Substrate

This is absolutely incredible from Lance Robertson:

I made a short speculative film as my final project for a course on Deepfakes and Synthetic Media that I took this quarter at UC San Diego, as part of my Speculative Design major.

…

Prof. Kenneth Payne (King's College London) has published a paper entitled AI Arms and Influence: Frontier models exhibit sophisticated reasoning in simulated nuclear crises and summarised it on his blog. It's game theory on steroids.

This is such great satire from McSweeney's on the state of AI company financing. We all know it's a bubble, it's just whether it's going to pop or slowly deflate...

Benjamin owns a farm. He employs 100 workers plowing his fields. His total payroll is $10 million/year. One day, he buys a mule, which provides the worker who uses it with a modest 10 percent productivity gain. Benjamin fires 99 of his workers and purchases 99 mules, expecting a 1,000 percent productivity gain. The driverless mules cause plow damage to his property in excess of $50 million. Benjamin loses another $5 million due to the loss of productivity from his one remaining employee, who no longer guides a plow but instead spends 100 percent of his time shoveling mule shit. Goldman Sachs builds an altar to Benjamin in their lobby and cuts out the heart of a junior analyst on it every Friday. They call it “Blood Sacrifice Friday.” The name isn’t catchy, but the event becomes a management favorite nonetheless.

…

This article by Justin Garson on the American Philosophical Association's blog is extremely relevant to my life at the moment. A few months ago, shortly before receiving a diagnosis of Overtraining Syndrom from a consultant, my GP prescribed me an SNRI deal with both physical and mental symptoms that I was experiencing. I'm still on it.

The ball doesn't always go in the hoop, the can doesn't always come out of the vending machine, the firework sometimes fails to go off, and the spoon sometimes sinks into the soup. That's life.

Paris-based motion design studio Parallel created a series of short animations which aim to do anything but impress, clips that highlight the frustrating day-to-day mishaps by turning them digital. The series, titled UNSATISFYING, aims to go against the trend of oddly satisfying videos that are currently pervading the internet, instead making audiences cringe with scenes that only deliver disappointment.

…

I love the idea of "places that only exist at a certain time of the day". If you've ever sat in a quiet place before it gets busy, you'll know that the whole meaning and feeling of a place can change quite suddenly depending not only on the atmospheric conditions, but who's there with you.

I like this as an encouragement to do things. Everyone thinks that the way they see the world is "obvious" but it's very much dependent on everything from your genetics through to your experience and interests.

Parenting is hard: you always feel like you're making mistakes, and looking back you always think that you could have done better. That's particularly true with my son, currently 19, who will be moving back in with us for next academic year.

I'm not sure how cool I am with the latest versions of Anthropic's Claude models (Fable 5 & Mythos 5) seemingly inventing their own internal language:

When we’re first starting to understand a new model’s behavior, the most abundant source of data we have to draw on is its behavior during reinforcement-learning training. Reviewing this evidence for signs of reward hacking (exploiting loopholes that go against the spirit of a task) or unexpected actions can inform what we should be looking out for in the model’s real-world behavior. The most notable finding was illegible reasoning in a few reinforcement-learning environments over long rollout, but little sign of deceptive or highly surprising actions, and no clear evidence of unexpected coherent goals.

…