November 29, 2024 • 47 mins
Join Lee and Gwilym as they navigate the complex role of AI in intellectual property with Ryan Abbott, Professor of Law and Health Sciences at University of Surrey. Ryan discusses how jurisdictions worldwide are grappling with AI's potential role as an inventor, highlights the lack of international consensus, and the ongoing debates that define this area of law. Get ready to uncover the intriguing journey of AI-generated inventions, the implications of AI's prowess in fields like drug discovery, and the pressing need for evolving patent law to keep pace.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 2 (00:03):
Lee Davis and Gwilym Roberts are the two IPs in a pod
and you are listening to apodcast on intellectual property
brought to you by the CharteredInstitute of Patent Attorneys.
Speaker 4 (00:20):
Hey Gwilym, how are you?
This fine I say fine day it wassnowing this morning.
I've just walked miles, I'mdamp In the snow or just in the
rain.
Speaker 2 (00:29):
No, it's just it doesn't.
It's warmed up, so it's wet.
It's just wet snow aka rain.
Speaker 4 (00:35):
Yeah, no, I came into Waterloo this morning so
excited because there werereally really thick flurries of
snow, and then you lose thatexcitement, don't you?
When it just turns to mush onyour coat and you yeah, yeah,
yeah.
And I thought about jumping ona bus.
I jumped on a bus last week andit took a ridiculous amount of
time to get out of the office,but I thought I'd be a little
bit drier and warmer.
But the queue was horrificallylong and I thought why are all
(00:55):
those people standing there toget on a bus that's going to
take them much longer to get towhere they want to go than it
would walking.
And they get, and they'regetting wet, and they're getting
wet anyway.
That's especially a consultantin yeah do you want a little bit
, little bit news on the seeperfront?
Yes, so do you know?
We've been supported ably overthe last couple of years by ria,
(01:16):
who always lurks in thebackground on these calls and
does the recording for us anddoes a big chunk of the editing
and all of those kinds of things.
Yes, this is her last week.
This is her last podcast no, itcannot be, she's on.
She's off to past his news.
He's off to another membershipassociation.
She's leaving me grill him.
Speaker 2 (01:34):
She's leaving me, she's listening, ria, we know
you're there we know what youdid I don't.
Speaker 4 (01:40):
I don't think we're going to be able to coax her on,
not at the start anyway.
Maybe we can get her on towardsthe end.
But I think we should say a big, big thank you from the podcast
.
I mean, we'll obviously saythank you for all of the work
that Ria's done for SIPA overthe years, but I think we should
just say a big, big thank youfrom the podcast for Ria for
supporting us.
Speaker 2 (01:56):
Actually, yeah, we should Absolutely.
And to say thank you to, weshould say thank you to oh, the
people in the background.
Speaker 4 (02:02):
When neil neil comes on occasionally doesn't know
what sub for me or you if we'renot around.
Yeah, we've got emily andopinia and fran, who, fran, of
course, started off the podcast,really, in terms of working out
the tech for us and doing theediting.
But yeah, emily, emily andopinia and fran in the
background doing all of theediting and making a sound like
we know what we're talking about, which is no mean feat no mean
(02:23):
feat, they're all great.
Speaker 2 (02:24):
Thank you everybody.
Speaker 4 (02:25):
Yeah, no thanks to everyone really, yeah, but yeah,
sorry to see maria go, but uh,hopefully, hopefully she'll find
an amazing new role that she'smoving into, so that's cool she
doesn't get to do a podcast no Ibet she doesn't.
Yeah, I bet there's no talent.
There be no.
So, speaking of speaking oftalent, should we get the guest
on?
Yeah, so, occasionally.
Occasionally, we have returningguests, which is for a couple
(02:54):
of reasons.
Is it so?
It'll either be because theywere so amazing last time that
we feel that we need to revisitthem and the subject, or that
we've just forgotten we've hadthem on before.
So, um, and, and I don't knowwhich, it is with Ryan.
Ryan, who are you?
Tell us a little bit aboutyourself.
Speaker 1 (03:09):
This is after the two of you were just telling me how
amazing I looked and was, andGwilym named his recent child
after me, so I assumed it wasamazingness and I was here for
the kind of final real podcast.
So I took that as a great markof honor.
But you know, it could also bethat we just all have short
memories.
Speaker 4 (03:30):
It was, it was a long time ago that we were last
together on the podcast.
It was May of 21, which waskind of deep, deep, dark
pandemic days.
Speaker 1 (03:39):
It all blurs together .
Speaker 4 (03:41):
Yeah, I had a little listen back earlier and we
certainly started off talkingabout the fact that we may never
see the real world again andwhere we confined to our houses
for for the duration turns out acouple of years, and we were
good yeah, now you can go backto getting wet waiting in a
queue yeah, yeah yeah, podcastsare a really good historical
document for that, actually,because that people just they're
(04:01):
recorded conversations aboutwhat was happening.
Speaker 2 (04:03):
I listened to a whole series of a rudely named one, a
huge long series, and itstarted off in the depths of
lockdown and it was reallyinteresting listening back to
some of the conversations.
Speaker 4 (04:14):
It's a good historical document.
I think so Many started then,didn't they?
Yeah?
Anyway, ryan, can you introduceyourself for the listeners,
because we know who you are?
They don't, unless they listento the last podcast, which is
unlikely.
Speaker 1 (04:27):
Yes, Well, I'm Ryan Abbott.
I'm a professor of law andhealth sciences at the
University of Surrey School ofLaw and a attorney partner at
Brown Harry Smith Conn in theUnited States, and a mediator
and arbitrator with JAMS.
And I assume that I'm here formy infamy associated with the
Thaler or Davis cases, or justbecause Willem thinks I'm very
(04:48):
handsome.
Speaker 2 (04:49):
It'd be great, but you also know that your
knowledge on all that stuff isalso super interesting and I
think we'll hear some updates,things happening, what's going
on All that?
Speaker 4 (04:59):
Yeah, so do you want to give us a potted history of
the project, and I know we didit in depth in the last podcast.
But let's start off with atleast reminding people about the
project, and I know since welast spoke it's been tested in a
number of jurisdictions withdifferent outcomes.
Speaker 1 (05:14):
Indeed, indeed.
Well, I'll give the slightlylonger rambling version, and you
can interrupt me when itbecomes even less interesting.
But we go back about 10 yearsin time and I was working as a
baby attorney in the lifesciences and noticing in drug
discovery that people seem to bedoing some interesting things
with AI and some vendors kind ofthe forerunners to AI-generated
(05:38):
drug discovery, whatever thatis were advertising that they
could do things like if we gavethem a therapeutic target,
they'd have a computer, gothrough a big antibody library
and pick the antibody out thatwould be best for us and this
could be our lead drug candidate.
This is a company I was workingfor and you know we could file
a patent on that and that's thefoundation of a new drug
(05:59):
portfolio.
And I thought well, that'sinteresting because it sounds
like an AI is automating.
You know at least some of whattraditionally makes a human
being an inventor, and whatwould happen if you had an AI
invent something?
And it turned out people hadbeen thinking about this for a
long time.
People have been writing aboutit since at least the 60s and
academically saying you know,because at the time only
(06:21):
academics took an interest inthis, you know you really
wouldn't need a patent becausean AI doesn't care about getting
a patent and patents are toincentivize inventors.
And I thought, well, that's notquite right because really all
of these inventors assign theirpatents to the pharmaceutical
company anyway.
It's the companies that needthe patents to hire the
inventors and to invest in doingclinical trials and to disclose
(06:43):
these things inventors, and toinvest in doing clinical trials
and to disclose these things.
So you'd need a patent just asmuch.
And why wouldn't you want acompany using AI and R&D if you
could get it to do a better jobthan a person at something which
it wasn't really doing,although it's all very
complicated?
But I wrote some law reviewarticles on it and for the first
time, people took an interestin things I was writing and in a
couple of years it had gonefrom.
(07:04):
This is vaguely academicallyinteresting to you know, we're
kind of concerned about this.
Could we pay you a bunch ofmoney to help us think it
through?
And at some point one of mycolleagues said well, why, if
this is an issue, aren't thereany cases on it?
And I said, well, you'd have toreally proactively disclose.
You had an AI do this.
There's a huge incentive not toand you know otherwise, only in
(07:26):
litigation in the US, butthat's probably going to take a
long time.
So I, you know, assembled agroup of patent attorneys to
file a couple of test cases forAI-generated inventions.
I spoke to Dr Thaler, who Iinterviewed for the academic
work, and said you know, wouldyou be interested in having your
AI system generate a couple ofinventions or an invention?
And he and others had done thispreviously, again kind of this
(07:50):
proof of concept experiments.
So he had an AI generate twoinventions.
While he was at it we filedapplications on those of the UK
IPO just to, you know, help theBritish patent attorney milieu.
We did this because they did asubstantive exam before looking
at inventorship.
The patents both passedpreliminary substantive exam.
(08:11):
We filed a form seven and a PCTapplication and said well,
actually these were invented byan AI and the person who owns
them is the owner of the AI, onthe principle that you own
property made by your property.
And then we filed it in 17places and some interesting
things happened.
I could continue to ramble.
Speaker 4 (08:34):
You carry on rambling .
Rambling is good at the moment.
Speaker 1 (08:37):
Ramble away.
So in 2021, although again itdoes all blur together South
Africa granted us a patent Withthe AI listed as the inventor,
which again isn't a matter ofgiving an AI rights.
Most inventors don't haverights to their patents,
although all of this is verycomplicated.
Plus, an AI is not a legalperson that couldn't own
(09:00):
property.
But it is to be transparentabout how an invention is made
and with the AI's owner as theowner of the patent.
South Africa doesn't dosubstantive examination, but
these have already beensubstantively examined and they
do do formalities examination.
So we got a patent there andlater that week the enlightened
Justice Beach and the FederalCourt of Australia issued a 40
something page reasoned decisionabout why, under the Australian
(09:22):
Patent Act, an AI could be aninventor and why at least we had
made the best showing ofentitlement to that patent.
That got overturned at an enbanc decision.
The federal circuit in the USrejected the application on the
basis that an inventor isdefined as an individual in the
patent act, which the SupremeCourt has held.
(09:42):
Usually means a natural person,we pointed out.
Usually they said well, thedictionary says an individual is
a person and we pointed outwell, the very next sentence
says an individual could be athing, so that doesn't really
work.
But there you have that.
And they said well, if you lookthrough the patent act, it
seems to assume an inventor is anatural person.
And we said well, the US patentlaw has this unique provision,
(10:06):
or provision anyway, that youcan't negate patentability on
how something was made, which iswhat they were effectively
doing.
And they said well, for thefirst time we're going to say
this applies only tonon-obviousness.
