Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 1 of 24
`
`Exhibit 36
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 2 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`Page 1
`
` UNITED STATES DISTRICT COURT
` SOUTHERN DISTRICT OF NEW YORK
`
`NETWORK-1 TECHNOLOGIES, INC.,
` PLAINTIFF,
` vs. No. 14 Civ. 2396 (PGG)
`GOOGLE LLC AND YOUTUBE, LLC, 14 Civ. 9558 (PGG)
` DEFENDANTS.
`_____________________________
`
` CONFIDENTIAL - OUTSIDE ATTORNEYS' EYES ONLY
` VIDEOTAPED DEPOSITION OF
` JIM SCHREMPP
` Wednesday, September 25, 2019
`
`Reported By: KATHLEEN WILKINS, CSR #10068,
` RPR-RMR-CRR-CCRR-CLR-CRC
`
` BE IT REMEMBERED that on Wednesday, September
`25, 2019, commencing at the hour of 11:04 a.m.
`thereof, at CROWELL & MORING, Three Embarcadero
`Center, 26th Floor, San Francisco, California, before
`me, Kathleen A. Wilkins, RPR-RMR-CRR-CCRR-CLR-CRC,
`a Certified Shorthand Reporter, in and for the State
`of California, personally appeared JIM SCHREMPP, a
`witness in the above-entitled court and cause, who,
`being by me first duly sworn, was thereupon examined
`as a witness in said action.
`____________________________________________________
` DIGITAL EVIDENCE GROUP
` 1730 M Street, NW, Suite 812
` Washington, D.C. 20036
` (202) 232-0646
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 3 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
` MR. LEDAHL: Brian Ledahl from
`Russ August & Kabat on behalf of the plaintiff,
`Network-1.
` MR. RAMSEY: This is Gabe Ramsey, along
`with my colleague Kayvan Ghaffari, from
`Crowell & Moring, counsel for the witness,
`Jim Schrempp.
` THE VIDEOGRAPHER: Will the court reporter
`please swear in the witness, and then we can begin.
` JIM SCHREMPP,
` having been duly sworn,
` was examined and testified as follows:
` EXAMINATION BY MR. DANG
`BY MR. DANG:
` Q. Would you please state your name for the
`record.
` A. Jim Schrempp.
` Q. Have you ever been deposed?
` A. Yes.
` Q. Okay. Just a quick refresher on how this
`will go.
` I've got a series of questions for you.
`
`Page 10
`
`If any of them are poor questions or could be
`rephrased, definitely let me know, and I'll try to
`ask a better question.
` I was going to try to take a break maybe
`every hour or so. But if you want a break sooner or
`you want to keep going, just let me know.
` A. Mh-hmm.
` Q. Does that sound good?
` A. Yeah.
` Q. All right. Let's start with your
`education.
` Where did you go to school?
` A. I went to college at Cal Poly, San Luis
`Obispo.
` Q. And did you receive your undergrad there
`or your master's or any other degrees there?
` A. Undergraduate degree in computer science.
` Q. Okay. And when did you graduate?
` A. 1980.
` Q. And did you take any classes on content
`recognition or machine learning while in school?
` A. To be honest with you, I can't recall.
`Page 11
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
` Q. Okay. Do you have any papers related to
`audio content recognition?
` A. No, I don't think so.
` Q. What about patents?
` A. Classes in patents? I'm sorry. I
`didn't --
` Q. Oh, sorry.
` Do you have any patents related to the
`field of audio content recognition?
` A. I do.
` Q. About how many, would you say?
` A. I think I'm named inventor on, I would
`guess, 20, 25. I'm not precise. Don't know a
`precise count.
` Q. Okay. And when did you start obtaining
`those patents on audio content recognition?
` A. Probably in the year 1990 -- '99, I think
`we started.
` Q. And where were you employed in 1999?
` A. Audible Magic.
` Q. Are you still employed at Audible Magic?
` A. No.
`
`Page 12
`
` Q. Do you have any current relation with
`Audible Magic?
` A. I am on the board of directors.
` Q. And when did you join the board of
`directors?
` A. In -- when I joined the company.
` Q. Okay. What was your position at Audible
`Magic when you joined the company? Was it just as a
`board -- on the board of directors, or did you have
`another job title?
` A. It was vice president of software
`development.
` Q. And around that time, in 1999, did you
`ever begin working on a product called Clango?
` A. Yes.
` Q. And in general terms, what was Clango?
