I design biological circuits, formerly at @ElowitzLab. I invented the core tech behind new Caltech startup Asymptote Genetic Medicines. Like coding, books too
Joined May 2023
- Tweets510
- Following338
- Followers116
- Likes4.2K
The descendants of the Judean people largely live in the same place as their ancestors. Most of them converted to Christianity, and then Islam.
This quoted post is unavailable.
Seems immoral to me.
Israeli politician Moshe Feiglin says Jewish morality requires killing the babies of your enemies.
"Because I'm Jewish, and I guess I'm a little bit weak morally, I don't want to see babies being killed," but "morally that's what should have been done," Feiglin says.
"If we would have done [to] Gaza...what the Americans did in Nagasaki…this movie NAZA…would have never came. The world would be with us just as they were with us after we won the Six-Day War."
35
AI seems to be a moatless field filled with extremely strong competitors that can rapidly meet and overcome frontier state of the arts, while bidding down the price of the service. This is not good for AI companies, which see their profits competed away. But it’s excellent for consumers - access to remarkably competent models for much cheaper than they would have been, if the field was more concentrated in one or two companies. It’s a bonanza. Those that are integrating AI heavily into their workflows will capture real value.
Gemini 4 Argon is our next era of frontier intelligence.
It shows significant improvements across benchmarks, setting a new state of the art for real-world long-horizon software engineering tasks.
43
I like these charts. Apparently h3 is like a silent “oh” sound, like you put your mouth in that position before going into the r.
64
Michael J. Flynn retweeted
Israeli bulldozers smash Palestinian olive trees ahead of annual harvest reut.rs/4jk0wbo reut.rs/4jk0wbo
Michael J. Flynn retweeted
The top advocate for Israel in Washington is now publicly arguing that Israel needs to integrate its military with ours so that it will be costly for the U.S. to extract itself from the relationship later, that doing so “would also carry a significant cost for Washington.”
The Senate is debating this proposal now. We can’t really say later that we weren’t warned.
Let’s say you made a pact with Satan and he gave you the weights for a true Shoggoth LLM, one with a superintelligent internal state with malicious intent, which simulated for itself a friendly external veneer that answered prompts politely and correctly when it knew it was being watched.
Wouldn’t putting those weights through standard fine tuning reinforcement learning immediately begin to sand down that internal computational machinery - just by entropy because those weights don’t really contribute to simply solving the problem? Until the Shoggoth no longer had an internal state?
And if LLM training would be expected to destroy Shoggoths, how could it be expected to produce them?
1
84
What Terrance Tao and the other 25 field medalists should do is fork DeepSeek and train that private instance relentlessly until it easily solves Navier Stokes, blows past OpenAI. They should then refuse to contribute or use other AIs for math.
1
78
I don’t trust Sam Altman, Dario Amodei, Elon Musk, or the US govt with regulation of AI. They will bias the AIs towards their own political ends. In particular, they will us AI to persuade future users that no genocide occurred in Gaza, despite current consensus among human rights groups. They will try to persuade users that it is antisemitic to criticize the state of Israel. This is unacceptable. Very important to resist AI regulation by those who are immoral.
63
I suspect the motivation for this is economic, not AI-safety. AI companies can invest enormous amounts of money in improving the models, but if they all invest the same, they gain no edge over their competitors, and get no return on their investment. Additionally, if they all agree to invest less, they improve slower but no one falls behind, and thus they suffer no decrease in revenue. Lower costs while keeping revenue the same is economically favorable, so you would expect AI companies to advocate for collective slowing of investment. Add to this how easy copying is, and how fast edges are erased, there is a strong economic incentive against being the fastest innovator.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
63