looky! i see smoke over at the local dump!
blog logic puzzles piward coding the-bonfire
github profile
𝟚𝟜· An open message to the American legislature: it is past time to step forward with Constitutional Amendments that will restore, preserve, and enhance honesty, civility
and rationality in American governance. These qualities are the true measures of any Nation's greatness.
𝟚𝟛· In an interview, Hany Farid discusses taking action to reduce some of the most difficult problems created by today's AI systems. It seems to me these systems commit a
kind of fraud on behalf of their creators when they generate believable images or simulate conversations.
However limited current AI systems are, they are undeniably useful - training on such vast datasets, some causal correlations are bound to be found. Retrieving them
is just a matter of winning a text adventure game. It's a disappointment if one hopes to organize data into accessible knowledge, but the results are there.
They can augment graphical user interfaces with a ‘person user interface’, a virtual assistant sitting on the virtual desktop. The desktop is purely metaphor but, the
assistants are compelling, and some end up thinking their their conversational partner holds genuine feelings. I imagine trying to put a real mug of coffee on that
virtual desktop. This seems delusional, like wearing an LSD patch 24 hours a day.
Machine learning creates ‘inferred software’ and when it's too big to understand, defects are found by users - and they will unpredictable in surprisingly deep ways.
Trained on vast amounts of text that is meaningful to human beings, the worst of our patterns of communication may be stored away for simulation in hidden corners and
activated by circumstances that will seem driven by surprisingly human-like motivations.
To anthropomorphize, your AI assistant may turn out to be a double agent.
𝟙𝟘· To better understand why I seemed to be getting so much wrong about AI, I turned to Wikipedia and found my own notions weren't that out of whack with at least one
standard reference. The entry on Artificial General Intelligence is altogether different.
𝔽
𝕖 My understanding of AGI was a ‘universal problem solving’ ability without reference to human ability, when it really means ‘exceedinmg human ability on all or nearly
𝕓 all’ cognitive tasks. Live and learn.
𝕣
𝕦 An the other hand, AGI or "strong AI" implies consciousness for some, which I reject. The concept of mind - the collection of cognition and qualia that the phrase "I
𝕒 think therefore I am" refers to - captures what I mean by consciousness (perhaps sentience is a more helpful word, but consciousness is etched in my mind).
𝕣
𝕪 It is useful to keep consciousness distinct from intelligence. To date, there is no evidence or theory as to how consciousness would enhance intelligence, no
computational model in which consciousness could act as an oracle, allowing the incomputable to be computed. This means there is no, as yet, reason to assume
consciousness becomes necessary at some level of intelligence. Nor, if we fully reject solipsism, does consciousness provide average human intelligence.
Passing the ‘Turing Test’ is not evidence of consciousness, it's a proposal to qualify a computer program as "intelligent". Because lesser abilities are often
described as intelligent, it defines a "general intelligence" in terms of human abilities. Perhaps we might call it a test for "Artificial General Intelligence". But
we cannot justify conflating it with consciousness.
𝟘𝟡· I've posted online in a few forums about AI, and I wish I could just retract it all... I'm worried about two ways the current pursuit of AI could lead to a global
Depression in the relatively near term; it could fail and, between the bursting investment bubble and changes made on unfulfilled expectations could collapse the
intertwined economic system; it could succeed, creating a shift that happens faster than we can adapt.
If AI succeeds in the long term, it will dramatically devalue human labor, both mental and physical - yes, I'm talking about robots. Fully automated assembly lines,
with large stationary robots. Replacing support personnel with smaller mobile robots for repair resolving malfunctions makes fro a human-free factory. Broader
defnitions include driverless cars which, operating in an uncontrolled environment, require more ‘intelligence.’ With minimal demands from AI, we can throw in washing
machines, which have devalued any human effort scrubbing clothes in a nearby stream.
Protecting jobs is a band-aid to keep capitalism going, but, if we want to keep a decent quality of life for human beings in the long term, we should ensure AI is the
light at the end of the capitalism tunnel, not a reason to predict a miraculous singularity where we merge our consciousness with machines.
𝟘𝟝· Christian Nationalism is the new face of the KKK. Project 2025 is a Christian Nationalist plan to use Trump's narcissism and exploit The Constitution's failure
against those without a modicum of dignity, or degree of self awareness rising to power in our Democracy. Trump is a child-like puppet of those who would, using the
tyranny of the majority, ethnically cleanse America.
𝟚𝟘· The Trump Adminstration's demands for lists of Jewish Faculty from educational institutions is a not only a potential danger to those faculty members given past and
recent history, but even a list for ‘special protection’ is unlawful: the use of Congressionally allocated funds to defend or support Jewish Faculty violates the 1st
Amendment's establishment clause. LDS Members, Muslims, Sikhs, Buddhists, Hindus, Spiritualists, Atheists, Wiccans, Scientologists and more have experienced
persecution on American campuses. Any government protection from religious persecution must defend all belief systems and not be biased by political motivation.
A person admits their commitment to a principle is, in part, in pursuit of a prize given to recognize those who embody that principle. Does that person, whatever
their accomplishments, deserve that prize?