` A. Clango was a media recognition tool that
`would run on a personal computer, monitor a stream
`of audio, return -- if it could identify audio, it
`would. It would return title, artist, a link to
`purchase. I think you could add it to a favorites
`list, you know, that kind of stuff.
`
`Page 13
`Pages 10 to 13
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 4 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
` Q. Is it fair to say that Clango identified
`audio by its content?
` A. Yes.
` MR. LEDAHL: Objection. Form.
`BY MR. DANG:
` Q. And what was your role in working on
`Clango in 1999?
` A. I was vice president of software
`development. I was responsible for the whole
`product.
` Q. Okay. And who else at Audible Magic was
`working on Clango at that time?
` A. I don't know if I have a complete list. I
`would say Erling Wold, Tom Blum, Doug Keislar,
`Jim Wheaton, Alan Edel, Lou Kvitek. We had some
`contractors in at the time. They didn't -- they
`didn't stay long. That's all I can -- there may
`have been more, but, yeah.
` Q. And why try to identify songs based on
`their audio content?
` MR. LEDAHL: Object to the form.
` THE WITNESS: State -- could you give me
`Page 14
`
`the question again.
`BY MR. DANG:
` Q. Yeah. Why -- why were you working on a
`product that sought to identify audio by its
`content?
` MR. LEDAHL: Same objection.
` THE WITNESS: So we had -- we had this
`idea that there was a market for -- audio was being
`streamed, and people didn't have metadata
`associated -- they were starting to listen to audio.
`It was either being streamed or they were playing it
`from files on their system, and that they wouldn't
`know necessarily what the name of the exact title or
`the artist was for the particular piece of music or
`whatever audio they were listening to at the time.
`I think they were music focused.
` So we thought, well, if we could identify
`it, then we could link those people to some kind of
`a purchase, and we could make money, you know,
`from -- by selling -- I think we were originally
`selling Amazon CD -- you know, we would buy the CD
`at Amazon and pass it on to the person.
`
`Page 15
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`BY MR. DANG:
` Q. And do you remember around when you first
`had this idea?
` MR. LEDAHL: Objection. Vague.
` THE WITNESS: 1999, I believe. There
`was -- yeah. I think the first idea for this --
`it's pretty clear in my mind. I don't know the
`date. But we were on a ski trip in Tahoe, and we
`had this idea of pushing a button on the radio.
` Somebody said, "Boy, I don't know what
`that song is on the radio."
` And another person said, "Boy, I wish you
`could push a button on the radio and get the
`answer."
` And I was sitting in the back seat of this
`Tahoe, you know, Ford -- whatever the brand is, you
`know, SUV. And I remember thinking, well, actually,
`we could do that. You know, that could be done.
` And then it morphed from there over time
`into identifying content off the -- that people were
`listening to on their laptops or their PCs rather
`than being over the radio. But that was -- that was
`Page 16
`
`the genesis point. So pretty clear in my mind for
`that.
`BY MR. DANG:
` Q. Do you have an understanding of what the
`term "watermarking" means?
` A. Yeah. Generally familiar with that.
` Q. And what do you -- what is your
`understanding of the term "watermarking" in this
`field?
` A. In this field, my belief is watermarking
`means inserting some -- some binary files into a
`file that can later be extracted to identify that
`file or that stream of bits. Tip -- yeah.
` Q. At the outset of the Clango project, did
`you give any consideration to using watermarking to
`identify audio?
` MR. LEDAHL: Object to form.
` THE WITNESS: I don't know that we -- we
`probably thought of -- tried to think of a lot of
`things, but I don't think watermarking was practical
`for that kind of application. It just didn't --
`didn't make sense to us, or I don't think it did,
`Page 17
`Pages 14 to 17
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 5 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`yeah.
`/ /
`BY MR. DANG:
` Q. And could you describe a little more why
`you thought that watermarking wasn't practical.
` A. Well, we were trying to identify
`content -- excuse me -- that would be available in
`the wild. So we couldn't necessarily -- we wanted
`to identify content that had been released you know,
`years ago, so a CD of The Doors that had been
`released years before watermarking was even
`available, it wouldn't have a watermark in it.
` We wanted to identify content that didn't
`require us to insert anything into the initial
`stream. We wanted to identify content that was
`available from anywhere.
` Q. And did Clango have any competitors at the
`time it was being developed?
` MR. RAMSEY: Objection. Vague.
` THE WITNESS: Yeah, not that I'm aware of.
`I mean not that I recall, I should say.
