Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Google Patents

Google Faces Multibillion-Dollar US Patent Trial Over AI Tech (reuters.com) 27

Posted by msmash from the up-next dept.
Alphabet's Google is set to go before a federal jury in Boston on Tuesday in a trial over accusations that processors it uses to power AI technology in key products infringe a computer scientist's patents. From a report: Singular Computing, founded by Massachusetts-based computer scientist Joseph Bates, claims Google copied his technology and used it to support AI features in Google Search, Gmail, Google Translate and other Google services. A Google court filing said that Singular has requested up to $7 billion in monetary damages, which would be more than double the largest-ever patent infringement award in U.S. history.

Google spokesperson Jose Castaneda called Singular's patents "dubious" and said that Google developed its processors "independently over many years." "We look forward to setting the record straight in court," Castaneda said.
This discussion has been archived. No new comments can be posted.

Google Faces Multibillion-Dollar US Patent Trial Over AI Tech More

Google Faces Multibillion-Dollar US Patent Trial Over AI Tech

Comments Filter:
  • If only independent creation were a valid defense against patent infringement. Perhaps they intend their AI magic to argue novel law in a way that gets the Supreme Court's attention? But not in the usual way of "don't have your briefs written by an AI" of course.

    • Re:New law via AI (Score:5, Interesting)

      by phantomfive ( 622387 ) on Tuesday January 09, 2024 @10:57AM (#64144215) Journal
      It wasn't independent creation. The timeline is like this:

      1) Singular made chips that were more efficient for AI.
      2) They introduced their chips to Google, who brought them to Mountain View and looked at them.
      3) Instead of buying Singular's chips, Google decided to make their own chips.

      Of course, now there are a bunch of companies designing chips that are optimized for AI in various ways, but at the time I don't think there were.
      • If only the summary and the article hadn't contained the statement, from Google, that "Google developed its processors 'independently over many years.'" Perhaps you intend your reference to a timeline to indicate that in your reality, they didn't say this?

        • If only the summary and the article hadn't contained the statement, from Google, that "Google developed its processors 'independently over many years.'"

          Google lies, has lied, and will continue to lie in the future.

          • If only they were somehow incapable of also lying to their lawyers. Perhaps you intend to continue your irrelevant observations as infinitum?
            • It's irrelevant whether Google says they developed their processors independently. The lawsuit says they did not develop their processors independently. Google has a track record of lying.

              Who do you believe? Actually that's a joke, you don't have to believe either one of them.

              • Re: (Score:2)

                by dpille ( 547949 )
                If only you had read the original post and the summary, you might understand the relevance of independent creation to a discussion associated with the summary of an article about the case. Perhaps you've paid for, downloaded, and read Google's answer to the amended complaint in this case and can confirm the words "independently created" and "independent creation" never appear in it?

      • Re: (Score:2)

        by HiThere ( 15173 )

        But under patent law, it wouldn't matter. Perhaps it would matter for the size of the penalty, but I don't think so. For it to matter I think it needs to be a civil law case, though IANAL, so I don't quite know the grounds. "Tortuous interference with business" comes to mind, though.

        • But under patent law, it wouldn't matter.

          Yeap, all that matters is whether Google is infringing on a valid patent.

          If they ripped off another company's ideas, that's just Google being evil, not illegal.

  • [the privacy intruders] said that Google developed its processors "independently over many years."

    Riiiiight. And if they (those privacy intruders) claimed infringement by someone else, they'd dismiss that same claim they're now making in their own defense now. Fucking hypocrites.

  • Does anyone have a link to the "two patents" that were allegedly infringed?

    • No, because google patent search doesn't recognize their validity :)

      But you could find them from the case docket online.

      Singular Computing LLC holds U.S. Patent Nos. 8,407,273 (“the '273 Patent”), 9,218,156 (“the '156 Patent”), and 10,416,961 (“the '961 Patent”), which each describe a method of “Processing with Compact Arithmetic Processing Element[s].” Singular has sued Google for infringing on those patents.

      • Have you looked at the patents? What do you think?

        I've read through a fair number of patents and have written a few patent applications myself. Sometimes I feel like the 'inventor' snuck one through on the patent examiner, the patent is too broad and/or vague. You could claim that a whole lot of things infringe it just because it has a few relevant keywords in it.

    • Here you go:

      https://portal.unifiedpatents.... [unifiedpatents.com]
      https://www.singularcomputing.... [singularcomputing.com]
      https://www.singularcomputing.... [singularcomputing.com]

      Presumably they cover Google TPU [wikipedia.org] technology.

      • Thanks for the links, interesting!

        I 'asked Patty' and she said; "The novelty of this patent lies in the use of at least one first low precision high-dynamic range (LPHDR) execution unit within a computing device." That's pretty broad, and it may be too broad. If so the patent could be successfully challenged.

  • Sign-exponent only floating point? (Score:5, Insightful)

    by Pinky's Brain ( 1158667 ) on Tuesday January 09, 2024 @12:46PM (#64144617)

    I refuse to even think about the claims in US8407273B2 (I have never seen patent claims as obnoxious as that, truly a tour de force) but from the description I'd guess they patented the idea of using sign-exponent floating point for implementing neural networks.

    That seems not novel, take for instance the paper "Multilayer feedforward neural networks with single powers-of-two weights" from 1993.

    • I refuse to even think about the claims in US8407273B2 (I have never seen patent claims as obnoxious as that, truly a tour de force) but from the description I'd guess they patented the idea of using sign-exponent floating point for implementing neural networks.

      That seems not novel, take for instance the paper "Multilayer feedforward neural networks with single powers-of-two weights" from 1993.

      Sorry, but nope. The independent claims of US8407273B2 are clear and solve a technical problem. The application appears to revolve around the idea of adding statistical noise to computations in order to simulate analog computing, thereby enabling the useful exploitation of a large number of computational cores in parallel. A corresponding limitation is at least strongly implied in the claim features "repeated execution" and "statistical mean" and further underlined by the specified minimum percentages of a

      • This is clear according to you :

        comprising at least one first low precision high dynamic range (LPHDR) execution unit adapted to execute a first operation on a first input signal representing a first numerical value to produce a first output signal representing a second numerical value,
        wherein the dynamic range of the possible valid inputs to the first operation is at least as wide as from 1/65,000 through 65,000 and for at least X=5% of the possible valid inputs to the first operation, the statistical mean, over repeated execution of the first operation on each specific input from the at least X % of the possible valid inputs to the first operation, of the numerical values represented by the first output signal of the LPHDR unit executing the first operation on that input differs by at least Y=0.05% from the result of an exact mathematical calculation of the first operation on the numerical values of that same input.

        I will continue to refuse to even dig into that level of bullshit, so lets stick to the description which at least was mostly written by an engineer before lawyers put it in their cauldron to deliver mind poison as above.

        The noise mentioned in the description was only a test case to compare the Logarithmic Number System implementation against (aka sign-exponent aka Power-Of-Two), an approximation of a theoretical analogue implementation. This is clear if you read for instance

  • It does not allow for faster human progress, the whole idea of patenting an idea and/or thing are absurd, and the only people benefiting drom it are patent trolls

    • Patents do what they say, but they also do a lot of other things that lead to the opposite of what they say. On balance they are arguably a net negative. But if they conformed to the original rules and intent, where obviousness and uniqueness mattered, they would be worth having around again.

Slashdot Top Deals

C'est magnifique, mais ce n'est pas l'Informatique. -- Bosquet [on seeing the IBM 4341]

Close