After market close on 10/26/2022, Meta guided an increase in annual capex of ~$4B (from 32-33 for 2022 to 34-39 for 2023), “with our investment in AI driving all of that growth”. NVDA shot up 4% afterhours on this news. (Before you get too alarmed, I read somewhere that most of that is going towards running ML on videos, which is apparently very computationally expensive, in order to improve recommendations, in order to compete with TikTok. But one could imagine all that hardware being repurposed for something else down the line. Plus, maybe it’s not a great idea (for us humans, collectively) to train even narrow AIs to manipulate humans?)
How much do you expect Meta to make progress on cutting edge systems towards AGI vs. focusing on product-improving models like recommendation systems that don’t necessarily advance the danger of agentic, generally intelligent AI?
My impression earlier this year was that several important people had left FAIR, and then FAIR and all other AI research groups were subsumed into product teams. See https://ai.facebook.com/blog/building-with-ai-across-all-of-meta/. I thought this would mean deprioritizing fundamental research breakthroughs and focusing instead on less cutting edge improvements to their advertising or recommendation or content moderation systems.
But Meta AI has made plenty of important research contributions since then: Diplomacy, their video generator, open sourcing OPT and their scientific knowledge bot. Their rate of research progress doesn’t seem to be slowing, and might even be increasing. How do you expect Meta to prioritize fundamental research vs. product going forwards?
Zuckerberg has made a huge bet on VR/”The Metaverse”, to the tune of multiple times the cost of the Apollo Program. The business world doesn’t seem to like this bet, people are not bullish on VR but are very bullish on AI. So the pressure is on Mark to pivot to AI, but also to pivot to anything that is productizable.
Spot check: the largest amount I’ve seen stated for the Metaverse cost is $36 billion, and the Apollo Program was around $25 billion. Taking into account inflation makes the Apollo Program around 5 times more expensive than the Metaverse. Still, I had no idea that the Metaverse was even on a similar order of magnitude!
The $36b number appeared to be extremely bogus when I looked into it the other day after seeing it on Twitter. I couldn’t believe it was that large—even FB doesn’t have that much money to burn each year on just one thing like Metaverse—and figured it had to be something like ‘all Metaverse expenditures to date’ or something else.
It was given without a source in the tweet I was looking at & here. So where does this ‘$36b’ come from? It appears to first actually be FB’s total‘capex’ reported in some earnings call or filing, which means that it’s covering all the ‘capital expenditures’ which FB makes buying assets by building datacenters or underseas fiberoptics cables; $36b seems like a pretty reasonable number for such a total for one of the largest tech companies in the world which is doing things like cables to Africa, so nothing odd about it. Techcrunch:
Meta also noted in the 8-K that it is narrowing capital expenditures for 2023 by $2 billion at the top end. Capex estimates are now between $34 billion and $37 billion, versus $34 billion and $39 billion previously. Meta doesn’t detail here which areas will be hit by those cuts — capex can include any number of things such as data centers and network infrastructure and AI, but not strictly Meta’s costly “metaverse” effort (which may have server and AI investments but is mostly an R&D investment, as Ben Thompson notes). It does note that the latter of these is not looking very bright.
Then second, if you wonder what it means that capex doesn’t “strictly include” Metaverse “but is mostly [something else]”, and ask how much of that is ‘Metaverse’ as an upper bound, apparently the answer is ′ as low as $0′, because R&D is defined by the US GAAP financial accounting standard to not be ‘capex’ but a different category altogether, ‘opex’: it’s treated as an operating expense you incur, not as purchasing an asset. (Many argue that R&D is in fact more like ‘buying an asset’ than ‘spending money on operating normally’ and should be under ‘capex’ rather than ‘opex’ - but AFAICT, in the numbers FB is reporting, it would not be.) So “$36b” is not only not just the Metaverse expenses, it’s none of the Metaverse expenses by definition.
(The actual Metaverse number is something like $10b/year, IIRC. Which is pretty staggering on its own—as more than one person has asked, where is it all going? - but a lot smaller.)
After market close on 10/26/2022, Meta guided an increase in annual capex of ~$4B (from 32-33 for 2022 to 34-39 for 2023), “with our investment in AI driving all of that growth”. NVDA shot up 4% afterhours on this news. (Before you get too alarmed, I read somewhere that most of that is going towards running ML on videos, which is apparently very computationally expensive, in order to improve recommendations, in order to compete with TikTok. But one could imagine all that hardware being repurposed for something else down the line. Plus, maybe it’s not a great idea (for us humans, collectively) to train even narrow AIs to manipulate humans?)
How much do you expect Meta to make progress on cutting edge systems towards AGI vs. focusing on product-improving models like recommendation systems that don’t necessarily advance the danger of agentic, generally intelligent AI?
My impression earlier this year was that several important people had left FAIR, and then FAIR and all other AI research groups were subsumed into product teams. See https://ai.facebook.com/blog/building-with-ai-across-all-of-meta/. I thought this would mean deprioritizing fundamental research breakthroughs and focusing instead on less cutting edge improvements to their advertising or recommendation or content moderation systems.
But Meta AI has made plenty of important research contributions since then: Diplomacy, their video generator, open sourcing OPT and their scientific knowledge bot. Their rate of research progress doesn’t seem to be slowing, and might even be increasing. How do you expect Meta to prioritize fundamental research vs. product going forwards?
Zuckerberg has made a huge bet on VR/”The Metaverse”, to the tune of multiple times the cost of the Apollo Program. The business world doesn’t seem to like this bet, people are not bullish on VR but are very bullish on AI. So the pressure is on Mark to pivot to AI, but also to pivot to anything that is productizable.
Spot check: the largest amount I’ve seen stated for the Metaverse cost is $36 billion, and the Apollo Program was around $25 billion. Taking into account inflation makes the Apollo Program around 5 times more expensive than the Metaverse. Still, I had no idea that the Metaverse was even on a similar order of magnitude!
The $36b number appeared to be extremely bogus when I looked into it the other day after seeing it on Twitter. I couldn’t believe it was that large—even FB doesn’t have that much money to burn each year on just one thing like Metaverse—and figured it had to be something like ‘all Metaverse expenditures to date’ or something else.
It was given without a source in the tweet I was looking at & here. So where does this ‘$36b’ come from? It appears to first actually be FB’s total ‘capex’ reported in some earnings call or filing, which means that it’s covering all the ‘capital expenditures’ which FB makes buying assets by building datacenters or underseas fiberoptics cables; $36b seems like a pretty reasonable number for such a total for one of the largest tech companies in the world which is doing things like cables to Africa, so nothing odd about it. Techcrunch:
Then second, if you wonder what it means that capex doesn’t “strictly include” Metaverse “but is mostly [something else]”, and ask how much of that is ‘Metaverse’ as an upper bound, apparently the answer is ′ as low as $0′, because R&D is defined by the US GAAP financial accounting standard to not be ‘capex’ but a different category altogether, ‘opex’: it’s treated as an operating expense you incur, not as purchasing an asset. (Many argue that R&D is in fact more like ‘buying an asset’ than ‘spending money on operating normally’ and should be under ‘capex’ rather than ‘opex’ - but AFAICT, in the numbers FB is reporting, it would not be.) So “$36b” is not only not just the Metaverse expenses, it’s none of the Metaverse expenses by definition.
(The actual Metaverse number is something like $10b/year, IIRC. Which is pretty staggering on its own—as more than one person has asked, where is it all going? - but a lot smaller.)
Thank you for this. I was going by statistics shared in a recent episode of the All-In Podcast, and I took those stats for granted.