`BY MR. DANG:
`
`Page 18
`
` Q. Did Audible Magic work with a company
`called Muscle Fish in developing Clango?
` A. We did, yes.
` Q. What was Muscle Fish?
` A. My understanding, Muscle Fish was a
`consulting group of four engineers that had
`developed an audio identification technology. And
`we worked with them to apply that technology to --
`to our solution for Clango.
` Q. I know it's a while ago. Do you remember
`the four engineers who --
` A. I do.
` Q. -- formed the consulting group?
` A. I do. It was Tom Blum, Erling Wold,
`Jim Wheaton, and Doug Keislar.
` Q. And what was the role of Muscle Fish in
`developing Clango, then?
` MR. LEDAHL: Objection. Vague.
` THE WITNESS: Muscle Fish supplied the --
`just as a form here, they object, I still answer?
` MR. RAMSEY: Yeah, yeah, yeah.
` THE WITNESS: Okay. Thanks.
`
`Page 19
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
` So you will tell me if I shouldn't answer?
` MR. RAMSEY: Yes, sir, I will. I will be
`very clear about that.
` THE WITNESS: I thought so. I was just
`making sure.
` So could you repeat the question, then.
`BY MR. DANG:
` Q. Yeah.
` What was the role of the folks at Muscle
`Fish in developing Clango?
` A. So they were responsible for the
`underlying audio identification technology, the core
`thing that would actually make it work.
` Q. And did you work with everyone at Muscle
`Fish, or were there particular individuals at Muscle
`Fish who you worked with?
` A. Well, in '99 -- or 2000 -- early, I guess
`my primary contact was Erling Wold in those days.
` Q. And what would Erling -- what was the form
`in which Erling Wold would provide you with work
`product related to Clango?
` MR. LEDAHL: Object to form.
`
`Page 20
`
` THE WITNESS: At the time, we would be --
`you know, so we were developing software that would
`wrap around a library. And he was supplying a
`library to us that we would incorporate in our
`product. So I would -- his delivery would probably
`be -- probably be email or FTP. I'm not -- I
`actually don't remember exactly how we got the
`product delivered.
`BY MR. DANG:
` Q. Okay. Well, let's take a step back.
` What in -- what in general terms is a
`library?
` A. So a library would be a piece of software
`produced that could be linked into or somehow called
`from another piece of software to do some particular
`function, is my just rough definition.
` Q. And generally speaking, how often was
`Erling providing you with these libraries?
` MR. LEDAHL: Objection.
` THE WITNESS: That would -- pretty
`frequently. I mean, I think we used to get them
`sometimes -- it could be a day-by-day things if
`Page 21
`Pages 18 to 21
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 6 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`there were bugs that were being fixed. If there
`were major enhancements, it could also -- you know,
`a major enhancement might not come every day, but it
`might come every week. It was pretty fast.
` I mean, back in those days, we were
`doing -- it was pretty fast-running development. We
`were trying to get this thing out the door, so ...
`BY MR. DANG:
` Q. So is it fair to say that you would
`incorporate Erling's libraries into some version of
`Clango?
` MR. LEDAHL: Objection.
` THE WITNESS: Yeah, I -- yes. There --
`yeah.
`BY MR. DANG:
` Q. And based on all of this experience, do
`you have personal knowledge of the development of
`the Clango system?
` MR. LEDAHL: Object to form.
` THE WITNESS: I do.
` MR. DANG: I'm handing the witness what's
`been -- if you could mark this as Exhibit 1.
`Page 22
`
` (Whereupon, Deposition Exhibit 1
` was marked for identification.)
` MR. DANG: And I'll note for the record
`that this was previously marked as Exhibit 1 at the
`deposition of Erling Wold.
` Q. Do you recognize this email?
` A. Well, it looks like -- yes. It looks like
`a -- looks like an email I would have sent to
`Erling, setting expectations. It looks like this
`one -- actually, as I -- I'm going to correct again.
` I think Tom Blum was actually our business
`contact at Muscle Fish, and Erling was our technical
`contact, but -- yeah. And I would have mostly
`worked with Erling. Yeah. So this looks like --
`this looks like a document I would have sent, yeah.
` Q. And does this refresh your recollection as
`to you sending this email -- as to whether you did,
`indeed, send this email?
` MR. LEDAHL: Object to form.
` THE WITNESS: It looks like -- it looks --
`this looks like an email I would have sent, yes. I
`mean, I -- looking at -- I don't know how to say it.