As a result of that decision,the Supreme Court didn't take
the case.
The US Patent and TrademarkOffice amended their guidance
from AI nothing to see here in2019, to a human being has to
(10:28):
make a significant contributionto each claim in an invention
where the claim is unpatentableand potentially the entire
patent.
You know, inventorship isuniquely important in the US as
a substantive thing which we cancome back to talk about.
So we have one model there.
It went up to the UK SupremeCourt, which I got to argue,
which was great fun, andAmericans just asked if I got to
(10:49):
wear a wig.
I didn't, which was unfortunate, but you'll know that I thought
we had a pretty good chanceunder UK patent law, which
basically said the inventor isan actual divisor which is
agnostic to entity nature, andthat you have to disclose all
individual.
You know inventors who arenatural persons, and we simply
(11:11):
submitted that the designationwas no one.
We could talk more about thatcase that we lost, although the
case is now back at the highcourt.
Just a couple months ago, though, we won in Germany at the
highest civil court, whichresolved a circus split between
two appellate German courts onethat held these inventions are
inherently unpatentable, onethat said they are patentable.
(11:32):
And the German Supreme Courtsaid all you have to do under
German law is have a personcause an invention to be made,
and that is sufficient forinventorship.
So, on our facts, we had that.
So you now have a lack ofharmonization about this,
although inventorship law isalready pretty disharmonized.
In Austria and Israel, forexample, you don't have to
(11:52):
disclose an inventor.
In Cyprus and Monaco, which areEPO member states, they've
reported EPO that they allowcorporations to be inventors,
and then in most otherjurisdictions the cases kind of
very slowly trudge on.
That was a pretty good size,ramble Lee, you look like you're
checking your email.
Speaker 2 (12:17):
No, not me, Not guilty I was.
I was actually just droppingGwilym a little message in the
chat.
It's been a roller coaster,that's for sure.
When we last spoke was beforethe Supreme Court, a long time
before the Supreme Court, wasn'tit?
And a lot of things are more upin the air, but I think around
then perhaps there were somereversals going on.
It's interesting because, Ihave to say, after the Supreme
Court decision, I'm afraid tosay Ryan, I thought that the
(12:39):
story might be coming to an end,but it clearly isn't.
That's great news.
Still, it's great news becauseit's creating huge debate about
something we're going to have toface up to in due course yeah,
that was, that was a big.
Speaker 1 (12:51):
You know that was one of the three reasons we did
this, you know, because thiswasn't really on people's radars
or if it was, they weren'ttalking about it.
You know the law historicallykind of lags behind technology
to the point where you haveproblems about this, as for
example, right now with, youknow, copyright training issues,
with machine learning, which isjust a hot mess in different
(13:14):
places, especially the UnitedStates, and I would submit more
or less a disaster, because noone seems to have seriously
thought about this, you know,before it came up.
So you know, part of the goalwas that you not have pharma
companies losing patents leftand right, at least in the
United States, over this in adecade and that instead you at
least have guidance forstakeholders, and I think the
(13:35):
product has been very successfulin doing that.
It's been named explicitly inall these consultations that AI
and IP offices are doing andpeople have different normative
views or statutory views on whatthe solution should be, but it
is, I think, for the best thatwe're not thinking about it for
the first time, once we havekind of widely available
inventive AI generating sociallyuseful stuff.
Speaker 4 (14:00):
So you mentioned there, Ryan, about the way that
technology runs ahead of the law.
Is there a danger that themarket ultimately decides and
actually this renders the wholesystem pointless?
Speaker 1 (14:16):
Well, that sounds very European in the framing of
it.
Right, the idea that the marketwould decide would horrify our
EU colleagues who just passedthe EU AI Act, whereas, you know
, the American colleagues tendto think that's probably the
right solution.
I mean, there are, to be sure,a lot of regulatory forces on
how AI is used to develop.
You know the market, you knowsoft regulation, you know EU
(14:40):
regulations, us regulations,technological constraints on
things.
You know all of these have animportant role to play.
You know, and you know IP lawgenerally is a government
intervention on a market that wedon't believe functions well
enough without a governmentmonopoly, a state sanction
monopoly.
Right, we think we won't haveenough of the right sort of
(15:04):
activity if we don't have patentlaw and therefore we have
patents which are, you know, bytheir very nature as
anti-competitive, as cool as youcan get.
So you know again.
You know, and then this kind ofgoes back to what do we want
for a system?
What do we think the benefitsand the costs of having patents
are, which aren't necessarilythe same for every jurisdiction.
(15:25):
Some jurisdictions, like the UK, are big IP exporters.
Some are big IP importers.
Ai is maybe changing this ininteresting sorts of ways.
You know, on the copyrightfront, I realize the audience
probably doesn't care aboutcopyright so much, but it is
very interesting to leverage forthese discussions.
You know people werehistorically to leverage for
these discussions.
(15:45):
You know people werehistorically when I was writing
about this, before thewidespread use of LLMs like chat
, gtp a couple of years ago,when, you know, ai making art
was something that generallyacademics and a few artists were
interested in, you know.
But academics thought, oh, thisis gonna be awful when you have
, you know, google Gemini.
Google is going to own all theart in the world, right?
Whereas instead what hashappened is, you know, they've
(16:06):
been platform companies that are, if anything, democratizing the
ability to be creative in a waythat, you know, people other
than music and movie studiosdidn't have before.
And so now I can ask GoogleGemini to make a creative image,
and it will, and it will be apretty good one.
Speaker 2 (16:20):
Now I usually find right, and so this is allowing
people in jurisdictions thataren't the uk to leverage these
technologies in ways that peopleyou know I don't think expected
would happen.
A long answer, lee.
I'm going to dig out a littlebit of that, though it's
actually quite broad.
A picture thing is that I thinkyou're saying in there
(16:42):
somewhere that you can't justlet the market decide some stuff
, because it makes some reallybad decisions, and that's when
the government steps in, and sothis might be one of the areas
where the market now might notlike it, but the market needs to
be looking a bit furtherforward kind of thing.
Speaker 1 (16:57):
Yeah, I mean, I think that's kind of generally a
defense of regulatory activity,and different regulators take
different views on whether amarket-based or a state-based
framework tends to be moreappropriate, entirely a
(17:23):
state-based mechanism, nowsubject to global norms through
trips, among other things.
But you know, and so you know,companies have now, in the past
five years, tended to takepositions on protection of AI
generated content and areactively lobbying lawmakers
around the world for theirpreferred solution, and, by and
large, most of the attention hasfocused on copyright, and I
(17:44):
think that this is because it'snow easy for everyone to see AI
generating creative stuff.
It's also a lot easier to makesomething creative than
something inventive and kind ofAI in the creative industry has
blossomed as a giant socialconcern.
The creative industry hasblossomed as a giant social
concern, for example, amotivating factor behind the
(18:05):
writer strikes of last year.
So a lot of the lobbying is onthings that are very pressing
right now, like training deepfakes.
To a lesser extent, ownershipof AI-generated content.
The patent stuff remains, Ithink, niche for industrial
engineering and life sciences,where, you know, many companies
claim to be making AI generateddrugs and there's a vision that
(18:28):
AI is going to totally changethe way drugs are discovered and
used.
You know that vision hasn'tcome to fruition yet, but people
are putting large sums of moneyinto it yet, but people are
putting large sums of money intoit.
Speaker 2 (18:42):
Just closing off, the Davos story you mentioned.
It's reappeared in the HighCourt in the UK.
What's?
Speaker 1 (18:48):
happening.
Well, a couple of interestingthings are happening.
I mean, I refuse to give up asa matter of course.
Generally right, so that's partof it.
Most people would just sort oftake the Supreme Court case and
be done with it.
But no, we don't roll that way.
You know, a couple of it shouldbe done, but at this, you know,
at the initial hearing,officers said look, uk patent
(19:12):
law doesn't protect us.
It's all very interesting, butgiven what you told us about how
the invention was made, I can'tsee how you could get a patent
from us.
Right, and and a fair positionfor them to take.
But then, before the SupremeCourt, they said we could care
less.
What's on this form, as long asit appears to be a human being,
we won't look behind it.
And if you'd only put DrThaler's name whether it was
(19:35):
because he used or owned the AIor he didn't really have to have
any involvement in it, thatwould have been fine for us.
So that was a submission theymade in oral arguments.
And so we amended Form 7 andsaid, ok, well, if that's your
position, here's the new form.
And they said, well, not inlight of the new Supreme Court
jurisprudence.
So we are.
You know our case deliberatelytook the factual position that
(20:01):
you know the AI was the actualadvisor of the invention, as it
did the thing that wouldtraditionally make a human being
an inventor.
Now, what that thing is Avaries by jurisdiction quite
substantially.
So the example in Germany itcan be I pushed a button.
In the US it has to have been asignificant contribution to a
(20:23):
claim, has to be conceptionversus reduction to practice In
the UK.
What does it mean to be anactual advisor exactly, and what
does a person have to do versusa machine, and where is the
line at which something becomesunpatentable?
Now, I think that will kind ofbe the next subject of cases,
Although it is probably,although patent attorneys take
(20:46):
different views on this.
Well, it is certainly lessimportant in the UK because
there's limited grounds tochallenge inventorship or
entitlement.
You have to be yourself anentitled party and if you don't,
the general view seems to bethat if I put my name on this
thing and then sued you forinfringing it, you couldn't
challenge that I wasn'tappropriately entitled to it.
(21:08):
But some attorneys do thinkthere is some equitable remedy
out there somewhere for this.
In the US it is somethingsomeone can challenge in an
infringement proceeding at anypoint and it can render a patent
invalid or unenforceable.
And so, you know, this iscommonly tested in litigation,
you know, and when you depose apharmaceutical researcher who's
(21:31):
a listed inventor and she says,well, I use this platform from
this company and no, I don'tknow how it works, but I asked
it to generate a new drug and itdid, and it said it made a
million of them in this, youknow, had all kind of the
traditional criteria we look for, and I don't know how it
predicted that.
And you know, it just told methis was the best drug.
So I gave it to our patentattorneys and they filed it.
You know you would now losethis patent in the United States
(21:54):
, but anyway.
So back to the high court.
So that's thing one.
So we are, you know, trying toexplore where these lines lay
and arguing.
You know, fundamentally, if theUK IPO wants to take the
position that it doesn't care,which is whether the
applications were, whether adivisional filed while the
(22:23):
applications were live but afterthey were eventually deemed
withdrawn, is a valid divisionalbased on the timing of that.
Speaker 2 (22:33):
Let's not do that one today.
Can I go back to your pointabout the UK in particular that
we're not that bothered by somementorship in the sense that, as
you said, there's limitedopportunities to challenge it?
No-transcript.
(23:01):
My understanding of his part ofhis point was that, so you know
, no one's challenging it.
