Obviously it depends a *lot* on where you draw the starting line, I'm surprised it's 2009 for wind!
I think they're all still injections though, presumably what they really want to be producing is a pill of some sort.
This is a drawing of the current best practices for fully protected bike lanes. Useful when you see other designs and think "wait, this doesn't seem safe, but I'm not sure what the right way is"
Threads
• Terrific thread of the result a competition between developers to make the worst possible volume control UI.• Someone hooked Chatgpt up to a browser and asked it to find him some money, and it succeeded. (Via the California unclaimed property website)
• AGI researcher roon takes a shot at arguing that the worst-case evil LLM possibilities seem implausible. Here's a second one, using a process reminiscent of the drake equation to argue LLMs are unlikely to be an existential risk to humanity.
• Andrej Karpathy argues that AutoGPTs are the future of engineering. Someone else says it's like taking loop where the AI writes code and the human looks at it and decides what to ask for next, just take the human out. Also, here's a paper just published on the same concept.
• Story about police killing a little girl's goat. It's really sad, not sure I recommend reading it, TL;DR: a fair didn't want to let a 9 year old girl back out of selling her goat for slaughter, managed to get the police involved.
• Thread comparing Adobe's new AI image generator with Midjourney. Adobe worked hard to only use material they could license, Midjourney doesn't worry about it. The difference in the results is *quite* striking.