Meta needs more than just open source llama at this point. With Chinese free models on the heels, with reddit and tiktok stealing general attention, and even training from scratch with bert, gpt2, clip becoming accessible.
Meta has stunted VR development by monopolizing the talent and the marketplace. They act like they are the benevolent stewards of the future metaverse, but just like with Facebook, all they want is to dominate and productize the world’s attention.
If you go by the chart at the link, they've lost 54.9 billion dollars since Q4 2020. [1] Do they have any products that can, even in principle, recoup that kind of loss? Is there some kind of tax scheme that can explain this?
[1] - From Q1 2020 to the present: (2.1 + 1.83 + 2.43 + 2.63 + 3.3 + 2.96 + 2.81 + 3.67 + 4.28 + 3.99 + 3.74 + 3.74 + 4.65 + 3.85 + 4.49 + 4.43) -- The trend is clearly accelerating. Could build a cool scatterplot.
The iPhone is a product that is easily worth $1 trillion. That is the goal here, 30% of all money that changes hands while using VR for most consumers.
My theory all along was that they spent maybe 20 billion on Reality Labs and the rest was for Zuckerberg's secret volcano lair but it's getting to a point where that can't really justify it either. Maybe they're working on the weather control laser now?
When you setup a quest, there's an option to allow Meta to use the point cloud data that it collects. I have a strong suspicion they're training some very heavy spatial models (point cloud and camera) that will be used for understanding and content generation augmented onto/around that understanding, in AR scenarios. So, my unfounded guess is that a big chunk of it is compute. Otherwise, there's got to be a hole in the bucket.
Cutting edge technology. Those Orion smart glasses use a totally novel display using micro-projectors bouncing light through waveguides etched into the silicon carbide lens. The glasses also apparently use 10 custom chips.
Both have the same 70 degree diagonal field of view. You're comparing Meta's diagonal to ML2's horizontal.
Its lighter, but years newer and not in significant production.
The full wireless and self contained stuff is impressive, but not really tied to the display is it? Maybe more efficient for less battery drain? Do we have numbers on the display power consumption at a given brightness, and is it from the optics or the image source?
Meta needs more than just open source llama at this point. With Chinese free models on the heels, with reddit and tiktok stealing general attention, and even training from scratch with bert, gpt2, clip becoming accessible.
They got $40 in a quarter. That's pretty insane. Loss of 4.4 is around 10%. Not all that bad.
Meta has stunted VR development by monopolizing the talent and the marketplace. They act like they are the benevolent stewards of the future metaverse, but just like with Facebook, all they want is to dominate and productize the world’s attention.
I am curious what are the deliverables from RL after spending all these money? Have they calculated the real ROI?
Serious question. WHAT EXACTLY sits in Reality Labs? You can't expect me to believe the oculus vr glasses cost 4.4bn in R&D.
All the LLM money is in here. Theyre spending a billion a quarter on training compute alone.
Their "Orion" AR headset and the EMG stuff they acquired from Ctrl Labs
Considering the stocks gain, is it really a loss tho?
Stock owners PnL isn't part of profit/loss calculations.
…how?!
Can someone smarter than me please do the math and figure out how a company burns through >$17 billion dollars in a single year developing anything?
If you go by the chart at the link, they've lost 54.9 billion dollars since Q4 2020. [1] Do they have any products that can, even in principle, recoup that kind of loss? Is there some kind of tax scheme that can explain this?
[1] - From Q1 2020 to the present: (2.1 + 1.83 + 2.43 + 2.63 + 3.3 + 2.96 + 2.81 + 3.67 + 4.28 + 3.99 + 3.74 + 3.74 + 4.65 + 3.85 + 4.49 + 4.43) -- The trend is clearly accelerating. Could build a cool scatterplot.
The iPhone is a product that is easily worth $1 trillion. That is the goal here, 30% of all money that changes hands while using VR for most consumers.
My theory all along was that they spent maybe 20 billion on Reality Labs and the rest was for Zuckerberg's secret volcano lair but it's getting to a point where that can't really justify it either. Maybe they're working on the weather control laser now?
Looks too big to fail now
When you setup a quest, there's an option to allow Meta to use the point cloud data that it collects. I have a strong suspicion they're training some very heavy spatial models (point cloud and camera) that will be used for understanding and content generation augmented onto/around that understanding, in AR scenarios. So, my unfounded guess is that a big chunk of it is compute. Otherwise, there's got to be a hole in the bucket.
Cutting edge technology. Those Orion smart glasses use a totally novel display using micro-projectors bouncing light through waveguides etched into the silicon carbide lens. The glasses also apparently use 10 custom chips.
Same field of view as years old Magic Leap 2 though. Was Silicon carbide an expensive way to work around ML's patents and get the same FOV?
No, it's a much wider field of view than the magic leap 2. 45 degrees vs 70 degrees. And more importantly it's in a much smaller form factor.
https://www.magicleap.io/hubfs/ML_22214-104-1.jpeg
https://techcrunch.com/wp-content/uploads/2024/09/GettyImage...
The magic leap has a much narrower FOV, it's bulkier, and has a wired connection. The Orion are wireless, have a wider FOV, and are slimmer.
Both have the same 70 degree diagonal field of view. You're comparing Meta's diagonal to ML2's horizontal.
Its lighter, but years newer and not in significant production.
The full wireless and self contained stuff is impressive, but not really tied to the display is it? Maybe more efficient for less battery drain? Do we have numbers on the display power consumption at a given brightness, and is it from the optics or the image source?
I hope Zuckerberg is not funneling this into some sort of secret slush fund that will blow up like Enron.
Its crazy, the whole of SpaceX Starship development is under 2 billion $ a year.