You said it was who it was.
You believed that everything'sokay.
Um, it was I.
I see seeper liked thatargument.
We actually put a submission tothe supreme court, as you know,
kind of supporting that, thatline of thinking.
Um, obviously it didn't comethrough.
But do you think that's thedirection that the legislation
(23:24):
should go?
It's just to keep away from theAmerican prevalence of the
invention, as it were, and tokeep it between anyone who
thinks there might be aninvention.
Leave everyone else out of it.
Speaker 1 (23:36):
DAVID BASTIAN yeah, I mean it's a good question.
It is clearly a broader pointthan one related to AI-generated
inventions.
That is, you know, and I thinkit was actually I just got back
from China the AI PPI and thatwas one of the study questions
there on harmonization ofdisclosure requirements and this
sort of thing ventureship.
It generally just matters toentitlement and very few people
(23:59):
can challenge it.
You know, Burse's point wasthat and also that the UK IPO
shouldn't be in the business ofverifying these forms.
The form was filed.
Their job is to process theform.
There was a bona fide beliefthat the form was accurate
anyway, but really none ofthat's the IPO's business.
We don't have a concept offraud on the patent office here.
(24:20):
The way that we do in thestates and I think first was
kind of looking over at americaand seeing everyone suing
everyone else about everythingand thinking we probably don't
want that here in the uk.
Um, you know that that'sprobably right.
Um, I think it probably isright.
(24:40):
The other components of thiskind of issue and question being
you know, what rule do we wantto have for this sort of thing?
And the UK is kind ofinteresting, just again to bring
copyright into it.
They were the firstjurisdiction anywhere in 1988 to
explicitly change copyright lawto say, when you have a
(25:02):
traditionally authorless work,we're going to provide
protection for computergenerated work under the special
scheme, and I did think thatthat was particularly forward
looking and future proofing.
And granted, people in the 80sapparently thought this needed
urgent answering and they were35 years ahead of their time.
But but, on the ai generatedissue, it would be nice for
(25:23):
parliament to kind of step inand say this is what we want and
we want it for good policyreasons and here's the law on it
.
But I do think also that youknow the patent act as it is did
have an answer in that,particularly if you looked at it
purposefully rather than kindof a a literalist, uh approach
to the question so you this isbecoming?
Speaker 2 (25:45):
can I call this a passion?
Is this a passion?
This whole ai thing for yourown yeah.
Speaker 1 (25:51):
Yeah, it sounds a little, a little odd.
If I'm passionate about ai,that sounds like I have an ai
girlfriend.
But uh, I don't.
Speaker 2 (26:03):
I'm fishing for that.
That's good to know.
Speaker 1 (26:08):
I was going to talk about.
Are we still plugging your book?
Oh, please, I have a couple ofnew books but yeah, plug the
Reasonable Robot.
Speaker 2 (26:13):
The Reasonable Robot is a really, really good read
everybody listening but also Ithink it's one of the early kind
of full discussions about whathappens when we start accepting
ai as almost as a person, andobviously the book was quite
took it beyond ip significantlyinto.
I know, definitely carinsurance was an interesting
topic and various other thingsas well, but really interesting.
(26:34):
But it feels like that debatehas become more mainstream now.
There are more discussionsabout what if, what if?
I saw in the US there's one atthe moment.
The USPTO has got aconsultation out about AI and
the risk of AI creating loads ofprior art.
Is this familiar?
Speaker 1 (26:56):
Yes, it is.
Yes, it is so the well, youpicked up two veins there, but I
will remember to something forplugging the reasonable robot
and historically you've had acopy of it to wave around or hit
Lee with.
But I'm sure it's theresomewhere, it is not?
Yeah, that clearly is it Right?
(27:17):
So you know, ip agencies andregulators have been thinking
about AI and IP for a long timeand again, it's always been kind
of very academic, largely, youknow the UK being a notable
example of a different event incopyright.
But the USPTO kind of set a newwave of this off in 2019, where
they released two requests forcomments, and the summary of
(27:39):
which was not much to see here,and then kind of a lot of other
people got in the mix, like UKIPO.
You know they've recently donea 180 on this, as I mentioned,
and now they're thinking of allthe other interesting ways that
AI is going to change IP law anddigging into a couple of these,
and there really are a lot ofthem you know with how.
You know the book broadly dealtwith the phenomenon of AI
(27:59):
behaving like a person and howthe law deals with that.
You know whether it's inventingor creating something or
setting standards for humanbehavior, like reasonableness,
including in the self-drivingcar field, you know.
But we have kind ofreasonableness standards in
patent law with the personhaving ordinary skill in the art
.
And as you have an AI thatstarts automating routine R&D
(28:20):
activities, you know how doesthat change standards?
You either have an AI representthe person having ordinary
skill in the art or the personhaving ordinary skill in the art
is a human researcher using agenerative AI system.
And you know semantically thoseare different, but functionally
not so much.
You know other.
You know questions include youcan have an AI make a huge
(28:42):
amount of stuff.
You know is that prior art andis that a problem?
And you know in the US you knowprior art has to be enabled.
So you know, for you know youmay have.
We've had a company, forexample, all prior art that's
been working on having an AImake every conceivable sentence
(29:03):
for many years.
You know a claim has to be onesentence in the US.
So you know, if in a databaseof a quadrillion sentences you
have an identical, you knowpassage, you know or claim, is
that prior art?
Is it enabled?
Is it accessible?
(29:23):
You know, one of the reallyinteresting areas that I don't
work in so much but is AI makingdesign patents or registered
designs?
Because AI is really good atspitting out variations on
industrial designs and thoseapplications are just drawings.
So, you know, is AI preemptingwhole fields of activity and is
that a good or a bad thing?
(29:43):
And do we want lawmakers tocarve out that sort of stuff?
My own view on this was, youknow, you don't want a two-track
system for AI and people.
You know, if you have AI makinguseful prior art, that is
something that should beconsidered prior art.
It would be very difficult todeal with otherwise, all the
incentives in the right place,but that if you do have this
(30:05):
gigantic database, no one couldpossibly use and has no value
and just has sentences thatisn't enabled prior art yeah,
it's also weird.
Speaker 2 (30:14):
Why would somebody use ai to create loads of prior
art?
It's just an old concept.
They're gonna hope to monetizeit, or something.
Speaker 1 (30:22):
Well, for a few things I mean.
One is people do deliberatelypoison the well as a research
strategy.
So if you aren't planning topatent in a space that you're
active in, you might want torelease prior art to keep other
people from patenting in thespace.
If you don't want to, you know,enforce patents.
You know some people you knowhave a view of patents as
(30:46):
inherently bad and kind ofacademically want to blanket
spaces and publications toprevent, you know, monopolies
growing up in the space.
Those are kind of two of thebig reasons, I think.
Speaker 2 (31:00):
Yeah, it just interests me that it never
occurred to me that it would besomething that needed
legislation but clearly can Ichuck one in that.
Speaker 4 (31:07):
So you know that I usually ask the daft question,
so let me ask a daft question.
Um, so we've been doing quite alot of work both on the podcast
and in conversations around, uh, non-practicing entities.
So your patent troll typearrangement If AI gets
(31:28):
widespread inventorship, couldyou foresee a future where AI is
used in that kind of aggressiveway to just flood the market
with patents just for thepurposes of litigation?
Speaker 1 (31:37):
Sure, although there's a lot to unpack in that
quote, unquote, daft questionand I don't think it was
terribly daft and I have beensticking pretty patented 30
years.
But I assume that is yourtarget market here and we're not
too much hitting up the sixthform listeners.
I do have a daughter in sixthform who's not listening to this
or I'm sure will not listen tothis and Well, so there's a kind
(32:03):
of different views on whethernon-practicing entities or you
know the derogative term, patenttrolls, are a good or a bad
thing for the patent system andwhether you know right.
So there's all that too Right.
But you know, yes, this gets tokind of the concern which I kind
of mentioned in the copyrightspace.
But let's say you get.
You know, ai isn't that greatyet, but we've seen how very
(32:24):
quickly it improves in creativestuff.
Let's say, in five or 10 yearsyou really do have an AI where
you can say you know, I'd like adrug to treat this or optimize
this design, and it just does it.
And it does it extremely welland it is really prolific at
doing that sort of thing.
And, yes, you would be able toget lots and lots of patents on
(32:47):
that activity.
Now I tend to think that's agood thing because you know,
let's say you were able to takea sample of someone's cancer and
run it through an AI and the AIcan make a personalized
antibody to treat your cancer.
Right, you know, you couldmonopolize the field for
(33:10):
treating cancer.
On the other hand, that meanswe've now got the cure for
cancer.
So if AI is spitting out allthis really useful stuff at
volume, that just means we'regetting what the system was
intended to do socially valuablestuff.
And you know, there is a 20year kind of limit on how much
protection someone can have, andafter that, then we all have
free access to the cure forcancer in a way that we didn't
previously.
(33:31):
You know someone could, you know, do some you know bad stuff
with that?
If you had the cure for cancer,you know, only let someone have
it in return for a billiondollars, you know, per person.
You know are safeguards builtinto patent law, like compulsory
licenses to deal with that sortof thing.
But also, again, what we saw inthe copyright was a little bit
(33:52):
less.
One company is going to controlthe stuff and more platform
companies are making thisavailable so that you could
potentially use AI to makesomething inventive.
But we will see how it washesout.
But yes, as you say, there is aconcern someone could try and
really dominate certainscientific areas with these
technologies.
Speaker 2 (34:13):
RAOUL PAL.
I'm going to just touch onsomething that has been one of
the big themes Is this as yousay, ai just treats it like it's
a person.
But also AI is the skilledperson which begins to introduce
a whole bunch of interestingpatenty things.
For the non-patent specialist,the skilled person is the
nominal, non-inventive butknowledgeable person who never
(34:37):
has an inventive thought, whoknows how to follow the
instructions in a patentdocument to make it work, and
various other kind of tests allsit around this nominal skilled
person.
And, ryan, I know you had agood think about what it means.
If AI becomes a skilled personor if that becomes the test,
then that causes all kinds ofproblems later on because
(34:58):
everyone else is basically not agenius.
An example, going back to thepharmaceuticals that.
So if you suddenly say, well,just because ai did it, um,
there's no inventive in in there, um, it just messes, messes
things up, um, is this debatestill going on about that
conflation?
Speaker 1 (35:17):
oh yes, I, I will remark.
You seem to look right up at mewhen you said a person who's
ever had an inventive thought.
But but Well, messes it up orfixes things Right.
So you know we have this bar inpatent law, which is we don't
want to give a patent forsomething that you know an
average researcher would havecome up in the ordinary course
(35:39):
of things with, because you know, economically speaking, we
didn't really need thatincentive for that invention to
come about.
