if you're not following the experiments happening with gemini, you are missing out. there are some deeply new & weird ideas here (mostly coming out of google creative lab) that are gonna be around long after the ghibli memes are gone 🧵 what is the future of human-AI interaction? is it virtual avatars/talking heads that feel like talking to another human? i don't think so. i think this is a fairly limited use case, maybe good for therapy, education, or customer service grisoon.github.io/INFP/

i'm building an experimental tool for exploring 25 years of my old sketchbooks, with image and text recognition powered by gemini the first step could not be automated—i pulled out my box of 33 sketchbooks, and took a photo of every single page. around 4300 pages in total.

a new study on bitcoin energy use gives us the most accurate picture yet, and it basically confirms what we already knew: bitcoin is using about as much energy as the entire internet (around 12GW or 100TWh/year).

https://twitter.com/karimhelpme/status/1668592294440890368

previous work on bitcoin energy has been “top-down”, broadly based on market-driven data, assuming that miners are spending a % of their profit on electricity. but when electricity prices go down or bitcoin price goes up, top-down approaches overestimate—by up to 50% for CBECI.

one of the biggest features @OpenAI could add right now would be contextual generation constraints for their LLMs. for example, forcing responses to match a JSON template. this is only possible when you have access to the inner sampling loop. the first company to offer this at scale is going to open up a huge market, because templated responses will lure in data scientists trying to make sense of big natural language corpora.

is there any research on, like, LLM societies? 1. ask an LLM for unique 10-word descriptors for 1000 different people 2. have it fill out each person with a backstory 3. track their locations in a virtual space and run simulated interactions between them when they meet. the sims, feat gpt-4: turing’s purgatory

one year later, i've just published to arxiv a final post-mortem of ethereum's proof-of-work era emissions arxiv.org/abs/2112.01238 the final damage: 18.1 million tons of CO2, 20% more than the nord stream pipeline leaks kylemcdonald.github.io/ethereum-emiss… i opted not to add any new mining equipment benchmarks (which would have gone in the red circle), and not to make any other changes that would have modified my previous predictions. i just ran the model for the remaining amount of time until the merge.

seeing this anti-web3 link go around, "keep the web free, say no to web3". i think this takedown misunderstands some details, so i'm going to try and clarify which critiques i think are actually valuable yesterweb.org/no-to-web3/ind… 1 "quadratic voting means people vote with money" yes, but it favors 10 people spending $1 over 1 person spending $10. it's better than existing lobbying systems that only empower people w a ton of money. quadratic voting works when each person gets a fixed set of votes.

New research & tracker for Ethereum energy + emissions. I estimate that every day Ethereum delays PoS, it emits ~20ktCO2. That’s comparable to two to three coal power plants. kylemcdonald.github.io/ethereum-emiss… This is where I say “what I found will shock you!” and give some more scary numbers and comparisons. But basically, I found confirmation that previous estimates are probably in the right range. I’ll recap in this thread, but I also wrote a short summary kcimc.medium.com/ethereum-emiss…

sometimes i think about all the failed 3d prints i made circa 2010. no one told me that they would likely outlast any other impact i could have in my life or career. i don't understand how the permanence of plastic is not the single defining factor of 3d printing aesthetics. wow, apparently a nyt photographer stopped by the @makerbot open house while i was in residence. any bets on whether the nyt or my shitty plastic trinket will decompose first? nytimes.com/slideshow/2011…

new work! "sakoku explorer" with @ycam_jp is a new way of visualizing the data that google and facebook collect about you sakoku-explorer.ycam.jp this is part of the "sakoku [walled garden] project" (advised by @suryamattu) special.ycam.jp/sakoku/ which included a workshop where we used this explorer to facilitate conversations about the data collection practices that drive surveillance capitalism.

gps metadata of 68k videos uploaded to parler some parler users uploaded videos from US military bases. and one near the googleplex.

in memory of one the most magical people i've met @jtnimoy JT was my earliest introduction to computer vision, via jmyron for @ProcessingOrg webcamxtra.sourceforge.net from the little time we spent together, i felt that JT had a unique way of approaching the world that opened up the certainty of things into something more undefined, surreal, sublime.

i’m helping to document “te lapa”, an unexplained luminous phenomena used for navigation by polynesian voyagers ⚡️⛵️ so excited to work with this team + support from @LACMA ❤️

https://twitter.com/VakaTaumako/status/1294375235643351040

the work to document and share voyaging knowledge from taumako started in 1993 when chief koloso kaveia recruited dr. marianne george. my contribution will be a small piece of something much larger, connecting beyond taumako to revivalist polynesian sailing.

at the “facial machines” seminar in aarhus, denmark for two days. i’ll be tracking some of the presentations in this thread. denmark has had a face covering ban for only a few months. there are some practical exceptions: if it's really cold, if it's halloween, or if you are wearing the extremely badass traditional dress of people of fanø island.

thread: since moving to LA i've been sailing a lot. i took @memotv yesterday and his analytical approach reminded me what a multifaceted activity it can be. it spans a number of disciplines and traditions, from polynesian wayfinding to cutting edge engineering. polynesian sailors mastered the science of navigation. one of their traditional tools is based on swell interference patterns: they use the boat rocking to determine their position and orientation in known geography, or to estimate the location of unknown islands.

reading the wikipedia page for speech synthesis i fell into a rabbit hole following the history of "brazen heads": legendary bronze objects with supernatural foresight. i noticed a number of connections to contemporary issues in AI. while golems and moving sculptures are typically mobile but speechless, brazen heads speak but are immobile. they don't follow the familiar plot of abusing their strength to run amok or destroy their creator.