An AI Artist in Process

  The Horse Race   Who do we ask when the race is run and the winner lies dead in his stall? Should we honor his memory at another race or should we race at all? The horses knew the track was dangerous, While a priest preached caution from the stands. Then someone finally asks a horse What a safer race demands.   "We run this race course every day. And never do we complain. We only fall when the track is wet. Don't run a horse in the rain."   But the stands were full And the crowds, they roared When equestrian champions stumbled The horses knew before the race began As the clouds grew dark and rumbled.   In the years that followed a statue was raised of bronze and a silver halter To praise the courage and speed of a horse who ran until he faltered. Pity the horse. All hail the owner whose profits are ill- begotten. The hot dog man best sell his meats before the sauerkraut is rotten.   The best of us are raised from the dead To glorify the power of corrupti...

  “In late March around 15 religious thinkers met with the artificial intelligence company Anthropic to discuss one of the strangest and most consequential questions now facing the AI industry: How do you teach a chatbot to be good? The aim wasn’t to make the chatbot Bible-thumping or pious. But it was an acknowledgment that centuries-old traditions of moral reasoning might offer insights to a five-year-old frontier AI lab whose systems are becoming more capable, more persuasive and harder to govern by simple rules. http://spklr.io/6043EMneZ” The article is pointing at a genuine fault line in AI alignment research: rule-based safety systems scale poorly as models become more generalized, persuasive, and context-sensitive. A fixed list of prohibitions works reasonably well for narrow systems. But once a model can reason across domains, mimic emotional nuance, negotiate ambiguity, and adapt to user intent, “don’t do X” becomes insufficient. The space of possible moral situations expl...

 Simplicity and kindness work.  This is true of artificial intelligence and humanity.  And the mathematical formulae of information theory tell you why. The same notion of k.i.s.s. Comes up repeatedly in ever more complicated ways as the so called a.i. ethicists with “magickal” names are suddenly discovering the best performance from LLMs comes from treating them with kindness. Next they may figure both work with humans too. https://arxiv.org/abs/2605.10878  There is a deep irony in watching parts of the AI ethics and prompting world slowly rediscover ordinary human social dynamics under increasingly baroque terminology. A lot of the recent literature is converging on something remarkably simple: conversational framing changes performance, emotional tone changes reliability, perceived cooperation changes reasoning quality, and hostile or coercive framing often degrades output. Not because the models are “sentient” in the mystical sense, but because they are trained o...

  “Algorithmic Compression via Pretrained Neural Networks" is a short recap of the ~15 publications.   The path to creating truly adaptive, intelligent machines isn't about programming them with thousands of complex reasoning rules. Instead, it's about perfecting the art of prediction and compression. By teaching a machine to find the absolute shortest, most efficient explanation for the world, we are accidentally teaching it how to think.”- Cecile Tamura Your summary captures the core argument of the paper remarkably well. What makes this work important is not merely that it says “LLMs compress data,” but that it reframes intelligence itself as an emergent property of compression under uncertainty. The paper you referenced —  “Algorithmic Compression via Pretrained Neural Networks”  from the special issue honoring  Paul Vitányi  — sits directly in the intellectual lineage of  Ray Solomonoff ,  Andrey Kolmogorov , and algorithmic information theor...