LLMs needing less compute would actually be a good thing for Nvidia due to Jevons paradox. Right now token costs are an impediment to using AI more broadly, and more efficient models would help adoption in cases where AI has proven to be useful, like coding.
Jevon’s paradox is a common talking point but it is not a law of nature. LED lightbulbs use 80% less energy than incandescent but you don’t see people using 5x more lights on their homes. The overall energy used to light homes has decreased.
And even if compute demand were perfectly elastic it’s only a good thing insofar as it drives demand for new Nvidia hardware. If tokens can be served from Apple hardware or Google hardware or Huawei hardware that doesn’t help Nvidia.
As I look up and see three lightbulbs side-by-side over my desk and another one in the lamp on the desk… which is dramatically more light than the old 100W bulb used but also lower energy consumption.
The more apt analogy is between pre-electric and post-electric lighting (~1850 to 1920), when efficiency of light generation rapidly increased, restabilizing at ~222x as much light for the same unit of human labor. [0]
Across that period, first world countries massively increased their demand for light.
Further efficiency increases only matter to individual choices if they take a use case from {economically impossible} to {economically possible}.
I'd offer that by the 1920s, most goings-on in first world urban environments were no longer price constrained in terms of their light usage.
Perhaps. But their huge valuation is based on them supplying the massive buildout of data centers that’s happening / planned.
If that dies because a lot of people’s needs turn out to be met by a system at home they can run a 30b-150b model on, a lot more of that money goes to apple or intel or amd.
It would likely get voided as unconscionable if they just unilaterally demanded it, but it might hold up in specific circumstances (if the user is well-aware of the salary demand when they accepted the contract, and the user gets some proportionate value out of giving Apple a percentage of salary).
A lot of big manufacturers have had recalls (for things like laptop batteries, vehicle batteries, the infamous Hoverboards, etc) so I wonder what Anker's batting average is compared to others. It's clearly a hard problem and squeezing in the level of density that customers expect means potentially thinner safety margins.
It's not just a theory. It's well documented that Apple has a bunch of APIs and protocols (like AirPods proprietary low latency wireless instead of Bluetooth, NFC was not allowed until years after Apple Pay) that are not available to 3rd party developers. They will sometimes open things up after they've given their own products years of head start because apparently owning most of the ecosystem and having undying brand loyalty from their users isn't enough.
Actually for many years even to build regular Bluetooth devices that did anything besides audio, you needed to add a special chip (Made for iPhone chip) to your hardware to verify that it was an authorized Apple Accessory. Pebble had one, but any 3rd party apps that wanted to send data to the watch (like Uber app, sports apps, random indie apps) had to get allow listed to communicate with accessory devices like Pebble.
Ambulances are expensive enough that people are hesitant to call them, sometimes even in life threatening situations.
And if the person is unsure whether the situation is critical, they might try to "sleep it off" rather than driving or getting a ride, because ER is also kind of expensive and you could be stuck there all day.
And Waymo doesn't currently operate on highways for passenger service (I think they have authorization to, but they're only testing on highways right now).
They should be able to get to SFO from Millbrae Ave and San Bruno Ave without getting on the highway proper, although it'll likely be a lot slower unless you're getting a ride from nearby. While SJC can serve downtown SJ and Santa Clara without getting on a highway.
Most likely both agreements had been in negotiations for a while and not something they just pulled together last week in response to SJC, although it's possibly they could have used it as leverage (hey we've talking to SJC ...)
Taxi services can potentially complement public service by filling in the gaps: last-mile connections (home to train station) and backup service late at night when transit runs less frequently or not at all.
There's a risk that robotaxis could become too cheap and people use them for point-to-point transportation because it's faster. This could be mitigated through taxes on robotaxis (with incentives to connect people to transit) and/or car usage in general, or maybe using robo-buses to provide a middle ground between personal convenience and system efficiency.
https://en.wikipedia.org/wiki/Jevons_paradox