I designed / built / installed an 11kw system with 55kwh of batteries in 2024. It was a fun hobby for a year, I loved it. Grid buyback is only $0.03-$0.04/kwh (while cost is $0.37-$0.42/kwh) so, not really worth it. We live among tall trees, but still managed to generate 75% of our yearly solar needs last year.
Permitting was new for me and it was tough to navigate alone, but definitely possible.
What does the lightbulb have to do with the projection? Obviously the lightbulb is not the camera... nor the projector (nothing clear is getting through a frosted white shell). Is the lightbulb just a prop and imagined form-factor?
I feel like the demos are disingenuous because they suggest that the lightbulb is responsible for the entire thing.
Tested your example, the confidence is 0. In smart home context, I can think of an application where the low confidence answers can be forwarded to cloud, whereas the vast majority generic queries solved locally, if the confidence is reliable enough. The response is quite fast by the way.
Mine too! I was not serious. I was just answering your napkin joke. But when I reread now, it is not that obvious. I guess with text you can pass only so much of emotions.
Ah no harm, then. I considered /s, but I really thought it should be plain that I'm not crazy enough to believe one would need any more than a single local device to process such a relatively low complexity command.
Considering this is a 14MB model running almost in a microcontroller, I am fine with such ‘ambigiuous’ queries cannot be handled, as long as the model confidence score is accurate. By the way, I did not test this model thoroughly. I am speculating on the potential of a small model like this. I don’t know if this one is good enough or not.
Turning a voice command into a tool call should be self evidently useful, being able to do that on a small embedded device is the novelty here. In this theoretical example, the thermostat is hosting the model on device and would use no external services.
I confused by the dispatch model. Tool calls typically need some reasonability to be deterministic and, more importantly, predictable in response (o/w GIGO). Why would the thermostat need to interpret a voice command that the node capturing the voice command would not interpret instead?
The node capturing the voice command could be the thermostat. From my understanding they are targeting very small devices.
So that could be a master home automation node, but why not also a single purpose device? I can think of more bad examples than I can good ones, but maybe I am doing some soldering and I need my soldering iron turned up a bit; my hands are full, so doing that by voice would be useful enough. Something I would never link up to a big AI model or home automation network, but could be useful to control by voice.
If it's something that can be burnt directly into a chip and shipped with the products for cheap, maybe that's a more pragmatic way to get AI into small devices (see taalas for a much bigger model doing that, althoug not yet cheap).
Oh I agree that's not unreasonable. I wonder about the harnessing heft required to make it feasible though. If I say I'd like it a bit warmer, a tool can deterministically bump a few degrees while an LLM might bump it 10C. So do we limit the tool's range?
Now you are just making stuff up. The original query was "Make it a little warmer in here". Just because you can come up with some sort of context where the word warmer could mean that you want it cooler does not mean it fits in this scenario which is exactly what people are using LLMs for - to actually understand that context, if it can't do that it's worthless
Permitting was new for me and it was tough to navigate alone, but definitely possible.