Right, we only want to rewardextraordinary effort.
You know something that we thinkwouldn't have been invented but
for a patent incentive.
You know, or at least a patentincentive significantly sped up
the introduction of thisinvention to society in a
(36:00):
trade-off we find worthwhile.
Now let's fast forward 10 or 20years and assume that we have
AI that is really vastly morecapable than human beings in
certain areas like drugdiscovery.
Like you have a disease and youjust ask an AI to make a
treatment for it and it does,that then becomes routine
(36:21):
activity that you don't need apatent to incentivize, and so it
probably isn't something thatwe want to grant the patent for.
You know, it's something thatreally anyone could have done,
or any average researcher.
Now, in this hypotheticalfuture in which AI is
outperforming people, you know,yes, it becomes very difficult
for an unaided natural person tomake an inventive thing, you
(36:45):
know.
But again, otherwise, youbasically have a protectionist
system where we are rewardingcertain kinds of human effort,
even though we could have had anAI do it very cheaply and
quickly, just to encouragesomeone to be an inventor, and I
don't think protectionism hasreally ever been the route to go
with dealing with technology.
I don't remember if I used thismetaphor in the last podcast,
(37:07):
so I'll use it again.
Well, not metaphor, but example.
I was in Brazil some years agoand they had elevator attendance
in all of the commercialbuildings and I asked why that
was and they said, well, it wasa union thing to give people
jobs.
And that did give people jobs,but it wasn't a very productive
use of their labor.
We could have had elevatorpanels and they could have been
out doing something moreproductive.
(37:28):
And you know, as people arestruggling with this very much
so in the creative industryright now, I don't think the
solution is like you havespecial incentives for people or
you ban the use of AI.
You know AI will replacecertain jobs.
It will make other jobs.
You know we should havepolicies that help people who've
(37:55):
been rendered technologicallyunemployed with more benefits
and retraining, you know, butthat is, have an unlimited
amount of wealth and leisure asa society and the goal there is
to just keep all of the benefitsfrom flowing to Elon Musk and
Sam Altman and making sure thatwe get some of them.
You know, speaking of your kindof future question about you
(38:16):
know, ai doing things in thefuture, sam Altman and Elon Musk
say a lot of things, but theyalso think we're going to have
AI that's broadly at theintellectual capabilities of a
human being in a couple of yearsand if they're right about that
, that really is a prettydisruptive social concept, but
also for patent law.
Speaker 4 (38:37):
So, Ryan, one of my jobs on the podcast is to kind
of keep an eye on the clock andyou'll either be delighted or
terribly frustrated.
They will be over 45 minutesnow.
I've got at least one otherquestion.
I don't know whether Gwilym'stotally exhausted.
Are you exhausted or can I?
Should I, because, because mineis a bit of like taking us
towards the end question, Idon't want to take a stay if
(38:57):
you've got more to to ask Gwilym.
Speaker 1 (38:59):
I was going to ask that question about plumbing,
but I guess we do that after thepodcast at the pub.
Speaker 2 (39:05):
No, go for it, Lee.
Speaker 4 (39:07):
So you were talking kind of like end game there in
terms of human level artificialintelligence and what that might
mean generally as like a socialconstruct.
But what's end game for you,ryan?
How long do you keep this going?
Is this end game that thereends up being some kind of
international harmonizationaround AI?
As an inventor, when'd you stepoff the roundabout?
Speaker 1 (39:31):
Yeah, no, I think I'm on the roundabout for a good
while, which is nice because itmeans I don't have to think up
anything new.
This is really the case thatjust keeps on paying dividends
and I just keep writing the samelaw review over and over and
giving the same talk over andover, and this is really sort of
the academic stream.
But when I first spoke toreporters about this in 2019,
when everyone thought this was aridiculous publicity stunt and
(39:53):
you know I no longer get thattoo much you know the reporters
are like so when will this allbe done?
Six weeks, and I'm like youknow, probably at least 10 years
and they seemed unable toaccept that as an answer.
But here we are, five years in,and I think we started kind of
the work in 2018.
So we're at least halfwaythrough.
(40:14):
I mean, look, jurisdictions haveto decide for themselves what
they want their rules to be.
I think this case law hasreally helped stakeholders.
It's promoted discussion.
It's advanced one view of howthings might want to be.
It goes down from courts topolicymakers to kind of have
public debates and discussionsabout this and then ultimately,
(40:35):
yes, the ideal would be that youdo want some sort of
international harmonization,because it shouldn't matter if
you're using AI to make aninventive thing in the uk or the
us or china or you knowwherever you are.
Uh, an invention's an invention, and so you know, hopefully at
some point in the next 30 yearswe'll get something like that,
and in the meantime I get tokeep appearing on podcasts 30
(40:59):
years, guillem?
Speaker 4 (41:00):
have you got another 30 years in you mate?
Speaker 2 (41:02):
oh, yeah, yeah, yeah, I'm nothing stopping me.
Speaker 4 (41:06):
I'll be nearly 90.
No, that's kind of nothappening.
I won't be here, so I thinkwe're done or thereabouts.
Except I know that Gwilym'ssitting on a question that
you're desperate to ask, aren'tyou?
Speaker 2 (41:17):
I am.
I've just realised I close huh.
I'm going to have a closingquestion.
Ask Lily first, then you.
Ryan, I've got one today, butI'm going to be surprised if
it's the same one we did lasttime.
It doesn't matter, because thatwas 2020.
Speaker 4 (41:31):
Is it?
Speaker 2 (41:32):
I can't remember what we did last time, so my answer
won't be consistent, but it'sokay, because then we can
basically compare our answers,which are probably going to be
different.
It's very simple.
It's always 42 42.
(41:52):
By the way, lee, that's deepthought.
Now that's an ai, that was um.
So, first of all, um, we'vetalked lots and lots about ai
becoming, you know, treated likea person and replacing the
skilled person and all theseother driving cars.
We've talked about that.
But you read the reasonablerobots very interesting, um.
But um, oh yeah, what yourother?
What are your other books, ryan?
That's what I wanted to askbefore I get to my closer, sorry
.
Speaker 1 (42:10):
Oh, the Research Handbook on AI and IP by Edward
Elgar, and the InternationalIntellectual Property in a
World-Integrated Economy byAspen Publishers.
So those books are even moreriveting than the Reasonable
Robot, one of which is aresearch handbook and one of
which is a textbook.
But I'm now working on anencyclopedia of AI and law, so
(42:33):
that's going to be the most pageturning yet.
Speaker 2 (42:38):
Typical name the Reasonable Robot.
You've got to read that, anyway.
Closing question AI.
Speaker 4 (42:51):
Lee, what job that you do would you most like AI to
do instead?
Oh, I guess the obvious answeris none, because I don't like
working for a living.
That's not true.
I quite like getting paid,which is slightly different,
isn't it?
You don't like money, do you?
I love my job.
I love working with people likeyou, gwilym, so it's probably
something I don't know whetherAI would be able to do that I'm
not the world's greatestnetworker.
I hate being stuck in a roomwith a load of people I don't
(43:13):
know and being forced to makekind of polite conversation,
perhaps with a view to maybeforging one or two connections
that might be useful at somepoint in the future.
Can't be doing with it, so ifAI could somehow come in and do
all of that networking for meand just give me a bunch of
perfectly formed connections,that would suit me in the future
.
Job done.
Speaker 2 (43:34):
I believe that's referred to as neural networking
.
Thank you.
Ryan, did you set that one up,will?
No?
No, I'm just here, I'm all week.
Uh, you, you're on fire.
So it's what, ryan, what jobthat you do would you?
Would you prefer ai did?
Speaker 1 (43:54):
but before I give you an answer, and it's an easy one
, lead have you guysexperimented at all with the
google?
You know, notebook LM.
Speaker 4 (44:02):
Not, no, what, no, what no.
Speaker 1 (44:04):
Yeah well, so you know, take a paper or something
and you drop it in a folder anda Google podcast emerges from
this of two kind of emotionallyintelligent sounding, human
sounding voices debating thesubject of this thing.
It's basically an automaticpodcast generator and it takes
about two minutes to make one,but no one would listen to that
(44:27):
and replace you two at this job.
I'm sure Me grading papers Imean my students use LLMs to
write the papers.
I should be able to use LLMs tograde the papers, at which
point one kind of wonders aboutthe purpose of the whole system.
But anyway, I I don't likegrading lee, I'll jump in.
Speaker 4 (44:48):
I know you're going to ask me my answer I was going
to come back on the gradingthing first, because I used to.
I I was I was a teacher infurther and higher education and
I always found it easy just tostand at the top of the stairs
and lob papers down the stairs,and the further they flew, the
heavier they were, so the higherthe grade yeah, I mean.
Speaker 1 (45:04):
Well, to be fair, that is more or less what I do.
But it would be even easier tojust upload it all to an ai and
have it put the numbers in, oror to simulate the stair
experiment I'm sure we're reallyinfusing your students as we
speak, so um they're notlistening, don't worry perfect.
Speaker 4 (45:21):
Yeah, no one does grill them.
Come on then.
What's yours?
Speaker 2 (45:25):
I'm gonna be festive christmas shopping.
I do it all online anyway, justwant ai to go and do it.
Speaker 4 (45:30):
It's just innovating, it's innovating do you find
that more and more people areusing these blooming online list
things for christmas shopping?
We we do it now with likesecret santa at work.
Now we do it in our family,where you kind of upload a wish
list and someone will buy you agift.
What's the point?
What's the point in me tellingpeople things that they can buy
(45:52):
me when I can just buy themmyself?
Speaker 2 (45:54):
that's a good point, although my family approach is
that if you ask for somethingyou you can't have it, so you
have to ask for something youdon't want and then then you
won't get it and you'll get whatyou might want.
It's quite complicated.
Speaker 4 (46:06):
Oh, some subtle kind of psychology going on.
I've not really thought aboutthat.
Speaker 2 (46:09):
It's quite messy.
Speaker 1 (46:10):
Well, it sounds better.
I'm supposed to guess whatpeople want, and if I guess
wrong I get punished.
Speaker 4 (46:18):
So I prefer your list of urges.
I don't have enough time in mylife to sit down and build an
Amazon gift list.
It's just not going to happen.
Speaker 2 (46:25):
You're the man who owns everything.
Anyway, let's face it.
Speaker 1 (46:27):
Well there you go.
Speaker 4 (46:29):
Yeah, yeah Well, ryan , thank you for coming back and
gracing the podcast.
Always a pleasure to have youon.
Gwilym, thank you for well justbeing the consummate co-host,
as you always are, and thanks toall of our listeners who are
now going to rush off and leaveus a little review on the
podcasting platform of theirchoice so that other people can
find us.
