Creativity is for startups. If you run the 8th or so most valuable company on the planet, your competitive advantage is mostly your scale. You should mostly be using that to maximize return on proven stuff (either by buying or copying proven things), not pissing away billions of dollars on stuff that a startup can iterate on for a tiny fraction of that price.

That’s an incredibly depressing way to run a world. What on earth is a billion dollars for if not doing something interesting?

Doing something interesting comes with risk. That billion dollars is likely cash from investors who want a return. They won't be happy if the interesting thing you want to do isn't going to return a profit, hence they stick with what they know best and is least risky.

If you have a billion dollars and want to do something creative, invest it in 100 startups. But if you want to spend a billion dollars through a big corporation, you should use that corporation's comparative advantage -- which is generally scale, not creativity.

Why would it be bad that all risky initiatives start small instead of eg accidentally ruining the whole healthcare, education or policing system with a single decision?

Metaverse? Llama? They’re doing both.

How well they’re doing those is up for debate, but they’re doing interesting stuff too.

Metaverse is exactly my point -- they've incinerated vast amounts of investor money with nothing to show for it. They could've and should've just plowed that money into hundreds of startups.

LLAMA is also exactly my point -- LLMs are pretty well-proven at this point. If you spend $X on machine learning researchers, you can get an LLM with Y parameters that is useful for various tasks like sentiment analysis, machine translation, etc. Facebook should be spending money on that because they have lots of cash, data centers, and ML researchers so they can do this at scale that only a few others can (Google, Microsoft/OpenAI and maybe Amazon).

That seems like a lot of gross simplifications with hindsight bias.

It's really about whether you can incrementally assess P&L.

Facebook has epic amounts of user-generated imagery and text, and incremental improvements in NLP and computer vision generate incremental revenue (and profit) because you can put up better ads and introduce new features like translation. As an investor, that is very appealing.

There's no incremental revenue from Metaverse. It's a huge money pit, with no obvious end date for when enough billions have been wasted to call it quits. It could well work out, but if I'm an investor why would I want to invest in this through Facebook rather than through a bunch of startups pursuing a diverse set of strategies.

In this case, I think there was more than a little traction for a specific kind of social site, which turned into a crater, so it's not surprising somebody would build exactly what existed before.