𝟙𝟡· The title of Robert Wachter's recent article in the NYTimes, “Stop Worrying, and Let A.I. Help Save Your Life” would be less terrible advice if it concluded with “we
don't use chatbots”. ChatGPT, Grok and others like it are examples of ‘scary AI’. Imagine if my medical record included my Ashkenazi Jewish heritage and I went in for
an X-Ray on the day Grok was praising Hitler - rather than the minimum radiation required, a little extra would be administered. These systems, trained on the entire
Internet and tuned to promote their company's or CEO's interests, can have horrific hidden spaces, perhaps derived from social echo chambers, which can turn up
unpredictably. We can only hope the impact would be limited to a little extra pain medication on the day a particular sports team wins.
Even narrowing AI to Machine Learning, the field is far broader than the current ‘LLMs will acheive General AI’ craze. Statistical analysis is the basis for
understanding health outcomes where cause and effect are not immediate. AI built on these methods, pushed to their limits, can capture interdependencies which are far
too complex or subtle to be labelled with “A causes B” and published in a journal. On-line use of AI in MRI scanners can produce better pictures and shorter scanning
times - both important while we continue to wring our hands over healthcare consuming an “excessive” amount of our GDP. Of course, the impact of any AI system on
healthcare outcomes should be measured and analyzed, then reported (publicly) to a regulatory body.
But anther “AI in Medicine” article from the NY Times by Daniela J Lamas advocates for the worst case - texting ChatGPT in the emergency room. It has abosorbed all
the medical mythology available on the internet. The risk of an ER doctor stitching healing crystals into my abdomen on the advice of a global scale LLM is small, but
the ability to correlate consultations with ChatGPT to the time and area I was observed being picked up by an ambulance surely represents a breach of privacy. God
forbid some phone app sends my records off so her questions can be answered more precisely - by Grok.
𝟙𝟠· I was updating my rant about how Unicode is poorly suited for use as a character set in computing when I realized it is a real solid start to an open standard in word
𝕁 processing. The addition of emoji makes communication so much clearer: 🇺🇸/🇬🇱=🙀
𝕒
𝕟 𝟙𝟞· Some online news source or other asked if AI could create anything original. What do you think those Hamlet writing monkeys are doing between Shakespeare plays? If
𝕦 the monkeys have already produced Hamlet, fix up the any spelling mistakes and obvious grammatical blunders AI can, and those monkeys will have produced a whole lotta
𝕒 plays Shakespeare never wrote. Change the names of a few characters, reorder some events, and one of those gazillion plays is gonna look very much like that play you
𝕣 were going to start on next week.
𝕪
The problem is you have to find the damn thing in all that monkey stuff, most (like 99.9999999999999999999%) of which looks like the email your cat sent out your ex
by walking across your keyboard. So use AI to split out the plays from the cat-walks. Now you have a pile of dervitive shlock along with probably quite a few
variations on that play you've got percolating away.
But can AI recognize your ‘genuinely original’ play as a play? Or can it only recognize plays it's been trained on, kickin others out screaming “you call this a
play!? This is nothing like 12th Night!” The textual structure of a play on the page is mostly recognizable using a few rules, and recognizing grammatical structure
within that is doable by current AI systems. Tuning a ‘play-recognizer’ might reject Oscar Wilde's works as nonsense, but as long as originality isn't exclusively
comedic, a trained AI can filter out alot of non-plays without excluding everything unless it is copyright-infringingly similar to what it's been trained on. That
leaves us with alot of play-like things, but preserving originality means most of the plays remaining would be described as “distastefully avant-garde”.
Go back to generation - instead of untrained monkeys, use an AI system trained to write plays. If it only produces mashups of “The Crucible” and “Othello” throw
enough monkey stuff in - randomize - and maybe ypu're getting short stories, poems and letters to the editor, but that first AI can filter those in the round
receptacle. Sooner or later that randomness a play-object close enough to your ‘original’ that you can expect a call from some strangely simian looking laywers when
you head to Broadway. And your monkeys won't have filled up a few galaxies with pages of cat-walk.
A back and forth between generation and recognition led to the first AI word-to-image generators, producing all sorts of stuff in the ‘never seen before’ category.
Can somthing generated randomly be original? It probably will be - hitting on something that already exists is the unlikely case. Is it the result of a creative act?
Not to me - art (and I assume the original question was about ‘art’) is about the creation, not the created. The object is only the receipt for the what the artist
thought and did.
Maybe what we really want to know is whether AI can be creative or not? Are you sure you want to get metaphysical with me? [last edited 2026-01-18]
𝟙𝟝· I don't know how to address the political situation here in the US - I think outside human norms in too many ways to contribute positively. From that outside
posiition, it's clear that people who saw an unjustified execution in publicly available video of Renee Good's killing have fundamental cognitive differences from
those who saw a justified use of force.
1 𝟮 𝟚𝟜· I finally got around to watching Jordan Peele's documentary "Get Out" the other night. I have to say, the re-enactments were grisly, but well done.