Maybe that's something that aicould do for us.
(46:49):
Perhaps we could get some kindof ai crawly bot thing, or
whatever they're called, tocrawl all over the podcasting
world and leave lots of reviewsabout two ips.
Thanks both cheers, cheers.
We'll see you next time.
Lee Davis and Gwilym Roberts are the two IPs in a pod
and you are listening to apodcast on intellectual property
brought to you by the CharteredInstitute of Patent Attorneys.
Speaker 4 (00:20):
Hey Gwilym, how are you?
This fine I say fine day it wassnowing this morning.
I've just walked miles, I'mdamp In the snow or just in the
rain.
Speaker 2 (00:29):
No, it's just it doesn't.
It's warmed up, so it's wet.
It's just wet snow aka rain.
Speaker 4 (00:35):
Yeah, no, I came into Waterloo this morning so
excited because there werereally really thick flurries of
snow, and then you lose thatexcitement, don't you?
When it just turns to mush onyour coat and you yeah, yeah,
yeah.
And I thought about jumping ona bus.
I jumped on a bus last week andit took a ridiculous amount of
time to get out of the office,but I thought I'd be a little
bit drier and warmer.
But the queue was horrificallylong and I thought why are all
(00:55):
those people standing there toget on a bus that's going to
take them much longer to get towhere they want to go than it
would walking.
And they get, and they'regetting wet, and they're getting
wet anyway.
That's especially a consultantin yeah do you want a little bit
, little bit news on the seeperfront?
Yes, so do you know?
We've been supported ably overthe last couple of years by ria,
(01:16):
who always lurks in thebackground on these calls and
does the recording for us anddoes a big chunk of the editing
and all of those kinds of things.
Yes, this is her last week.
This is her last podcast no, itcannot be, she's on.
She's off to past his news.
He's off to another membershipassociation.
She's leaving me grill him.
Speaker 2 (01:34):
She's leaving me, she's listening, ria, we know
you're there we know what youdid I don't.
Speaker 4 (01:40):
I don't think we're going to be able to coax her on,
not at the start anyway.
Maybe we can get her on towardsthe end.
But I think we should say a big, big thank you from the podcast
.
I mean, we'll obviously saythank you for all of the work
that Ria's done for SIPA overthe years, but I think we should
just say a big, big thank youfrom the podcast for Ria for
supporting us.
Speaker 2 (01:56):
Actually, yeah, we should Absolutely.
And to say thank you to, weshould say thank you to oh, the
people in the background.
Speaker 4 (02:02):
When neil neil comes on occasionally doesn't know
what sub for me or you if we'renot around.
Yeah, we've got emily andopinia and fran, who, fran, of
course, started off the podcast,really, in terms of working out
the tech for us and doing theediting.
But yeah, emily, emily andopinia and fran in the
background doing all of theediting and making a sound like
we know what we're talking about, which is no mean feat no mean
(02:23):
feat, they're all great.
Speaker 2 (02:24):
Thank you everybody.
Speaker 4 (02:25):
Yeah, no thanks to everyone really, yeah, but yeah,
sorry to see maria go, but uh,hopefully, hopefully she'll find
an amazing new role that she'smoving into, so that's cool she
doesn't get to do a podcast no Ibet she doesn't.
Yeah, I bet there's no talent.
There be no.
So, speaking of speaking oftalent, should we get the guest
on?
Yeah, so, occasionally.
Occasionally, we have returningguests, which is for a couple
(02:54):
of reasons.
Is it so?
It'll either be because theywere so amazing last time that
we feel that we need to revisitthem and the subject, or that
we've just forgotten we've hadthem on before.
So, um, and, and I don't knowwhich, it is with Ryan.
Ryan, who are you?
Tell us a little bit aboutyourself.
Speaker 1 (03:09):
This is after the two of you were just telling me how
amazing I looked and was, andGwilym named his recent child
after me, so I assumed it wasamazingness and I was here for
the kind of final real podcast.
So I took that as a great markof honor.
But you know, it could also bethat we just all have short
memories.
Speaker 4 (03:30):
It was, it was a long time ago that we were last
together on the podcast.
It was May of 21, which waskind of deep, deep, dark
pandemic days.
Speaker 1 (03:39):
It all blurs together .
Speaker 4 (03:41):
Yeah, I had a little listen back earlier and we
certainly started off talkingabout the fact that we may never
see the real world again andwhere we confined to our houses
for for the duration turns out acouple of years, and we were
good yeah, now you can go backto getting wet waiting in a
queue yeah, yeah yeah, podcastsare a really good historical
document for that, actually,because that people just they're
(04:01):
recorded conversations aboutwhat was happening.
Speaker 2 (04:03):
I listened to a whole series of a rudely named one, a
huge long series, and itstarted off in the depths of
lockdown and it was reallyinteresting listening back to
some of the conversations.
Speaker 4 (04:14):
It's a good historical document.
I think so Many started then,didn't they?
Yeah?
Anyway, ryan, can you introduceyourself for the listeners,
because we know who you are?
They don't, unless they listento the last podcast, which is
unlikely.
Speaker 1 (04:27):
Yes, Well, I'm Ryan Abbott.
I'm a professor of law andhealth sciences at the
University of Surrey School ofLaw and a attorney partner at
Brown Harry Smith Conn in theUnited States, and a mediator
and arbitrator with JAMS.
And I assume that I'm here formy infamy associated with the
Thaler or Davis cases, or justbecause Willem thinks I'm very
(04:48):
handsome.
Speaker 2 (04:49):
It'd be great, but you also know that your
knowledge on all that stuff isalso super interesting and I
think we'll hear some updates,things happening, what's going
on All that?
Speaker 4 (04:59):
Yeah, so do you want to give us a potted history of
the project, and I know we didit in depth in the last podcast.
But let's start off with atleast reminding people about the
project, and I know since welast spoke it's been tested in a
number of jurisdictions withdifferent outcomes.
Speaker 1 (05:14):
Indeed, indeed.
Well, I'll give the slightlylonger rambling version, and you
can interrupt me when itbecomes even less interesting.
But we go back about 10 yearsin time and I was working as a
baby attorney in the lifesciences and noticing in drug
discovery that people seem to bedoing some interesting things
with AI and some vendors kind ofthe forerunners to AI-generated
(05:38):
drug discovery, whatever thatis were advertising that they
could do things like if we gavethem a therapeutic target,
they'd have a computer, gothrough a big antibody library
and pick the antibody out thatwould be best for us and this
could be our lead drug candidate.
This is a company I was workingfor and you know we could file
a patent on that and that's thefoundation of a new drug
(05:59):
portfolio.
And I thought well, that'sinteresting because it sounds
like an AI is automating.
You know at least some of whattraditionally makes a human
being an inventor, and whatwould happen if you had an AI
invent something?
And it turned out people hadbeen thinking about this for a
long time.
People have been writing aboutit since at least the 60s and
academically saying you know,because at the time only
(06:21):
academics took an interest inthis, you know you really
wouldn't need a patent becausean AI doesn't care about getting
a patent and patents are toincentivize inventors.
And I thought, well, that's notquite right because really all
of these inventors assign theirpatents to the pharmaceutical
company anyway.
It's the companies that needthe patents to hire the
inventors and to invest in doingclinical trials and to disclose
(06:43):
these things inventors, and toinvest in doing clinical trials
and to disclose these things.
So you'd need a patent just asmuch.
And why wouldn't you want acompany using AI and R&D if you
could get it to do a better jobthan a person at something which
it wasn't really doing,although it's all very
complicated?
But I wrote some law reviewarticles on it and for the first
time, people took an interestin things I was writing and in a
couple of years it had gonefrom.
(07:04):
This is vaguely academicallyinteresting to you know, we're
kind of concerned about this.
Could we pay you a bunch ofmoney to help us think it
through?
And at some point one of mycolleagues said well, why, if
this is an issue, aren't thereany cases on it?
And I said, well, you'd have toreally proactively disclose.
You had an AI do this.
There's a huge incentive not toand you know otherwise, only in
(07:26):
litigation in the US, butthat's probably going to take a
long time.
So I, you know, assembled agroup of patent attorneys to
file a couple of test cases forAI-generated inventions.
I spoke to Dr Thaler, who Iinterviewed for the academic
work, and said you know, wouldyou be interested in having your
AI system generate a couple ofinventions or an invention?
And he and others had done thispreviously, again kind of this
(07:50):
proof of concept experiments.
So he had an AI generate twoinventions.
While he was at it we filedapplications on those of the UK
IPO just to, you know, help theBritish patent attorney milieu.
We did this because they did asubstantive exam before looking
at inventorship.
The patents both passedpreliminary substantive exam.
(08:11):
We filed a form seven and a PCTapplication and said well,
actually these were invented byan AI and the person who owns
them is the owner of the AI, onthe principle that you own
property made by your property.
And then we filed it in 17places and some interesting
things happened.
I could continue to ramble.
Speaker 4 (08:34):
You carry on rambling .
Rambling is good at the moment.
Speaker 1 (08:37):
Ramble away.
So in 2021, although again itdoes all blur together South
Africa granted us a patent Withthe AI listed as the inventor,
which again isn't a matter ofgiving an AI rights.
Most inventors don't haverights to their patents,
although all of this is verycomplicated.
Plus, an AI is not a legalperson that couldn't own
(09:00):
property.
But it is to be transparentabout how an invention is made
and with the AI's owner as theowner of the patent.
South Africa doesn't dosubstantive examination, but
these have already beensubstantively examined and they
do do formalities examination.
So we got a patent there andlater that week the enlightened
Justice Beach and the FederalCourt of Australia issued a 40
something page reasoned decisionabout why, under the Australian
(09:22):
Patent Act, an AI could be aninventor and why at least we had
made the best showing ofentitlement to that patent.
That got overturned at an enbanc decision.
The federal circuit in the USrejected the application on the
basis that an inventor isdefined as an individual in the
patent act, which the SupremeCourt has held.
(09:42):
Usually means a natural person,we pointed out.
Usually they said well, thedictionary says an individual is
a person and we pointed outwell, the very next sentence
says an individual could be athing, so that doesn't really
work.
But there you have that.
And they said well, if you lookthrough the patent act, it
seems to assume an inventor is anatural person.
And we said well, the US patentlaw has this unique provision,
(10:06):
or provision anyway, that youcan't negate patentability on
how something was made, which iswhat they were effectively
doing.
And they said well, for thefirst time we're going to say
this applies only tonon-obviousness.
As a result of that decision,the Supreme Court didn't take
the case.
The US Patent and TrademarkOffice amended their guidance
from AI nothing to see here in2019, to a human being has to
(10:28):
make a significant contributionto each claim in an invention
where the claim is unpatentableand potentially the entire
patent.
