This seems like the truth on the manufacturing debate. Productivity only went up dramatically before because it's a per capital measure.

We should be more obsessed with total output as a nation rather than pure productivity/person, especially as labor participation rate is low.

Whenever I see hardcore environmentalists advocate we shut down all growth, I think they don't realize they're... already winning.

So long as we survive the next 100 years without disastrous, irreversible catastrophe (possible), humans using too many resources seems unlikely.

People, in general, are far too eager to find disagreement with one another.

I think about this often. We are so damned hard-wired to debate, divide, and politic.

My toxic trait (in Silicon Valley culture) is that I really appreciate suits and fashion in general.

But if you wear this stuff to work here, you're kinda looked down upon. It does seem more people are open-minded today, but the 'rebel hoodie' is now the mainstream.

On one end, whenever people talk about the disgust of protesters taking down statues, I kinda get it. Feels over the top.

On the other end, taking down statues is precisely what the first Americans did. They just happened to also melt them down into bullets afterwards.

Engineers in Silicon Valley — you already know titles like “Staff” are rare.

What we are seeing happen very rapidly is the floor for good development raising dramatically due to AI. Big supply increase.

Expect a major salary squeeze over ~2 years.

hindustantimes.com/business/googl…

The “one flexible connect account” was an absolute dream when I worked there. I was on some of the early teams designing it, but left before it beta’d.

So stoked for what that becomes over the next 5 years. It’s a fun roadmap. Economic infrastructure ftw.

Call me old fashioned, but mass infiltration of US infrastructure by China feels awfully close to an “act of war” to me.

So much so that a kinetic response seems inevitable, as these cyber threats will keep ratcheting up and are already unacceptable.

theguardian.com/world/2024/apr…

This entire thread very much highlights and covers what we're building at Embra.

Though I would purely say, with a bunch of human input, these aren't really 'autonomous' agents.

Semi-autonomy for the win.

About 1 in every 8 Uber rides I get, it seems the driver thinks he’s training for formula 1.

Llama3 looks very promising. Although the benchmarks are always interesting, you really have to use the models in depth to fully understand their differences.

For example, you can compare Opus and GPT-4 — numerically they’re close.

But they’re nothing alike in practice.

I agree that it's useful for technologists to lead with optimism. What Humane was able to package into hardware is truly impressive.

I think the heat + software problems should be surmountable within a year or so. People forget how shitty the v1 Apple Watch was.

when buying a suit, hold the sales associate like this, cock back your fist, and then look for a collar gap while saying 'the menswear guy told me to do this'

Of all the various new ways AI is changing the game, my biggest concerns are with voice cloning.

The commercial applications are modest when compared to more generic text-to-voice. Businesses don't need to emulate specific people.

The abuse/damage potential is absurdly high.…

This picture has been sticking in my brain. This is a good exemplar of why LLMs as computing platforms don't make sense to me. They learn the input distribution biases and all, and not necessarily any underlying structure or meaning. It's the hallucination problem in another form