BREAKING: New study in PNASNews combines #remotesensing data, #machinelearning and #humanrights expertise to shed light on forced labor at sea bit.ly/2K2VyBT emLab UC Santa Barbara Liberty Shared SFG UC Santa Barbara Ren Cabral @TimHochberg David Kroodsma traceybit Oliver Zahn Walmart.org

Fishing is a major application of slavery. Satellite data and machine learning can identify high-risk vessels with high confidence.Global Fishing Watch, emLab UC Santa Barbara, Liberty Shared, SFG UC Santa Barbara, Ren Cabral, @TimHochberg, David Kroodsma, traceybit pnas.org/content/early/…

Voting behavior in the 2020 election is strongly correlated with per capita Covid-19 case prevalence, with red states leading total case counts (even California, which has the largest total case count is a distant 36th in terms of spread).

Thank you, black women and men of Georgia, for saving all of us.

Great article by Ezra Klein highlighting the importance of transitioning from animals to more efficient means of converting plants to the foods we love most. nytimes.com/2021/12/16/opi…

Republicans are getting very upset people keep retweeting this ad so I figured I would post for any of you who haven’t seen it yet and then I am going to retweet myself tonight.

Hillary haters, this is on you

"When a self-driving car...kills someone...[how do we stop] the media from [creating] a panic and fear of the new? (Self-driving cars will crash sometimes, but very much less often than human-driven cars that kill over 1 million people every year.)" warpnews.org/optimists-edge…

So much winning. 2028 can't come fast enough. linkedin.com/posts/jason-mi…

This interesting idea is being popularized especially by people in the crypto + AI intersection (e.g., Worldcoin’s mission, posts from Balaji Srinivasan, Vitalik Buterin and some xAI-adjacent thinkers) who argue that AI and crypto are symbiotic. AI needs crypto’s trust layer

The scientists who spend decades developing intuition about protein folding, materials behavior, or drug interactions - they should be partners in the AI revolution, not inputs to it. Right now there's no infrastructure for that partnership. We can build it.

Interesting tension in AI right now: the labs need expert knowledge to get better at hard problems, but experts have no real incentive to contribute. The current model is basically "give us your expertise for free and maybe we'll compete with you later." Doesn't seem stable.

The data speaks for itself

Data

My writing today in the The San Francisco Standard: Too many Silicon Valley leaders have divested themselves of the responsibility to speak out against the Trump administration. We can’t shrink away and hope the crisis fades. It’s time for all of us in tech to do and say more.

I’ve been encouraged in recent days that more Silicon Valley leaders (Paul Graham & Vinod Khosla, just to name a few) have spoken out to condemn the Minneapolis killings and the policy choices that led to them. It’s a good start to something America needs much more of right now.

We measured what happens when AI compresses your conversation to fit the context window: At 5x compression, 75% of the specific facts you told it are gone. At 10x, all of them. The model doesn't flag any of it. It just keeps working with full confidence on wrong assumptions.