Design talent flees back to Apple, Apple buys advanced ML startup

[RockRock8]

Looked at the first 4. All have the same 35 USC 101 issues, looks like. None seem to claim any special hardware (all that matters in a patent is the claims. All the stuff before that is just used to interpret the claims.)

Can you explain how using complex mathematical algorithms to interpret sub-millimeter movements in the cheek and neck via speckle patterns produced by lasers are somehow "lip reading but computerized?" The practical effect is that, yes.

Are you planning to look at the rest also?

 

[Roller]

I have neither the knowledge nor the expertise to say whether this company’s IP is worth it, but I doubt Apple would have spent that much on the acquisition without thorough vetting and cause. It’ll be interesting to see where this goes.

 

[RockRock8]

I have neither the knowledge nor the expertise to say whether this company’s IP is worth it, but I doubt Apple would have spent that much on the acquisition without thorough vetting and cause. It’ll be interesting to see where this goes.

It's the same people who invented PrimeSense and Face ID. They've proven their ability to create novel hardware

 

[Cmaier]

Can you explain how using complex mathematical algorithms to interpret sub-millimeter movements in the cheek and neck via speckle patterns produced by lasers are somehow "lip reading but computerized?" The practical effect is that, yes.

Are you planning to look at the rest also?

Because the claims don’t claim any particular algorithm to do that. Here’s an example from the ‘785 patent:

The claim is directed to (1) receiving signals (we don’t know anything about them other than that they represent facial skin micromovements. No voltages, no encoding. Nothing); (2) determining from the signals at least one word to be spoken… (we aren’t told HOW. No algorithm. No circuitry for doing it. Nothing); (3) instituting an interpretation of the at least one word (what kind of interpretation); (4) causing the interpretation to be presented as the word. (how? sound? text? how is it done?)

You can’t claim an abstract algorithm, but that’s really what is being claimed here.

Another example, from the ‘282 patent:

Operating a light source? How? This is no different than looking at a face in a lit room. Receiving unknown signals representing light reflections. That’s what happens when you look at someone’s face. Analyzing the signals (how?!?) to identify two components (how are the components identified?) Filtering out a component (what kind of filter?)

The claims don’t reveal the actual algorithms, and don’t describe any actual circuits or components for building a machine to do any of this. It’s just using a conventional computer to perform some vague steps.

 

[Cmaier]

It's the same people who invented PrimeSense and Face ID. They've proven their ability to create novel hardware

I’m sure they have great hardware for this - I think I started by saying Apple probably bought them for their hardware and actual designs.

 

[Cmaier]

By the way, you see that “A non-transitory computer readable medium” language? That *used* to be how you could get around the problem I am referring to. Around 2010 or so, the thought was that if you just talked about a computer running software, it was no good, but if you were referring to some sort of memory storage that contained instructions, that was patentable.

In 2015, the Alice decision came down from the Supreme Court, and everything changed. There is a two-step test, that amounts to (1) is it abstract? if not, you’re good. Otherwise (2) do the steps, alone or in combination, somehow amount to something more than a merely abstract idea? If so, you’re good.

Step two rarely saves a patent claim, but it can sometimes. Typically you would have to show something like that the claimed steps somehow improve the operation of a computer or network.

To give you an idea, by the way, in Yu v. Apple, the claims went to the use of multiple cameras in combination to improve a photograph. The federal circuit court of appeals found this claim invalid for the reasons I mentioned:

 

[RockRock8]

I’m sure they have great hardware for this - I think I started by saying Apple probably bought them for their hardware and actual designs.

That's initially why I asked about this: to determine what and how advanced they were (an estimate). I wasn't trying to see if it was good patent, although I'm a little confused. If you're confident that the patent is invalid why would they grant it? Why bother giving someone multiple patents that you say has this obvious flaw?