You know, inventorship isuniquely important in the US as
a substantive thing which we cancome back to talk about.
So we have one model there.
It went up to the UK SupremeCourt, which I got to argue,
which was great fun, andAmericans just asked if I got to
(10:49):
wear a wig.
I didn't, which was unfortunate, but you'll know that I thought
we had a pretty good chanceunder UK patent law, which
basically said the inventor isan actual divisor which is
agnostic to entity nature, andthat you have to disclose all
individual.
You know inventors who arenatural persons, and we simply
(11:11):
submitted that the designationwas no one.
We could talk more about thatcase that we lost, although the
case is now back at the highcourt.
Just a couple months ago, though, we won in Germany at the
highest civil court, whichresolved a circus split between
two appellate German courts onethat held these inventions are
inherently unpatentable, onethat said they are patentable.
(11:32):
And the German Supreme Courtsaid all you have to do under
German law is have a personcause an invention to be made,
and that is sufficient forinventorship.
So, on our facts, we had that.
So you now have a lack ofharmonization about this,
although inventorship law isalready pretty disharmonized.
In Austria and Israel, forexample, you don't have to
(11:52):
disclose an inventor.
In Cyprus and Monaco, which areEPO member states, they've
reported EPO that they allowcorporations to be inventors,
and then in most otherjurisdictions the cases kind of
very slowly trudge on.
That was a pretty good size,ramble Lee, you look like you're
checking your email.
Speaker 2 (12:17):
No, not me, Not guilty I was.
I was actually just droppingGwilym a little message in the
chat.
It's been a roller coaster,that's for sure.
When we last spoke was beforethe Supreme Court, a long time
before the Supreme Court, wasn'tit?
And a lot of things are more upin the air, but I think around
then perhaps there were somereversals going on.
It's interesting because, Ihave to say, after the Supreme
Court decision, I'm afraid tosay Ryan, I thought that the
(12:39):
story might be coming to an end,but it clearly isn't.
That's great news.
Still, it's great news becauseit's creating huge debate about
something we're going to have toface up to in due course yeah,
that was, that was a big.
Speaker 1 (12:51):
You know that was one of the three reasons we did
this, you know, because thiswasn't really on people's radars
or if it was, they weren'ttalking about it.
You know the law historicallykind of lags behind technology
to the point where you haveproblems about this, as for
example, right now with, youknow, copyright training issues,
with machine learning, which isjust a hot mess in different
(13:14):
places, especially the UnitedStates, and I would submit more
or less a disaster, because noone seems to have seriously
thought about this, you know,before it came up.
So you know, part of the goalwas that you not have pharma
companies losing patents leftand right, at least in the
United States, over this in adecade and that instead you at
least have guidance forstakeholders, and I think the
(13:35):
product has been very successfulin doing that.
It's been named explicitly inall these consultations that AI
and IP offices are doing andpeople have different normative
views or statutory views on whatthe solution should be, but it
is, I think, for the best thatwe're not thinking about it for
the first time, once we havekind of widely available
inventive AI generating sociallyuseful stuff.
Speaker 4 (14:00):
So you mentioned there, Ryan, about the way that
technology runs ahead of the law.
Is there a danger that themarket ultimately decides and
actually this renders the wholesystem pointless?
Speaker 1 (14:16):
Well, that sounds very European in the framing of
it.
Right, the idea that the marketwould decide would horrify our
EU colleagues who just passedthe EU AI Act, whereas, you know
, the American colleagues tendto think that's probably the
right solution.
I mean, there are, to be sure,a lot of regulatory forces on
how AI is used to develop.
You know the market, you knowsoft regulation, you know EU
(14:40):
regulations, us regulations,technological constraints on
things.
You know all of these have animportant role to play.
You know, and you know IP lawgenerally is a government
intervention on a market that wedon't believe functions well
enough without a governmentmonopoly, a state sanction
monopoly.
Right, we think we won't haveenough of the right sort of
(15:04):
activity if we don't have patentlaw and therefore we have
patents which are, you know, bytheir very nature as
anti-competitive, as cool as youcan get.
So you know again.
You know, and then this kind ofgoes back to what do we want
for a system?
What do we think the benefitsand the costs of having patents
are, which aren't necessarilythe same for every jurisdiction.
(15:25):
Some jurisdictions, like the UK, are big IP exporters.
Some are big IP importers.
Ai is maybe changing this ininteresting sorts of ways.
You know, on the copyrightfront, I realize the audience
probably doesn't care aboutcopyright so much, but it is
very interesting to leverage forthese discussions.
You know people werehistorically to leverage for
these discussions.
(15:45):
You know people werehistorically when I was writing
about this, before thewidespread use of LLMs like chat
, gtp a couple of years ago,when, you know, ai making art
was something that generallyacademics and a few artists were
interested in, you know.
But academics thought, oh, thisis gonna be awful when you have
, you know, google Gemini.
Google is going to own all theart in the world, right?
Whereas instead what hashappened is, you know, they've
(16:06):
been platform companies that are, if anything, democratizing the
ability to be creative in a waythat, you know, people other
than music and movie studiosdidn't have before.
And so now I can ask GoogleGemini to make a creative image,
and it will, and it will be apretty good one.
Speaker 2 (16:20):
Now I usually find right, and so this is allowing
people in jurisdictions thataren't the uk to leverage these
technologies in ways that peopleyou know I don't think expected
would happen.
A long answer, lee.
I'm going to dig out a littlebit of that, though it's
actually quite broad.
A picture thing is that I thinkyou're saying in there
(16:42):
somewhere that you can't justlet the market decide some stuff
, because it makes some reallybad decisions, and that's when
the government steps in, and sothis might be one of the areas
where the market now might notlike it, but the market needs to
be looking a bit furtherforward kind of thing.
Speaker 1 (16:57):
Yeah, I mean, I think that's kind of generally a
defense of regulatory activity,and different regulators take
different views on whether amarket-based or a state-based
framework tends to be moreappropriate, entirely a
(17:23):
state-based mechanism, nowsubject to global norms through
trips, among other things.
But you know, and so you know,companies have now, in the past
five years, tended to takepositions on protection of AI
generated content and areactively lobbying lawmakers
around the world for theirpreferred solution, and, by and
large, most of the attention hasfocused on copyright, and I
(17:44):
think that this is because it'snow easy for everyone to see AI
generating creative stuff.
It's also a lot easier to makesomething creative than
something inventive and kind ofAI in the creative industry has
blossomed as a giant socialconcern.
The creative industry hasblossomed as a giant social
concern, for example, amotivating factor behind the
(18:05):
writer strikes of last year.
So a lot of the lobbying is onthings that are very pressing
right now, like training deepfakes.
To a lesser extent, ownershipof AI-generated content.
The patent stuff remains, Ithink, niche for industrial
engineering and life sciences,where, you know, many companies
claim to be making AI generateddrugs and there's a vision that
(18:28):
AI is going to totally changethe way drugs are discovered and
used.
You know that vision hasn'tcome to fruition yet, but people
are putting large sums of moneyinto it yet, but people are
putting large sums of money intoit.
Speaker 2 (18:42):
Just closing off, the Davos story you mentioned.
It's reappeared in the HighCourt in the UK.
What's?
Speaker 1 (18:48):
happening.
Well, a couple of interestingthings are happening.
I mean, I refuse to give up asa matter of course.
Generally right, so that's partof it.
Most people would just sort oftake the Supreme Court case and
be done with it.
But no, we don't roll that way.
You know, a couple of it shouldbe done, but at this, you know,
at the initial hearing,officers said look, uk patent
(19:12):
law doesn't protect us.
It's all very interesting, butgiven what you told us about how
the invention was made, I can'tsee how you could get a patent
from us.
Right, and and a fair positionfor them to take.
But then, before the SupremeCourt, they said we could care
less.
What's on this form, as long asit appears to be a human being,
we won't look behind it.
And if you'd only put DrThaler's name whether it was
(19:35):
because he used or owned the AIor he didn't really have to have
any involvement in it, thatwould have been fine for us.
So that was a submission theymade in oral arguments.
And so we amended Form 7 andsaid, ok, well, if that's your
position, here's the new form.
And they said, well, not inlight of the new Supreme Court
jurisprudence.
So we are.
You know our case deliberatelytook the factual position that
(20:01):
you know the AI was the actualadvisor of the invention, as it
did the thing that wouldtraditionally make a human being
an inventor.
Now, what that thing is Avaries by jurisdiction quite
substantially.
So the example in Germany itcan be I pushed a button.
In the US it has to have been asignificant contribution to a
(20:23):
claim, has to be conceptionversus reduction to practice In
the UK.
What does it mean to be anactual advisor exactly, and what
does a person have to do versusa machine, and where is the
line at which something becomesunpatentable?
Now, I think that will kind ofbe the next subject of cases,
Although it is probably,although patent attorneys take
(20:46):
different views on this.
Well, it is certainly lessimportant in the UK because
there's limited grounds tochallenge inventorship or
entitlement.
You have to be yourself anentitled party and if you don't,
the general view seems to bethat if I put my name on this
thing and then sued you forinfringing it, you couldn't
challenge that I wasn'tappropriately entitled to it.
(21:08):
But some attorneys do thinkthere is some equitable remedy
out there somewhere for this.
In the US it is somethingsomeone can challenge in an
infringement proceeding at anypoint and it can render a patent
invalid or unenforceable.
And so, you know, this iscommonly tested in litigation,
you know, and when you depose apharmaceutical researcher who's
(21:31):
a listed inventor and she says,well, I use this platform from
this company and no, I don'tknow how it works, but I asked
it to generate a new drug and itdid, and it said it made a
million of them in this, youknow, had all kind of the
traditional criteria we look for, and I don't know how it
predicted that.
And you know, it just told methis was the best drug.
So I gave it to our patentattorneys and they filed it.
You know you would now losethis patent in the United States
(21:54):
, but anyway.
So back to the high court.
So that's thing one.
So we are, you know, trying toexplore where these lines lay
and arguing.
You know, fundamentally, if theUK IPO wants to take the
position that it doesn't care,which is whether the
applications were, whether adivisional filed while the
(22:23):
applications were live but afterthey were eventually deemed
withdrawn, is a valid divisionalbased on the timing of that.
Speaker 2 (22:33):
Let's not do that one today.
Can I go back to your pointabout the UK in particular that
we're not that bothered by somementorship in the sense that, as
you said, there's limitedopportunities to challenge it?
No-transcript.
(23:01):
My understanding of his part ofhis point was that, so you know
, no one's challenging it.
You said it was who it was.
You believed that everything'sokay.
Um, it was I.
I see seeper liked thatargument.
We actually put a submission tothe supreme court, as you know,
kind of supporting that, thatline of thinking.
Um, obviously it didn't comethrough.
