Way to talk in generalities Joe. Not saying there aren't a lot of fucked up things about pharma. That's why we created https://t.co/39PNLwZeYs. But to ignore that the same industry has saved who knows how many lives is bullshit and you know it.
It's also disrespectful to all the doctors, researchers and medical professionals that dedicate their lives to saving lives, like Dr Hotez and the 800k plus doctors in the country that believe vaccines save lives
You arent trying to find the ground truth on vaccines. If you were you would bring on someone who focused on debating the issue. Like @thereal_truther .
Trying to bully Dr Hotez is ridiculous. You have producers that will prepare you and you get to control the conversation.
@RobertKennedyJr also has a staff ready to prepare him, and these topics are what he talks about in every speech.
You both do this on a daily basis Dr Hotez works every day to try to find ways to help people.
Joe , you and @elonmusk's @twitter are the mainstream online media and your platforms have become everything supposedly wrong with MSM. You are driven by self interest. Just like the MSM always has been accused of.
And you both have earned that right. You busted your asses to be great at what you do and earned all you have accomplished.
But don't lie to yourselves and all of us and tell us you are different. You aren't.
UNSEALED INDICTMENT: The indictment starts with the stunning revelation Trump showed classified information to people not entitled to view it (and he was not entitled to possess it)
@rkosai Interesting, I think it moves me in the other direction? Seems that this could be extensible to more realistic simulations and points the way to an embodied agent with online learning. Short step from there to the Borg…
Was curious about the relative popularity of 1984 vs Brave New World over time. ChatGPT helpfully suggested Google Ngram viewer (which I didn’t know about). It also helped try to explain the spike in Orwell around 1977: Pink Floyd’s “Animals” was released, based on Animal Farm
@EMostaque Wouldn't Chinchilla optimality be ~175B parameters per Table 3 in the paper? Or have we moved on to better scaling laws now? Curious how you estimated fully optimal, doesn't seem to be based on https://t.co/nFvPRydd2K which would put you in the 6-7B range
@benthompson@benedictevans On review, you’re right, you said a larger and more diverse plugin ecosystem could be a small differentiator that can help with defensibility. I don’t think they can even serve that role as they’re not proprietary though. Either way, it was a very minor point in your piece.
Both @benthompson and @benedictevans referenced ChatGPT plugins as a moat this week. This misses the fundamental difference with how plugins work. ChatGPT just reads the API spec and figures out how to call it. No integrations to build, and other LLMs can use them too = no moat
@benedictevans@benthompson You’re right, I extrapolated (perhaps unfairly) from “textbook platform play.” I think the point stands though, as plugins aren’t “textbook” in a very key way.
@AnthropicAI “We believe that companies that train the best 2025/26 models will be too far ahead for anyone to catch up in subsequent cycles.”
Clearly self-serving, but could be true especially if chip shortages continue.
https://t.co/bebV1TfCQi