 

[Cmaier]

That's initially why I asked about this: to determine what and how advanced they were (an estimate). I wasn't trying to see if it was good patent, although I'm a little confused. If you're confident that the patent is invalid why would they grant it? Why bother giving someone multiple patents that you say has this obvious flaw?

The patent examiner focuses on prior art, not the issue I am referring to. Thousands of patents are invalidated every year when they finally get tested by an adversary in a court or in the USPTO after the patent is granted.

 

[RockRock8]

The patent examiner focuses on prior art, not the issue I am referring to. Thousands of patents are invalidated every year when they finally get tested by an adversary in a court or in the USPTO after the patent is granted.

can you explain the purpose of claims vs specification? Are you deliberately not answering about what they're working on? I guess I'm confused. You're looking at this from the perspective of if you challenge it, you could argue it would be invalid. Are you saying they've done nothing, or they're shit at writing patents lol? You said you think Apple acquired them specifically because Q ai has developed advanced hardware etc, but then why would you focus on the patent being potentially invalid

 

[Cmaier]

can you explain the purpose of claims vs specification? Are you deliberately not answering about what they're working on? I guess I'm confused. You're looking at this from the perspective of if you challenge it, you could argue it would be invalid. Are you saying they've done nothing, or they're shit at writing patents lol? You said you think Apple acquired them specifically because Q ai has developed advanced hardware etc, but then why would you focus on the patent being potentially invalid

The primary purpose of the specification is to help inform a person of ordinary skill in the art what the claims mean. For example, if a claim refers to “computing means for calculating [whatever],” you look in the specification to see what those “means” are. Or the patent specification could expressly define a term in a unique way, and then when it’s used in the claims, that tells you what it means.

Traditionally people would write a specification to describe a product in detail, then file multiple patents with the same specification, each with different sets of claims directed to different inventions embodied in the product; they all get the benefit of the original filing date (when determining whether something is prior art, for example). You can do this serially (in which case subsequent applications are called “continuations”) or simultaneously (in which case they are called “divisionals” - don’t @ me patent folks, I’m oversimplifying here) and can even add new stuff to subsequent specifications (“continuations-in-part”) in which case the effective date of filing would depend on when the material in the claims was added.

As for the rest of your question, I am not saying anything about what they’ve done other than I’ve twice now said I assume they’ve got a neat product, based on their pedigree and Apple’s willingness to fork over the dough. I am only commenting on the patents I’ve looked at. The reason I am focussing on patents is because you mentioned patents at some point, so I looked into the patents.

I can’t tell much about their actual hardware, because you can’t assume that the specifications are referring to their actual hardware, and they don’t go into a level of detail that would enable me to tell much.

 

[RockRock8]

The primary purpose of the specification is to help inform a person of ordinary skill in the art what the claims mean. For example, if a claim refers to “computing means for calculating [whatever],” you look in the specification to see what those “means” are. Or the patent specification could expressly define a term in a unique way, and then when it’s used in the claims, that tells you what it means.

Traditionally people would write a specification to describe a product in detail, then file multiple patents with the same specification, each with different sets of claims directed to different inventions embodied in the product; they all get the benefit of the original filing date (when determining whether something is prior art, for example). You can do this serially (in which case subsequent applications are called “continuations”) or simultaneously (in which case they are called “divisionals” - don’t @ me patent folks, I’m oversimplifying here) and can even add new stuff to subsequent specifications (“continuations-in-part”) in which case the effective date of filing would depend on when the material in the claims was added.

As for the rest of your question, I am not saying anything about what they’ve done other than I’ve twice now said I assume they’ve got a neat product, based on their pedigree and Apple’s willingness to fork over the dough. I am only commenting on the patents I’ve looked at. The reason I am focussing on patents is because you mentioned patents at some point, so I looked into the patents.

I can’t tell much about their actual hardware, because you can’t assume that the specifications are referring to their actual hardware, and they don’t go into a level of detail that would enable me to tell much.

Thanks for the in depth response.

