1/ Here’s the trouble with ONE centralized #AI agent offered by ONE company: People get to choose who they hang out with. We figure out the core values and beliefs of other people around us quite quickly as they speak openly and reveal their POV.

2/ #AI agents don’t reveal how they have been trained. What sort of POV and core values and beliefs they have developed. What biases will influence how they compose the words while pretending to think. And if there’s only one, we don’t get a change to leave & hang out w/ others.

3/ If there’s only one #AI agent, and everybody around us uses it and becomes faster, more detailed, more eloquent, etc., we quickly suffer such a competitive disavantage that we have no choice but to embrace whichever bias the AI agent has received during its training.

@elonmusk @MKBHD Huge respect, @MKBHD, but last time I checked, rule #1 was “Do not go bankrupt”.

Why do you want @elonmusk to turn Twitter into a Facebook? Isn’t enough what Meta does to its users behind the disguise of free?

Sure, comm could be handled slightly different. But different issue.

10/ reboot #4, @BoschGlobal. I’ll keep a public record so people know how many times your new dishwashers fail at simply turn on.

Still hoping to have a chat with the PM that shipped this and the VP that approved it.

@elonmusk What I love about the #AI community is how quietly and gently criticism is delivered.

@samcharrington Also, cheap.

How soft, peaceful criticism is delivered within the #AI community: https://t.co/RMSemENoVU

This was way easier.

It feels like #ControlNet gives you the opportunity to “reskin” reality.

By comparison, existing img2img techniques are more about taking inspiration from reality (= the starting image). There’s significantly less fidelity in results. https://t.co/6iri4trZKZ

If it happens here in London, the first thing that goes is the London Eye. It’s always the London Eye.

And the Thames, of course.

#ControlNet #stableDiffusion https://t.co/Q3U9hDp5kY