`Page 23
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`I mean, looking at it now, I think it looks like an
`email I sent, yes.
`BY MR. DANG:
` Q. And what were the circumstances under
`which you sent this email?
` MR. LEDAHL: Objection. Calls for
`speculation.
` THE WITNESS: So it looks to me like this
`is coming from a personal account, so this would
`have been very early in development as opposed to --
`and it's got the Wired Air, which is our early --
`the early name of Audible Magic.
` It looks like -- as I read it to refresh
`my memory here, it looks like they had asked --
`Erling had probably asked for a specification of
`what we wanted from them in the interface between
`our code and the library and what we wanted the
`library to perform. And I was trying to lay out a
`set of kind of rough specifications for him that he
`could work to.
`BY MR. DANG:
` Q. Okay.
`
`Page 24
`
` A. Kind of -- yeah.
` Q. Were you going to say anything else or --
` A. No. It just looks like I'm trying to set
`expectations on what it should do, how fast it
`should -- you know, how big a database it should
`cover, what the performance -- the false positive
`rate -- did I mention false positives in here? My
`eyes are a little --
` Q. Take your time.
` A. Yeah. Exactly. Thanks.
` An incorrect identification less than
`1 percent of the time and less than 4 percent of the
`time fail to identify -- yeah.
` So I'm trying to set some -- yeah. It's
`like working with a contracting company. I'm trying
`to set some parameters on what would be an
`acceptable outcome to us.
` Q. So the parameters in this email are things
`that you envisioned; is that correct?
` A. Yes.
` Q. Okay. And just to clarify for the record,
`the system that you're discussing here, is that the
`Page 25
`Pages 22 to 25
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 7 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`initial development of the Clango system?
` MR. LEDAHL: Objection. Vague.
` THE WITNESS: Yes. Mh-hmm.
`BY MR. DANG:
` Q. And let's discuss at a high level that
`program that you -- that you had envisioned here in
`this email. And I don't believe I asked you.
` Do you recall when it was sent?
` A. The date on this email is November 29th,
`1999, so -- 5:00 p.m. I believe that's when it's
`sent?
` Q. Okay. And does that refresh your
`recollection as to when you recall this email being
`sent.
` A. It -- yeah. It's -- it's in sync with my
`recollection, yeah.
` Q. Okay. And what was the purpose of this
`new application?
` A. This was the application we -- this was an
`application to identify audio that was intercepted
`on someone's PC and do the identification we
`discussed earlier and return that to the user.
`Page 26
`
`Yeah.
` Q. And just so I -- could you describe a
`little bit more how that identification would work.
` MR. LEDAHL: Object to form.
` THE WITNESS: So -- and, you know, I could
`quote a little bit from the email here, but -- so
`when the user clicks the "ID this button" -- so
`there would be a button on the application that
`says, you know, ID this, when the user clicks the ID
`this button, the client would transmit the last 180
`seconds' worth of analysis to a central server. So
`we're going to do some analysis on the local system
`and then send that in.
` And that would probably be to extract the
`fingerprint from the -- to compute a fingerprint --
`not extract, compute a fingerprint of the audio that
`we've intercepted. The server would then invoke a
`lookup system that would tell us if any of the
`fingerprints from the database on the server were
`present in the sample. And then we add the
`appropriate content metadata -- title, artist, and
`link. We return that to the client for display.
`Page 27
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`BY MR. DANG:
` Q. Okay. Let's -- let's break that down a
`little.
` So when you say it would transmit the last
`180 seconds' worth of audio to the server --
` A. I think I didn't say "audio." I didn't
`mean to say audio.
` Q. Okay. Got it. Got it.
` So what would it -- what precisely would
`it transmit? Would it be the audio, or would it be
`transmitting something that was analyzed?
` A. It was -- it would transmit the analysis
`of the audio. So there's -- there's -- there would
`be two parts to the Muscle Fish libraries that they
`would give us. One would reside on the client
`system, and that would be where the fingerprint
`would be computed. And then another part would
`rely -- would reside on the identification server,
`and that's where the lookup would be done.
` Our team would write wrappers on both of
`those, one wrapper for the client side, so that
`there's some user interface and behavior there,
`Page 28
`
`another wrapper on the ID server side that would,
`you know, expose this -- this lookup and the results
`to the internet and give some way for those two to
`communicate.
` Q. Okay. And just to clarify, what's a
`wrapper?
` A. That would be software that sits on --
`that calls -- that would be a software that calls a
`library and uses a library.