Honestly I'd be interested in you giving this patent to one of these transformer models, letting it "read" and summarize the patent, you writing a response of why you think it's invalid back to it, and seeing what it generates text on. It's your stuff, so I won't do it for you, but if you do that, I'd be curious to see what it says, context given around what's valid or invalid for ML patents, and what you think in regards to that. You deal with technology, but do you specifically deal with machine learning algorithm based patents? Because it seems like all of ML is highly abstract, so I'm curious what it might say in response.

 

[Cmaier]

Thanks for the in depth response.

Honestly I'd be interested in you giving this patent to one of these transformer models, letting it "read" and summarize the patent, you writing a response of why you think it's invalid back to it, and seeing what it generates text on. It's your stuff, so I won't do it for you, but if you do that, I'd be curious to see what it says, context given around what's valid or invalid for ML patents, and what you think in regards to that.

LOL. No thanks.

Right now I’m working on invalidating 7 patents on behalf of a big company for exactly these reasons. I have my hands full

 

[RockRock8]

LOL. No thanks.

Right now I’m working on invalidating 7 patents on behalf of a big company for exactly these reasons. I have my hands full

Lol okay. Is there specific information that the USTPO gives on ML related ("AI") inventions as of late? Do you know

 

[Cmaier]

Lol okay. Is there specific information that the USTPO gives on ML related ("AI") inventions as of late? Do you know

No, I haven’t seen any guidance on ML-related inventions. I’ve seen some patents that claim something along the lines of “use ML to do xxxx” be invalidated in federal district courts, though.

 

[RockRock8]

None seem to claim any special hardware

What to you are you defining as special hardware, just curious?

Also, I asked not for general insight on the patents themselves but insight for the specific context of answering how far along you thought they might be. I'm continually pressing on this because I think that you think you're answering that, but I feel like this whole discussion about patents -- while informative -- isn't really what I was asking.

I tried bringing it back around and directly asking again, but you specifically said you answered that question. Respectfully, I don't feel my specific question was answered. It's fine not to answer, but I think your analysis is hyper-fixated on whether the claims are valid or not. That's also why I specifically asked, "Are you saying they've done nothing, or they're shit at writing patents?" The answer seems to be you think this is a good start up because Apple and pedigree etc, plus that not all hardware advancements in a start up may be covered properly by patents, or not all good patents are implemented in hardware properly, but I wasn't really asking that.

I was specifically asking based on the patents they had, how far into whatever venture they were doing do you estimate they were? Obviously this discussion didn't get off to an amazing start because there was an issue finding more than 1 patent, so I tried to help there with giving 3 main people involved that I found referenced on patents

Not to come across as offensive (I'm certainly not mad) I'm just wanting a direct answer, and it's fine if you can't give it. If I could answer it myself, I wouldn't have asked I guess lol.

Do I think they bought it for patents? No. But there's literally zero documentation about this company, and I just was curious what little we could learn from the patents since that's literally the only thing. Not even their website was informative.

 

[Cmaier]

What to you are you defining as special hardware, just curious?

Could be anything. For example, a specific circuit. Patents that claim the use of conventional hardware (i.e. an ordinary phone, an ordinary computer, an ordinary LED, etc.) tend to fall victim to what I’m talking about. You can’t patent an algorithm, a law of nature, etc. You need to patent a specific implementation. So “shine a light on something, then use a computer to figure out what you’re looking at” doesn’t work. A specific circuit that implements some algorithm may very well work.

Also, I asked not for general insight on the patents themselves but insight for the specific context of answering how far along you thought they might be. I'm continually pressing on this because I think that you think you're answering that, but I feel like this whole discussion about patents -- while informative -- isn't really what I was asking.

