Tangentially related, Technology Connections published an interesting video just a few days ago, "The Antique Microwave Oven that's Better than Yours": https://www.youtube.com/watch?v=UiS27feX8o0
Yes you can make a microwave that cooks certain foods to perfection but it turns out that you can also get very close by smashing a +30second button or spinning a dial. You don't even have to read a manual.
That'd be quite awesome tbh.
Bonus if, after 18 months, you need to buy a replacement microwave because a security certificate has expired.
And then you get a Netflix vs. HBO Max problem where one subscription doesn't have potato recipes and the other doesn't have anything that contains butter.
The idea is good. It'll just take a while before someone gets the implementation right.
The June Oven does this, and it works pretty darn well IMO. It's using just a video camera to do food recognition, but it's fairly high-accuracy for recognizing supported foods. The first few times you do it, frankly, it feels kinda magical.
But, actually, a lot of my usage doesn't lean on the pre-set programs. Rather, I do a lot of 'reheat leftovers' by sticking a temperature probe into the food, hooking it in to the oven, and having the oven automatically reheat until the internal temperature is warmed.
Unless the microwave can auto detect what the food is I can't see how this can be fixed.
You don't need that, there are also bookmarks and an alphabetical index. And actually as every item listed on the screen has a separate button, it's almost always less button presses than smashing your +30 seconds button.
Pressing the same button rapidly 4 times is faster and easier than pressing 3 buttons after reading a menu
Isn't that only the case if you already know how long to microwave a potato?
I think some of the su comments suggesting AI/ML or IoT seem to miss his point: in a lot of cases, good old fashioned individual problem solving expressed as decision trees and lookup tables can do wonders.
Sure, it's a lot more ambitious as a VC-backed company to "solve" cooking by having the one algorithm to rule them all, but in most cases it's not even known that this is an algorithm that is decidable.
What I know we can do right now is to add a sensor to cook a potato, heat a hot water bottle, and cook popcorn with minimal logic and good testing. This way you'll also have a product that helps most users do 80% of their tasks while the cook-anything algorithm is still burning millions in stealth. If you're smart you may even collect enough data and operational knowledge building this product to put you way ahead of other teams, should you wish to go the AI route.
Imagine finding a base microwave that works, adding the humidity sensor, AND a thermal camera of some sort... you could bring food up to temperature and hold it there. The tricky bit is shielding the camera from the RF field.