` Q. Okay. And upon the identification of the
`piece of audio, did you envision sending anything in
`addition to the metadata back to the user?
` MR. LEDAHL: Objection.
` THE WITNESS: Well, it says -- it does say
`right here, "Title, artist, link," et cetera.
` So, yeah, we envisioned sending -- I mean,
`title, artist. We thought we'd send back genre. We
`thought we'd send a link where somebody could
`purchase it or a link that would -- yeah. We
`envisioned sending all kinds of stuff back, but ...
`BY MR. DANG:
` Q. And do you see in the fourth -- I believe
`Page 29
`Pages 26 to 29
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 8 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`it's the fourth paragraph, there's a sentence that
`says:
` "Brute force won't yield the
` performance or accuracy needed for
` this."
` Do you see that?
` A. Mh-hmm.
` Q. What did you mean by "brute force" here?
` A. Well, that would be -- early on, you know,
`we were thinking -- from talking with Erling, who
`was really the expert on all that, that as the size
`of the database grew, brute force would mean you'd
`have to just keep grinding through looking at every
`single place and everything, and that that would
`take too long.
` And so I think I was telling him that
`you -- I think I'm telling him:
` "Brute force won't yield the
` performance or accuracy needed for
` this. Your expertise in both the
` analysis needed" --
` (Reporter clarification.)
`
`Page 30
`
` THE WITNESS: So it says:
` "Brute force won't yield the
` performance or accuracy needed for
` this. Your expertise" -- and I'm
` referring to Erling at that point,
` his -- you know, Erling's
` expertise -- "in both the analysis
` and identification make you the
` perfect place to implement some
` kind of clustering of the
` fingerprints prior to lookup or
` some other strategy that optimizes
` the actual identification."
` That was all -- we just wanted to say -- I
`wanted to make sure, I think, in this that he
`understood we couldn't just do a brute force lookup.
`That was going to scale not -- wasn't going to scale
`up the way we needed with the number of
`fingerprints. And so what I wanted him to do is
`make sure he was putting some kind of scheme
`indexing on top of that to make it work faster.
`BY MR. DANG:
`
`Page 31
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
` Q. And when you say "scale up the way we
`needed," what sort of scale are you talking about?
` MR. LEDAHL: Object to form.
` THE WITNESS: Well, we -- we thought we
`were going to have to -- I mean, even in here I'm
`talking about 500,000 songs' fingerprints, you know,
`so 500,000 songs in the database.
` I think we believed -- we were trying to
`estimate how many -- we were thinking we were
`getting things that are played on the radio and
`streamed online, so we thought maybe 500,000, a
`million songs might cover most of the stuff that was
`streamed at that time.
`BY MR. DANG:
` Q. And did you continue working on the Clango
`system after this email?
` A. Yes. Yeah.
` MR. DANG: All right. Okay. Handing the
`witness what -- if you could mark as Exhibit 2.
` (Whereupon, Deposition Exhibit 2
` was marked for identification.)
` THE WITNESS: Thank you.
`
`Page 32
`
`BY MR. DANG:
` Q. Take a moment to review this email.
` A. 2000.
` Brings back good memories, yes.
` Q. Do you recognize the email?
` A. I do as -- yeah.
` Q. And who sent the email?
` A. An email from Erling Wold to me.
` Q. When did Erling send you the email?
` A. February 3rd, 2000, 2:40 p.m.
` Q. As you mentioned, does reviewing this
`document refresh your recollection of receiving this
`email?
` A. Mh-hmm, yes.
` Q. And what were the circumstances under
`which you -- under which Erling sent this email to
`you?
` MR. LEDAHL: Objection.
` THE WITNESS: It looks like Erling is --
`according to this, we asked earlier how we got
`libraries delivered. Looks like he zipped them up
`and put them in a zip file and emailed them to us.
`Page 33
`Pages 30 to 33
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 9 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`And this is -- this is a -- an email to me with
`the -- with the zip files for the current library
`that he -- whatever the latest version was he had,
`he was sending it to me, yeah.
`BY MR. DANG:
` Q. Okay. And do you see the first sentence
`discussing:
` "Attached a zip file with the
` current API implementation"?
` A. Um-hmm.
` Q. What's an API implementation?
` A. That would be the interface that our
`wrapper that we talked about earlier would use to
`call his library. And so the API is the
`specification between his library and our code, and
`it's implemented in his library. So we -- yeah.
` Q. And do you happen to know what "API"
`stands for?
` A. Application programming interface.