But do you think that's thedirection that the legislation
(23:24):
should go?
It's just to keep away from theAmerican prevalence of the
invention, as it were, and tokeep it between anyone who
thinks there might be aninvention.
Leave everyone else out of it.
Speaker 1 (23:36):
DAVID BASTIAN yeah, I mean it's a good question.
It is clearly a broader pointthan one related to AI-generated
inventions.
That is, you know, and I thinkit was actually I just got back
from China the AI PPI and thatwas one of the study questions
there on harmonization ofdisclosure requirements and this
sort of thing ventureship.
It generally just matters toentitlement and very few people
(23:59):
can challenge it.
You know, Burse's point wasthat and also that the UK IPO
shouldn't be in the business ofverifying these forms.
The form was filed.
Their job is to process theform.
There was a bona fide beliefthat the form was accurate
anyway, but really none ofthat's the IPO's business.
We don't have a concept offraud on the patent office here.
(24:20):
The way that we do in thestates and I think first was
kind of looking over at americaand seeing everyone suing
everyone else about everythingand thinking we probably don't
want that here in the uk.
Um, you know that that'sprobably right.
Um, I think it probably isright.
(24:40):
The other components of thiskind of issue and question being
you know, what rule do we wantto have for this sort of thing?
And the UK is kind ofinteresting, just again to bring
copyright into it.
They were the firstjurisdiction anywhere in 1988 to
explicitly change copyright lawto say, when you have a
(25:02):
traditionally authorless work,we're going to provide
protection for computergenerated work under the special
scheme, and I did think thatthat was particularly forward
looking and future proofing.
And granted, people in the 80sapparently thought this needed
urgent answering and they were35 years ahead of their time.
But but, on the ai generatedissue, it would be nice for
(25:23):
parliament to kind of step inand say this is what we want and
we want it for good policyreasons and here's the law on it
.
But I do think also that youknow the patent act as it is did
have an answer in that,particularly if you looked at it
purposefully rather than kindof a a literalist, uh approach
to the question so you this isbecoming?
Speaker 2 (25:45):
can I call this a passion?
Is this a passion?
This whole ai thing for yourown yeah.
Speaker 1 (25:51):
Yeah, it sounds a little, a little odd.
If I'm passionate about ai,that sounds like I have an ai
girlfriend.
But uh, I don't.
Speaker 2 (26:03):
I'm fishing for that.
That's good to know.
Speaker 1 (26:08):
I was going to talk about.
Are we still plugging your book?
Oh, please, I have a couple ofnew books but yeah, plug the
Reasonable Robot.
Speaker 2 (26:13):
The Reasonable Robot is a really, really good read
everybody listening but also Ithink it's one of the early kind
of full discussions about whathappens when we start accepting
ai as almost as a person, andobviously the book was quite
took it beyond ip significantlyinto.
I know, definitely carinsurance was an interesting
topic and various other thingsas well, but really interesting.
(26:34):
But it feels like that debatehas become more mainstream now.
There are more discussionsabout what if, what if?
I saw in the US there's one atthe moment.
The USPTO has got aconsultation out about AI and
the risk of AI creating loads ofprior art.
Is this familiar?
Speaker 1 (26:56):
Yes, it is.
Yes, it is so the well, youpicked up two veins there, but I
will remember to something forplugging the reasonable robot
and historically you've had acopy of it to wave around or hit
Lee with.
But I'm sure it's theresomewhere, it is not?
Yeah, that clearly is it Right?
(27:17):
So you know, ip agencies andregulators have been thinking
about AI and IP for a long timeand again, it's always been kind
of very academic, largely, youknow the UK being a notable
example of a different event incopyright.
But the USPTO kind of set a newwave of this off in 2019, where
they released two requests forcomments, and the summary of
(27:39):
which was not much to see here,and then kind of a lot of other
people got in the mix, like UKIPO.
You know they've recently donea 180 on this, as I mentioned,
and now they're thinking of allthe other interesting ways that
AI is going to change IP law anddigging into a couple of these,
and there really are a lot ofthem you know with how.
You know the book broadly dealtwith the phenomenon of AI
(27:59):
behaving like a person and howthe law deals with that.
You know whether it's inventingor creating something or
setting standards for humanbehavior, like reasonableness,
including in the self-drivingcar field, you know.
But we have kind ofreasonableness standards in
patent law with the personhaving ordinary skill in the art
.
And as you have an AI thatstarts automating routine R&D
(28:20):
activities, you know how doesthat change standards?
You either have an AI representthe person having ordinary
skill in the art or the personhaving ordinary skill in the art
is a human researcher using agenerative AI system.
And you know semantically thoseare different, but functionally
not so much.
You know other.
You know questions include youcan have an AI make a huge
(28:42):
amount of stuff.
You know is that prior art andis that a problem?
And you know in the US you knowprior art has to be enabled.
So you know, for you know youmay have.
We've had a company, forexample, all prior art that's
been working on having an AImake every conceivable sentence
(29:03):
for many years.
You know a claim has to be onesentence in the US.
So you know, if in a databaseof a quadrillion sentences you
have an identical, you knowpassage, you know or claim, is
that prior art?
Is it enabled?
Is it accessible?
(29:23):
You know, one of the reallyinteresting areas that I don't
work in so much but is AI makingdesign patents or registered
designs?
Because AI is really good atspitting out variations on
industrial designs and thoseapplications are just drawings.
So, you know, is AI preemptingwhole fields of activity and is
that a good or a bad thing?
(29:43):
And do we want lawmakers tocarve out that sort of stuff?
My own view on this was, youknow, you don't want a two-track
system for AI and people.
You know, if you have AI makinguseful prior art, that is
something that should beconsidered prior art.
It would be very difficult todeal with otherwise, all the
incentives in the right place,but that if you do have this
(30:05):
gigantic database, no one couldpossibly use and has no value
and just has sentences thatisn't enabled prior art yeah,
it's also weird.
Speaker 2 (30:14):
Why would somebody use ai to create loads of prior
art?
It's just an old concept.
They're gonna hope to monetizeit, or something.
Speaker 1 (30:22):
Well, for a few things I mean.
One is people do deliberatelypoison the well as a research
strategy.
So if you aren't planning topatent in a space that you're
active in, you might want torelease prior art to keep other
people from patenting in thespace.
If you don't want to, you know,enforce patents.
You know some people you knowhave a view of patents as
(30:46):
inherently bad and kind ofacademically want to blanket
spaces and publications toprevent, you know, monopolies
growing up in the space.
Those are kind of two of thebig reasons, I think.
Speaker 2 (31:00):
Yeah, it just interests me that it never
occurred to me that it would besomething that needed
legislation but clearly can Ichuck one in that.
Speaker 4 (31:07):
So you know that I usually ask the daft question,
so let me ask a daft question.
Um, so we've been doing quite alot of work both on the podcast
and in conversations around, uh, non-practicing entities.
So your patent troll typearrangement If AI gets
(31:28):
widespread inventorship, couldyou foresee a future where AI is
used in that kind of aggressiveway to just flood the market
with patents just for thepurposes of litigation?
Speaker 1 (31:37):
Sure, although there's a lot to unpack in that
quote, unquote, daft questionand I don't think it was
terribly daft and I have beensticking pretty patented 30
years.
But I assume that is yourtarget market here and we're not
too much hitting up the sixthform listeners.
I do have a daughter in sixthform who's not listening to this
or I'm sure will not listen tothis and Well, so there's a kind
(32:03):
of different views on whethernon-practicing entities or you
know the derogative term, patenttrolls, are a good or a bad
thing for the patent system andwhether you know right.
So there's all that too Right.
But you know, yes, this gets tokind of the concern which I kind
of mentioned in the copyrightspace.
But let's say you get.
You know, ai isn't that greatyet, but we've seen how very
(32:24):
quickly it improves in creativestuff.
Let's say, in five or 10 yearsyou really do have an AI where
you can say you know, I'd like adrug to treat this or optimize
this design, and it just does it.
And it does it extremely welland it is really prolific at
doing that sort of thing.
And, yes, you would be able toget lots and lots of patents on
(32:47):
that activity.
Now I tend to think that's agood thing because you know,
let's say you were able to takea sample of someone's cancer and
run it through an AI and the AIcan make a personalized
antibody to treat your cancer.
Right, you know, you couldmonopolize the field for
(33:10):
treating cancer.
On the other hand, that meanswe've now got the cure for
cancer.
So if AI is spitting out allthis really useful stuff at
volume, that just means we'regetting what the system was
intended to do socially valuablestuff.
And you know, there is a 20year kind of limit on how much
protection someone can have, andafter that, then we all have
free access to the cure forcancer in a way that we didn't
previously.
(33:31):
You know someone could, you know, do some you know bad stuff
with that?
If you had the cure for cancer,you know, only let someone have
it in return for a billiondollars, you know, per person.
You know are safeguards builtinto patent law, like compulsory
licenses to deal with that sortof thing.
But also, again, what we saw inthe copyright was a little bit
(33:52):
less.
One company is going to controlthe stuff and more platform
companies are making thisavailable so that you could
potentially use AI to makesomething inventive.
But we will see how it washesout.
But yes, as you say, there is aconcern someone could try and
really dominate certainscientific areas with these
technologies.
Speaker 2 (34:13):
RAOUL PAL.
I'm going to just touch onsomething that has been one of
the big themes Is this as yousay, ai just treats it like it's
a person.
But also AI is the skilledperson which begins to introduce
a whole bunch of interestingpatenty things.
For the non-patent specialist,the skilled person is the
nominal, non-inventive butknowledgeable person who never
(34:37):
has an inventive thought, whoknows how to follow the
instructions in a patentdocument to make it work, and
various other kind of tests allsit around this nominal skilled
person.
And, ryan, I know you had agood think about what it means.
If AI becomes a skilled personor if that becomes the test,
then that causes all kinds ofproblems later on because
(34:58):
everyone else is basically not agenius.
An example, going back to thepharmaceuticals that.
So if you suddenly say, well,just because ai did it, um,
there's no inventive in in there, um, it just messes, messes
things up, um, is this debatestill going on about that
conflation?
Speaker 1 (35:17):
oh yes, I, I will remark.
You seem to look right up at mewhen you said a person who's
ever had an inventive thought.
But but Well, messes it up orfixes things Right.
So you know we have this bar inpatent law, which is we don't
want to give a patent forsomething that you know an
average researcher would havecome up in the ordinary course
(35:39):
of things with, because you know, economically speaking, we
didn't really need thatincentive for that invention to
come about.
Right, we only want to rewardextraordinary effort.
You know something that we thinkwouldn't have been invented but
for a patent incentive.
You know, or at least a patentincentive significantly sped up
the introduction of thisinvention to society in a
(36:00):
trade-off we find worthwhile.
