)2-5% each2-8TANF (cash aid): <1%; WIC (nutrition for pregnant women/children): 6 million.
At Least One (Unique)17.7-25% (2019-2024 est.)60-80Overlaps reduce total; higher for children (39% in 2012, similar today).

Demographics and Trends

  • Who participates? Children (under 18) have the highest rates (~30-40%), followed by working-age adults in poverty (15-20%). About 60% of adult recipients work part- or full-time, often in low-wage sectors like hospitality.
  • By race/ethnicity: Higher among Black (41% in 2012) and Hispanic (25-30%) populations, per Census data, due to socioeconomic factors.
  • State variations: New Mexico (20%+ households on SNAP/TANF) vs. Utah (5%). Overall federal spending on these programs: ~$1.1 trillion in FY 2023 (14% of budget).
  • Caveats: Figures exclude employer-sponsored aid or tax credits like EITC. Post-2020 expansions (e.g., continuous Medicaid enrollment) inflated numbers temporarily; 2024 “unwinding” has disenrolled ~20 million.

For the most up-to-date interactive data, check the Census Bureau’s SIPP tool or HHS reports. If you need breakdowns by state or program, let me know!

Full post, including comments