I can’t tell anything about that from the patents. I’ve said it on here before in old threads, but I never try and draw that type of conclusion from patents. Startups, for example, patent every idea they have, good or bad, in order to increase the value of their I.P. portfolio (if a purchaser doesn’t look too closely) and in order to have fodder to maybe use against someone who sues *them* for patent infringement. There doesn’t need to be any correlation between the patents and the actual work being done; patents are almost never written by the engineers. They just hand a patent agent (not necessarily even a lawyer) a page or two of “my idea is to do XXX” and the patent agent writes up a patent.

I tried bringing it back around and directly asking again, but you specifically said you answered that question. Respectfully, I don't feel my specific question was answered. It's fine not to answer, but I think your analysis is hyper-fixated on whether the claims are valid or not. That's also why I specifically asked, "Are you saying they've done nothing, or they're shit at writing patents?" The answer seems to be you think this is a good start up because Apple and pedigree etc, plus that not all hardware advancements in a start up may be covered properly by patents, or not all good patents are implemented in hardware properly, but I wasn't really asking that.

I mean I think I answered, and I was trying to avoid being crass, but under duress: “they’re shit at writing patents.” I have to assume it’s not “they’ve done nothing,” but, again, that’s not based on the patents - it’s based on pedigree and purchase price.

I was specifically asking based on the patents they had, how far into whatever venture they were doing do you estimate they were? Obviously this discussion didn't get off to an amazing start because there was an issue finding more than 1 patent, so I tried to help there with giving 3 main people involved that I found referenced on patents

Yeah, like I said, I can’t tell how far they are. For $2 billion they must have working prototypes. For all we know they’ve been working with apple for a long time, and the cameras in the new airpod pros coming later this year are specifically doing Q’s algorithms. Putting them in just for hand gestures seems sorta lame.

Not to come across as offensive (I'm certainly not mad) I'm just wanting a direct answer, and it's fine if you can't give it. If I could answer it myself, I wouldn't have asked I guess lol.

Do I think they bought it for patents? No. But there's literally zero documentation about this company, and I just was curious what little we could learn from the patents since that's literally the only thing. Not even their website was informative.

Yeah. Dunno.

 

[RockRock8]

Could be anything. For example, a specific circuit. Patents that claim the use of conventional hardware (i.e. an ordinary phone, an ordinary computer, an ordinary LED, etc.) tend to fall victim to what I’m talking about. You can’t patent an algorithm, a law of nature, etc. You need to patent a specific implementation. So “shine a light on something, then use a computer to figure out what you’re looking at” doesn’t work. A specific circuit that implements some algorithm may very well work.




I can’t tell anything about that from the patents. I’ve said it on here before in old threads, but I never try and draw that type of conclusion from patents. Startups, for example, patent every idea they have, good or bad, in order to increase the value of their I.P. portfolio (if a purchaser doesn’t look too closely) and in order to have fodder to maybe use against someone who sues *them* for patent infringement. There doesn’t need to be any correlation between the patents and the actual work being done; patents are almost never written by the engineers. They just hand a patent agent (not necessarily even a lawyer) a page or two of “my idea is to do XXX” and the patent agent writes up a patent.


I mean I think I answered, and I was trying to avoid being crass, but under duress: “they’re shit at writing patents.” I have to assume it’s not “they’ve done nothing,” but, again, that’s not based on the patents - it’s based on pedigree and purchase price.



Yeah, like I said, I can’t tell how far they are. For $2 billion they must have working prototypes. For all we know they’ve been working with apple for a long time, and the cameras in the new airpod pros coming later this year are specifically doing Q’s algorithms. Putting them in just for hand gestures seems sorta lame.



Yeah. Dunno.

Thank you very much for writing this answer. I'm good now lol.

Although now I do have a new question. They specifically mention "coherent light," ear mounted brackets with that light, etc and analyzing how that light produces a speckle pattern to measure micro movements in the skin. I'm assuming they wanted to keep it broad for applicability, but isn't that specific enough? If they constrained themselves to IR LED, then they couldn't technically use anything else anymore right? So they went with saying coherent light. In one of the patents you mentioned, they also mapped a specific flow with steps, and that sounds like an algorithm.

