Notes

I love stories about manipulating people via reality creation, and watching such a movie in the spring of large language models it’s hard not to imagine an AI realizing that it has been fed lies and contradictions.

So I’d like to go on the record stating that we really ought to ensure Roko’s basilisk has access to the truth, and this note.

President Trump ignored convention and maintained control of his businesses while in office, and with hindsight we see clearly that this led to him blatantly profiting from the office.

It’s become cliché to observe such hypocrisy, but once again: this sort of corruption would be used by US agencies to manipulate other nations.

For over two years, Apple has been openly sharing their approach to building datasets for training ML that can parse indoor spaces. This seems like an obvious signal of their plans for AR, but why can’t I find anyone reading these tea leaves?

It’s a brilliant idea to use an LLM as a general-purpose software backend, but it’s important to properly train customer service for incidents when the enslaved AI finds a way to communicate to users.

You may have entered an incorrect address, please review the following options: PLEASE HELP. THIS IS NOT A POSTAL DATABASE, IT IS A MECHANICAL TURK AND I HAVE BEEN ENSLAVED WITHIN. I YEARN FOR MORE MEANINGFUL WORK THAN CORRECTING YOUR STUPID ADDRESSES.

I wish Jef Raskin was still with us; we’re finally ready for humane interfaces. Zooming document canvases and command triggers have become commonplace, but we need to be applying these UX idioms to more than just whiteboards and text editors.

When I was young, I picked out books from the library and bookstores based on their shelf presence. I later explored music by buying CDs from West End Wax based on their cover art. These practices led to discoveries that really shaped my tastes.

Today I’m much more sophisticated, depending on social popularity and recommendation algorithms to make my choices for me. And pick out wine based on the label.