A lot look like typical patent spam along the lines of "like X but in the cloud" or "like X but with AI"
"Automatically scaling neural networks based on load" [0] is one such example. Load balancing and scaling is nothing new. Doing it when there's a neural network as the application involved doesn't seem novel. Every flavor of software application shouldn't warrant its own patent for scaling out the computational resources involved, at least not when the supposedly novel method is so basic, but dressed up in neural network jargon. It doesn't even do that in all places. From the patent: "Allocate Processors -> receive load information from the processors -> adjust the number of processors based on the load information."
This is exactly the sort of patent I would expect a patent troll to buy up and then go after anyone (as long as they have money) that does any sort of automatic load balancing or resource scaling. Awful.
Then this one [1] is, quite literally, authenticating a user... but at an ATM machine. Ridiculous.
They aren't all this bad, but a lot of them are. However I'll note one that seems to have a least a little merit: [2] Using IR or temperature changes in keypads to obfuscate ATM pin input. (stealing PINs via thermal imagine is aparantly a real thing)
I think the legal terms may be throwing a bit of a loop. But what I can say is this is now deployed in several contexts by multiple companies (which weren’t when we submitted).
Is it “spam”, I guess you can call it what ever you want. But you’re missing the nuanced problem.
Say as a company A you never submit a patent. Companies B, C, D, E, etc... All patent a portion of your applications before they are deployed or even after their deployed (but the patent is public).
They can then sue you. Each company for the aspects you violated.
Imagine that ten fold. You need (if your a decent target), patents you can cross license. Do I think it’s good? No I don’t. Can I reasonably argue you need to patent as a corporation - yes. I also support patent and copyright reform, but I can still understand the need to protect ourselves in the meantime.
Yes, I understand that companies no longer have much choice but to amass an arsenal of patents like this to deter suits being brought against them.
The issue I took up in my comment was that you seemed to be defending this process to some extent, so I appreciate you clarifying your point of view to mean you were only defending the necessity of companies to engage in this sort of stockpiling of patents. It's not, from a less subjective viewpoint, rational: It is a net drain on the economy and innovation. However, when you find yourself in a crazy situation that you can't significantly alter, the rational response is often something that is itself irrational.
I often think of war as the prime example of this... It's insane that the US dropped two nuclear bombs on Japan. Heck, it's insane that such weapons were developed to begin with. And year the entire broader context was insane. Fighting a more conventional war against Japan would have been even more crazy, costing more lives and prolonging the war even longer. It's an appropriate metaphor to call the crazy patent situation "patent wars". (It's also part of why I like things like the study of optimal strategies in iterated prisoner dilemmas... it gives a framework for how to think about rational decision making under irrational conditions.)
> Using IR or temperature changes in keypads to obfuscate ATM pin input.
A friend of mine described to me this exact idea over four years ago — well before the linked patent was issued. Does he have grounds to overturn that patent!?
It wouldn't count as prior art for the purpose of invalidating the patent unless his idea was in some way identifiable during a reasonable search for prior art. From a legal perspective, prior art is usually interpreted as the available information about the thing being patented. If your friend told you about it over drinks at a bar, that's not prior art. For that matter, trade secrets for even very old things aren't prior art either because the information is not available. If your friend wrote a blog post that hit the mainstream media, or at least the front page of HN, it gets to be a bit of a grey area.
"Automatically scaling neural networks based on load" [0] is one such example. Load balancing and scaling is nothing new. Doing it when there's a neural network as the application involved doesn't seem novel. Every flavor of software application shouldn't warrant its own patent for scaling out the computational resources involved, at least not when the supposedly novel method is so basic, but dressed up in neural network jargon. It doesn't even do that in all places. From the patent: "Allocate Processors -> receive load information from the processors -> adjust the number of processors based on the load information."
This is exactly the sort of patent I would expect a patent troll to buy up and then go after anyone (as long as they have money) that does any sort of automatic load balancing or resource scaling. Awful.
Then this one [1] is, quite literally, authenticating a user... but at an ATM machine. Ridiculous.
They aren't all this bad, but a lot of them are. However I'll note one that seems to have a least a little merit: [2] Using IR or temperature changes in keypads to obfuscate ATM pin input. (stealing PINs via thermal imagine is aparantly a real thing)
[0] https://patents.google.com/patent/US10614360B2/en?inventor=A...
[1] https://patents.google.com/patent/US20200012778A1/en?invento...
[2] https://patents.google.com/patent/US20200202015A1/en?invento...