` Q. And how often did -- did Erling provide
`more than one of these API implementations to you in
`the course of developing this?
`
`Page 34
`
` A. Oh, yes. Oh, yes.
` Q. How -- approximately how often did he
`provide you with these API implementations?
` MR. LEDAHL: Object to form.
` THE WITNESS: It could be every day if he
`was bug fixing. If there were big changes, it
`could -- I mean, it could be every week. If we had
`an urgent need, he could do it sooner. Kind of was
`on his schedule.
` We had requirements and needs, as we
`talked about in this last email in Exhibit 1, and he
`was trying to get to that as quickly as he could.
`BY MR. DANG:
` Q. And do you recall what you did with the
`API implementation of this email after receiving it
`from Erling?
` A. Well, I can say what we -- what our
`standard practice would have been at the time, would
`be to unpack it, unzip it, do some little test.
` Notice in here he's including a test
`program called "aumTest." He called that the
`Audible Magic test because we were potentially only
`Page 35
`
`www.DigitalEvidenceGroup.com
`
`Digital Evidence Group C'rt 2019
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`one of their customers at the time, but aumTest.
`And so we could use that to verify on some files
`that it was working the way we expected.
` And then we would check it into SourceSafe
`or -- yeah, we were using SourceSafe at the time.
`Check it into SourceSafe, which is the source
`management and control system where we kept all of
`our -- our source code for our products.
` We would check in his library as a binary
`and then incorporate it in Clango and do another
`build of Clango based on the latest library.
` Q. And just for the record, how do you spell
`aumTest?
` A. A-U-M -- lowercase a-u-m, capital T,
`lowercase E-S-T.
` Q. If you look at the fifth paragraph here,
`the email says:
` "If you don't call MF database
` index, no index will be created,
` and a brute force implementation of
` MF database lookup will be used for
` searching. If you do call MF
`
`Page 36
`
` database index, a new tree-based
` faster lookup will be used."
` Do you see that?
` A. Mh-hmm. I do.
` Q. And did you understand Erling's reference
`to "brute force" to mean the same definition of
`"brute force" that we just discussed?
` A. Yes.
` Q. And what did you understand the term
`"tree-based faster lookup" to mean?
` MR. LEDAHL: Object to form.
` THE WITNESS: So for me -- you know, I
`don't know about the -- I don't know about the
`actual indexing. At this -- I can't remember the
`exact indexing underneath here, but it meant some
`kind of an indexing system that would just give
`us -- would give us faster results for a lookup with
`little or no degradation to the accuracy involved.
`BY MR. DANG:
` Q. And in general terms, how would the
`tree-based faster lookup compare with something like
`a brute force search?
`
`Page 37
`Pages 34 to 37
`202-232-0646
`
`

`

`Case 1:14-cv-02396-PGG-SN Document 239-18 Filed 11/12/20 Page 10 of 24
`
`9/25/2019
`
`Network-1 Technologies, v. Google LLC and Youtube LLC
`Confidential - Outside Attorneys' Eyes Only
`
`Jim Schrempp
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
` MR. LEDAHL: Object to form.
` THE WITNESS: It would be much faster. It
`would take less CPU time and complete more quickly.
`BY MR. DANG:
` Q. Do you -- is it fair to say that Erling
`did provide you with a brute force implementation of
`the search here?
` A. Yes, I think we had both available, both
`brute force and indexed approach.
` Q. What did you use the brute force
`implementation for?
` A. Sometimes -- well, depending on when we
`were testing accuracy, we might -- sometimes the
`indexing would not meet our specification. We would
`find -- we wouldn't get as many hits as we expected
`or we used to -- we had a large test bed of hundreds
`of files that we would run a new version of the
`library against with indexing, and we would compare
`those results to the previous results.
` Anything that was missing -- if we had
`gotten a hit in the previous run and we didn't get a
`hit on the current library, then we might turn --
`Page 38
`
`depending on what Erling wanted to do, he might ask
`us to turn off indexing, run the brute force
`approach because that would get all of the hits that
`were available, and then we could decide if -- if we
`turned off indexing and went for the brute force
`method and we still didn't get an identification,
`then he would know that there's some underlying
`problem in the feature comparison that had prevented
`us from getting an identification.
` If, on the other hand, with brute force
`enabled and we -- we got a recognition, then we turn
`on indexing and we don't get a recognition, then we
`know the problem is somewhere in the indexing code.
`And we want -- that would be our ...
` Q. So is it fair to say that the brute force
`imple