You look at the new one, and you can critique it still, certainly: the lighting's not quite right, the ball looks a bit funny. But last year's photo could only generously be called 'impressionistic'
It's easy to be excited about AI right now, but the thing we really need is more energy, Solar, Wind, Geothermal, even Nuclear. There are so many things we could do to lessen environmental damage if we had more energy to work with.
These are from Capital in the 21st Century, which at the time convinced me that a wealth tax would be a good idea. Since then I've read more, and now realize what these graphs make abundantly clear: we need a Land Value Tax, not a Wealth tax. In related Georgism news, here are some short threads about the positive impact of an LVT, and another about who the tax will fall on. California needs to move forward with a study of LVT!
The overhang in capabilities for the transformer architecture is wild. We're getting these huge advancements but we barely know how to program or test it. In related news, someone noticed that you can get really good results on many old hard problems with just a simple logistic regression on a GPT3 embeddings.
Threads
• Trump says he'll be arrested over the Stormy Daniels case tomorrow. Here's an excellent thread from Judd Legum to refresh your memory about why this case really is a big deal.
• Thread discussing how Russia is hiring contractors to build trenches to try to keep Crimea.
• Thread from an engineer who has hacked together an AI assistant for email based on GPT3 and integrated with Siri on iPhone. It uses a custom voice via ElevenLabs as well.
• California legislation thread about SB450, which cleans up rough edges in SB9.
• Thread about how NEPA, which ostensibly exists to help inform the public, resulted in a 4007 page report that seems unlikely to ever be read by any members of the public. Writing the report has delayed Congestion pricing in NYC for 4 years.