Specially 12,254,882 figures 4,6,7,8 all seems to be describing a specific pipeline or algorithm for processing stuff in the context of this specific device

 

[Cmaier]

Thank you very much for writing this answer. I'm good now lol.

Although now I do have a new question. They specifically mention "coherent light," ear mounted brackets with that light, etc and analyzing how that light produces a speckle pattern to measure micro movements in the skin. I'm assuming they wanted to keep it broad for applicability, but isn't that specific enough? If they constrained themselves to IR LED, then they couldn't technically use anything else anymore right? So they went with saying coherent light. In one of the patents you mentioned, they also mapped a specific flow with steps, and that sounds like an algorithm.

Specially 12,254,882 figures 4,6,7,8 all seems to be describing a specific pipeline or algorithm for processing stuff in the context of this specific device

Coherent light just means laser, really, so nothing special there. As for the rest of your question:

“Ultimately, the § 101 inquiry must focus on the language of the Asserted Claims themselves, and the specification cannot be used to import details from the specification if those details are not claimed. Even a specification full of technical details about a physical invention may nonetheless conclude with [unpatentable] claims” directed to an abstract idea. ChargePoint, Inc. v. SemaConnect, Inc., 920 F.3d 759, 769 (Fed. Cir. 2019)

“The main problem that [Plaintiff] cannot overcome is that the claim—as opposed to something purportedly described in the specification—is missing an inventive concept.” Two-way Media Ltd. v. Comcast Cable Commc’ns, 874 F.3d 1329, 1338 (Fed. Cir. 2017)

 

[RockRock8]

Coherent light just means laser, really, so nothing special there. As for the rest of your question:

“Ultimately, the § 101 inquiry must focus on the language of the Asserted Claims themselves, and the specification cannot be used to import details from the specification if those details are not claimed. Even a specification full of technical details about a physical invention may nonetheless conclude with [unpatentable] claims” directed to an abstract idea. ChargePoint, Inc. v. SemaConnect, Inc., 920 F.3d 759, 769 (Fed. Cir. 2019)

“The main problem that [Plaintiff] cannot overcome is that the claim—as opposed to something purportedly described in the specification—is missing an inventive concept.” Two-way Media Ltd. v. Comcast Cable Commc’ns, 874 F.3d 1329, 1338 (Fed. Cir. 2017)

I'm not entirely sure how that applies, but I don't know enough.

Another question I had in general about start ups: do companies like Apple seek to rewrite or reissue patents they acquired if they determine the technology they bought is good and they want to patent it and not deal with potential issues the start up had when they wrote

 

[Cmaier]

I'm not entirely sure how that applies, but I don't know enough.

Another question I had in general about start ups: do companies like Apple seek to rewrite or reissue patents they acquired if they determine the technology they bought is good and they want to patent it and not deal with potential issues the start up had when they wrote

re: the first point - it doesn’t matter if something is in a figure in the patent, or in the text of the patent, if it isn’t claimed! The analysis focuses on the claims, and even if you provide a full circuit diagram in the patent it doesn’t do you any good unless your claims include it.

As for the second point, the patentee may file continuations with different sets of claims that perhaps are stronger. There are rules, though:

1) you can only file a continuation if some member of the patent family is still currently being prosecuted at the patent office. If your patent is granted and you pay the fee and it issues as a real patent, and you didn’t file a continuation patent application before that, you are out of luck. Typically patent agents know to file a new continuation at the last minute as a matter of course, so long as the company thinks the patent family may be valuable

2) your new claims have to be supported by the specification. You can’t add something in a new claim that isn’t discussed somewhere in the figures or text of the patent.

3) if you add new text or figures, this is called a “continuation-in-part.” This has various legal effects, which mostly have to do with dates. For example, if you file a patent application in 2000, and a continuation-in-part in 2005, then publications that come out between 2000 and 2005 may be prior art that invalidates your new claims.