Now let's fast forward 10 or 20years and assume that we have
AI that is really vastly morecapable than human beings in
certain areas like drugdiscovery.
Like you have a disease and youjust ask an AI to make a
treatment for it and it does,that then becomes routine
(36:21):
activity that you don't need apatent to incentivize, and so it
probably isn't something thatwe want to grant the patent for.
You know, it's something thatreally anyone could have done,
or any average researcher.
Now, in this hypotheticalfuture in which AI is
outperforming people, you know,yes, it becomes very difficult
for an unaided natural person tomake an inventive thing, you
(36:45):
know.
But again, otherwise, youbasically have a protectionist
system where we are rewardingcertain kinds of human effort,
even though we could have had anAI do it very cheaply and
quickly, just to encouragesomeone to be an inventor, and I
don't think protectionism hasreally ever been the route to go
with dealing with technology.
I don't remember if I used thismetaphor in the last podcast,
(37:07):
so I'll use it again.
Well, not metaphor, but example.
I was in Brazil some years agoand they had elevator attendance
in all of the commercialbuildings and I asked why that
was and they said, well, it wasa union thing to give people
jobs.
And that did give people jobs,but it wasn't a very productive
use of their labor.
We could have had elevatorpanels and they could have been
out doing something moreproductive.
(37:28):
And you know, as people arestruggling with this very much
so in the creative industryright now, I don't think the
solution is like you havespecial incentives for people or
you ban the use of AI.
You know AI will replacecertain jobs.
It will make other jobs.
You know we should havepolicies that help people who've
(37:55):
been rendered technologicallyunemployed with more benefits
and retraining, you know, butthat is, have an unlimited
amount of wealth and leisure asa society and the goal there is
to just keep all of the benefitsfrom flowing to Elon Musk and
Sam Altman and making sure thatwe get some of them.
You know, speaking of your kindof future question about you
(38:16):
know, ai doing things in thefuture, sam Altman and Elon Musk
say a lot of things, but theyalso think we're going to have
AI that's broadly at theintellectual capabilities of a
human being in a couple of yearsand if they're right about that
, that really is a prettydisruptive social concept, but
also for patent law.
Speaker 4 (38:37):
So, Ryan, one of my jobs on the podcast is to kind
of keep an eye on the clock andyou'll either be delighted or
terribly frustrated.
They will be over 45 minutesnow.
I've got at least one otherquestion.
I don't know whether Gwilym'stotally exhausted.
Are you exhausted or can I?
Should I, because, because mineis a bit of like taking us
towards the end question, Idon't want to take a stay if
(38:57):
you've got more to to ask Gwilym.
Speaker 1 (38:59):
I was going to ask that question about plumbing,
but I guess we do that after thepodcast at the pub.
Speaker 2 (39:05):
No, go for it, Lee.
Speaker 4 (39:07):
So you were talking kind of like end game there in
terms of human level artificialintelligence and what that might
mean generally as like a socialconstruct.
But what's end game for you,ryan?
How long do you keep this going?
Is this end game that thereends up being some kind of
international harmonizationaround AI?
As an inventor, when'd you stepoff the roundabout?
Speaker 1 (39:31):
Yeah, no, I think I'm on the roundabout for a good
while, which is nice because itmeans I don't have to think up
anything new.
This is really the case thatjust keeps on paying dividends
and I just keep writing the samelaw review over and over and
giving the same talk over andover, and this is really sort of
the academic stream.
But when I first spoke toreporters about this in 2019,
when everyone thought this was aridiculous publicity stunt and
(39:53):
you know I no longer get thattoo much you know the reporters
are like so when will this allbe done?
Six weeks, and I'm like youknow, probably at least 10 years
and they seemed unable toaccept that as an answer.
But here we are, five years in,and I think we started kind of
the work in 2018.
So we're at least halfwaythrough.
(40:14):
I mean, look, jurisdictions haveto decide for themselves what
they want their rules to be.
I think this case law hasreally helped stakeholders.
It's promoted discussion.
It's advanced one view of howthings might want to be.
It goes down from courts topolicymakers to kind of have
public debates and discussionsabout this and then ultimately,
(40:35):
yes, the ideal would be that youdo want some sort of
international harmonization,because it shouldn't matter if
you're using AI to make aninventive thing in the uk or the
us or china or you knowwherever you are.
Uh, an invention's an invention, and so you know, hopefully at
some point in the next 30 yearswe'll get something like that,
and in the meantime I get tokeep appearing on podcasts 30
(40:59):
years, guillem?
Speaker 4 (41:00):
have you got another 30 years in you mate?
Speaker 2 (41:02):
oh, yeah, yeah, yeah, I'm nothing stopping me.
Speaker 4 (41:06):
I'll be nearly 90.
No, that's kind of nothappening.
I won't be here, so I thinkwe're done or thereabouts.
Except I know that Gwilym'ssitting on a question that
you're desperate to ask, aren'tyou?
Speaker 2 (41:17):
I am.
I've just realised I close huh.
I'm going to have a closingquestion.
Ask Lily first, then you.
Ryan, I've got one today, butI'm going to be surprised if
it's the same one we did lasttime.
It doesn't matter, because thatwas 2020.
Speaker 4 (41:31):
Is it?
Speaker 2 (41:32):
I can't remember what we did last time, so my answer
won't be consistent, but it'sokay, because then we can
basically compare our answers,which are probably going to be
different.
It's very simple.
It's always 42 42.
(41:52):
By the way, lee, that's deepthought.
Now that's an ai, that was um.
So, first of all, um, we'vetalked lots and lots about ai
becoming, you know, treated likea person and replacing the
skilled person and all theseother driving cars.
We've talked about that.
But you read the reasonablerobots very interesting, um.
But um, oh yeah, what yourother?
What are your other books, ryan?
That's what I wanted to askbefore I get to my closer, sorry
.
Speaker 1 (42:10):
Oh, the Research Handbook on AI and IP by Edward
Elgar, and the InternationalIntellectual Property in a
World-Integrated Economy byAspen Publishers.
So those books are even moreriveting than the Reasonable
Robot, one of which is aresearch handbook and one of
which is a textbook.
But I'm now working on anencyclopedia of AI and law, so
(42:33):
that's going to be the most pageturning yet.
Speaker 2 (42:38):
Typical name the Reasonable Robot.
You've got to read that, anyway.
Closing question AI.
Speaker 4 (42:51):
Lee, what job that you do would you most like AI to
do instead?
Oh, I guess the obvious answeris none, because I don't like
working for a living.
That's not true.
I quite like getting paid,which is slightly different,
isn't it?
You don't like money, do you?
I love my job.
I love working with people likeyou, gwilym, so it's probably
something I don't know whetherAI would be able to do that I'm
not the world's greatestnetworker.
I hate being stuck in a roomwith a load of people I don't
(43:13):
know and being forced to makekind of polite conversation,
perhaps with a view to maybeforging one or two connections
that might be useful at somepoint in the future.
Can't be doing with it, so ifAI could somehow come in and do
all of that networking for meand just give me a bunch of
perfectly formed connections,that would suit me in the future
.
Job done.
Speaker 2 (43:34):
I believe that's referred to as neural networking
.
Thank you.
Ryan, did you set that one up,will?
No?
No, I'm just here, I'm all week.
Uh, you, you're on fire.
So it's what, ryan, what jobthat you do would you?
Would you prefer ai did?
Speaker 1 (43:54):
but before I give you an answer, and it's an easy one
, lead have you guysexperimented at all with the
google?
You know, notebook LM.
Speaker 4 (44:02):
Not, no, what, no, what no.
Speaker 1 (44:04):
Yeah well, so you know, take a paper or something
and you drop it in a folder anda Google podcast emerges from
this of two kind of emotionallyintelligent sounding, human
sounding voices debating thesubject of this thing.
It's basically an automaticpodcast generator and it takes
about two minutes to make one,but no one would listen to that
(44:27):
and replace you two at this job.
I'm sure Me grading papers Imean my students use LLMs to
write the papers.
I should be able to use LLMs tograde the papers, at which
point one kind of wonders aboutthe purpose of the whole system.
But anyway, I I don't likegrading lee, I'll jump in.
Speaker 4 (44:48):
I know you're going to ask me my answer I was going
to come back on the gradingthing first, because I used to.
I I was I was a teacher infurther and higher education and
I always found it easy just tostand at the top of the stairs
and lob papers down the stairs,and the further they flew, the
heavier they were, so the higherthe grade yeah, I mean.
Speaker 1 (45:04):
Well, to be fair, that is more or less what I do.
But it would be even easier tojust upload it all to an ai and
have it put the numbers in, oror to simulate the stair
experiment I'm sure we're reallyinfusing your students as we
speak, so um they're notlistening, don't worry perfect.
Speaker 4 (45:21):
Yeah, no one does grill them.
Come on then.
What's yours?
Speaker 2 (45:25):
I'm gonna be festive christmas shopping.
I do it all online anyway, justwant ai to go and do it.
Speaker 4 (45:30):
It's just innovating, it's innovating do you find
that more and more people areusing these blooming online list
things for christmas shopping?
We we do it now with likesecret santa at work.
Now we do it in our family,where you kind of upload a wish
list and someone will buy you agift.
What's the point?
What's the point in me tellingpeople things that they can buy
(45:52):
me when I can just buy themmyself?
Speaker 2 (45:54):
that's a good point, although my family approach is
that if you ask for somethingyou you can't have it, so you
have to ask for something youdon't want and then then you
won't get it and you'll get whatyou might want.
It's quite complicated.
Speaker 4 (46:06):
Oh, some subtle kind of psychology going on.
I've not really thought aboutthat.
Speaker 2 (46:09):
It's quite messy.
Speaker 1 (46:10):
Well, it sounds better.
I'm supposed to guess whatpeople want, and if I guess
wrong I get punished.
Speaker 4 (46:18):
So I prefer your list of urges.
I don't have enough time in mylife to sit down and build an
Amazon gift list.
It's just not going to happen.
Speaker 2 (46:25):
You're the man who owns everything.
Anyway, let's face it.
Speaker 1 (46:27):
Well there you go.
Speaker 4 (46:29):
Yeah, yeah Well, ryan , thank you for coming back and
gracing the podcast.
Always a pleasure to have youon.
Gwilym, thank you for well justbeing the consummate co-host,
as you always are, and thanks toall of our listeners who are
now going to rush off and leaveus a little review on the
podcasting platform of theirchoice so that other people can
find us.
Maybe that's something that aicould do for us.
(46:49):
Perhaps we could get some kindof ai crawly bot thing, or
whatever they're called, tocrawl all over the podcasting
world and leave lots of reviewsabout two ips.
Thanks both cheers, cheers.
We'll see you next time.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.