It's bad. I believe them not to use it for training, but t means relevant data can and will be exfiltrated by US agencies or through court orders (see NY Times vs. OpenAI, where only traffic without any rentention was safe).
Agreed. The solution will likely be some vision foundation model that directly sends controls to the robot ("move here, grab, move there"), trained by Amazon with RL to integrate collision avoidance, object detection, grasping point detection, grasp verification etc.
> But electricity is under half (30%?40%?) and the rest of that energy isn't fossil-free.
The trick of course is that if you electrify heating and transportation they'll need much less energy. Your average car with an ICE has an efficiency of 20-40%, electric cars have 60-80%. Heating your house with gas has an efficiency of around 100%, heat pumps have 300%-500%.
> If you were to design an entire ATC system from scratch (pilot interfaces, sensors everywhere in the airport and planes etc) it can be automated.
Even then you'll probably run into the long-tail distribution issues, similar to self-driving cars. 99.9% of all situations are pretty standard, but once in a while something so abstruse happens that it's not pre-programmed and requires some creativity to solve.
> What you can probably do is create software which observes traffic and simulates it into the future and notifies the human ATCs about risks.
Fully agree. Some of the recent close calls really were "obvious" much earlier, meaning they were not caused by late course changes.
Is this some kind of calibration then? I'd expect that the probabilities automatically adjust during training, such that in "lock" mode, for example, syntax-breaking tokens have a very low probability and would not be picked even wich higher temperature.
I use them as cheap-man's VPN. A ssh server on a public IP but a non-obvious port brings you into the network, and port forwarding allows you to connect to relevant endpoints in your remote network via localhost:12345.
> My Weird Hill is that we should be building things with GPT-4.
Absolutely. I always advocate that our developers have to test on older / slower machines. That gives them direct (painful) feedback when things run slow. Optimizing whatever you build for an older "something" (LLM model, hardware) will make it excel on more modern somethings.
Putting sulfur into the right layers of the atmosphere seems to be the currently best viable options. It's not overly expensive, either. It acts fast and is reversible.
> Is this the shadow of something natural that we just couldn't see, or just a convenience?
They originally arose as tool, but complex numbers are fundamental to quantum physics. The wave function is complex, the Schrödinger equation does not make sense without them. They are the best description of reality we have.
The thing is that your home's heatpump has an efficiency of 300%-500%. So even if your power plant and power delivery only has say 50% gas-to-electricity-at-home, you are still looking at 150%-250% gas-to-heat-your-house efficiency.
Exactly. It is in general (much) more efficient to burn natural gas in a power plant and use the electricity for heatpumps compared to simply burning gas at home for heating.
> The "Elon process" relies specifically to the goal of getting rid of all dependencies. Musk has spoken extensively about building things from the ground up and not relying on other vendors (in this example complex software dependencies). He says he wouldn't be able to build SpaceX competitively if he had just bought rockets or components.
That I cannot believe. He might have shifted the make-or-buy decisions, but both Tesla and SpaceX do a lot of outsourcing.