prescod.bsky.social

prescod.bsky.social

@prescod.bsky.social

111 followers1K following21 posts

prescod.bsky.social reposted

Philip Gourevitch
@pgourevitch.bsky.social

No they don’t. Canada and Mexico want none of this. They want the president of the United States to respect and abide by the trade treaty he negotiated with them - and not make up bogus pretexts for plunging us all into a gratuitous no win economic debacle.

prescod.bsky.social reposted

Mark Cuban
@mcuban.bsky.social

I’m going to put my rich guy hat on and say I hope that Mexico and Canada issue equal, retaliatory tariffs and stick to them for an extended period. I apologize to all the people it will cost money and the businesses it will hurt. But it’s the only way for tariffs to be seen for what they are

prescod.bsky.social
@prescod.bsky.social

A friend shared these words by Aeschylus (in Agamemnon). I had to read them a couple of times. "Be healer to this perplexity that grows now into darkness of thought while again sweet hope shining from the flames beats back the pitiless pondering of sorrow that eats my heart."

prescod.bsky.social reposted

Katharine Hayhoe
@katharinehayhoe.com

I am a climate scientist and this is correct ⬇️

👻 Dead Chachra 💀@debcha.bsky.social

Occasional reminder that there’s no, “it’s too late, its over” for anthropogenic climate change. Every molecule of CO2 that doesn’t go into the atmosphere makes a difference. Preventing 0.1 degree of warming makes a difference. Every bit of climate resilience we build together makes a difference.

prescod.bsky.social reposted

👻 Dead Chachra 💀
@debcha.bsky.social

Occasional reminder that there’s no, “it’s too late, its over” for anthropogenic climate change. Every molecule of CO2 that doesn’t go into the atmosphere makes a difference. Preventing 0.1 degree of warming makes a difference. Every bit of climate resilience we build together makes a difference.

prescod.bsky.social
@prescod.bsky.social

My belief about what o3 seems to prove: given enough compute, you can now fine tune a model to do any task that a human can do which has a clear right or wrong answer. This model can then (usually? always? sometimes?) be used to train a much cheaper model to do it at scale. 1/2

prescod.bsky.social
@prescod.bsky.social

These extremely expensive, highly reliable models will be used to train the mainstream models that we use every day to do these right/wrong tasks much more reliably than a they do today. Subjective tasks may not improve much but they are already decent at most such tasks of industrial value. 2/2

You are up to date.