Latest / Coopers' Code / Meta, Google, and AI Liability: Evolving Legal Risks and AI's Dirty Little Secrets

Transcript

Meta, Google, and AI Liability: Evolving Legal Risks and AI's Dirty Little Secrets · Coopers' Code

0:00
Being informed is the best way to calm a lot of your fears
0:04
about this, right? Is that OK, let's understand
0:07
what's happening and let's see how it's developing and what
0:11
checks and balances are being put in place.
0:13
But ultimately the idea that this has to be an industry
0:18
that's unregulated and unchecked, someone has to think
0:22
about it differently because we know how that plays out when
0:26
industries have no boundaries. And I don't think we have faced
0:31
a tool or a product that's this powerful before, other than the
0:37
atom bomb maybe, right. There has to be some safeguards
0:41
put around it. And the owners of these products
0:45
have to understand, what do you want your product to have
0:48
longevity or do you want to just make the most powerful thing
0:51
that then just blows everything out?
0:53
And that's a a reality that they have to come to grips with.
0:57
Fortunately on the regulatory side and it and this just popped
1:00
up in my my feed just now open AI is coming out with a new
1:03
product called Skynet and it's supposed to to be a perfect
1:06
regulatory process. Howdy and welcome to the show.
1:12
This is another episode of the tech broadcast in terms of
1:15
covering AI and and all things tech and AI related in law and
1:19
and law. Jason Bye, blessed with Sherrod
1:23
Mullenfar's presence today. And unfortunately, Marshall
1:28
cannot join us. But thank you for joining us,
1:30
Sherrod. Of course.
1:32
So and we have quite an agenda today.
1:36
I think the first place that we should probably start in terms
1:40
of alpha and omegas is on social media trials.
1:44
Since we last spoke about them going forward, there have been
1:48
some verdicts that have come in. I think the last time we talked
1:51
about the fact that Meta and Google were in trial in Los
1:55
Angeles with Mark Lanier and Rahul Rabuti and saw their great
1:59
lawyers and they were able to obtain, I believe a $3,000,000
2:02
competitory damage award and $2.1 million punitive damage
2:07
award. And I believe a $900,000 award
2:10
against technically YouTube, which is a division of Alphabet.
2:15
And then there was another case in New Mexico that resulted in a
2:19
$375 million verdict as well. And you know, it's interesting
2:24
that the big verdict is coming out in New Mexico as opposed to
2:27
Los Angeles, but that just shows that other parts of the country
2:30
are waking up and holding corporations accountable.
2:34
The thing that. Interested me in New Mexico was
2:36
that was brought by the state itself in terms of violation
2:40
there. So there's a $5000 her violation
2:44
penalty that the jury found $375 million worth of penalties
2:48
against against Meta and and that again was for violating
2:53
consumer production laws and, and the lack of safety related
2:56
to to children and and addiction.
2:58
So one of the things that I found interesting about this was
3:04
these might be some of the first cases that we're going to see
3:08
with similar fact patterns. Did you have any sort of
3:10
takeaways as far as your practice in terms of potential
3:14
people calling or issue spotting when, when, when there are
3:18
potential calls? Yeah.
3:20
I mean, when I was doing defense work, we did a lot of product
3:22
liability work for the auto industry, right.
3:24
And you when you think about a traditional product, you think
3:27
about a car, a be it a blender or something that you're
3:32
tangible using. And now we have these AI systems
3:35
that are becoming far more advanced.
3:38
And I think the innovative thing that they did in that Los
3:41
Angeles case was go after it as a product liability case.
3:45
And these companies have to understand that while no product
3:49
is perfect, having products out there that addict people or harm
3:54
people or tell people to go do unthinkable things, you know,
3:58
and take their own lives are going to have consequences.
4:02
And that these are things that they have to consider in their
4:07
design of those products. But what's I think difficult and
4:11
interesting is that when you're creating a product that's
4:15
supposedly training itself and teaching itself, the question is
4:19
how are they going to put the proper boundaries or bumpers on
4:23
it so that the product doesn't make decisions that are going to
4:26
harm other harm human beings and to do it in a reasonable way.
4:31
And I think this is the part that this industry is struggling
4:35
with because I don't know if they understand what they're
4:40
unleashing in terms of how this product is evolving as it's
4:45
going along. Agreed.
4:48
And in the area of uncertainty and fear, whether rational or
4:56
irrational, probably that that the next place to go or are the
5:00
attacks on Sam Altman's house in San Francisco and the attempted
5:06
firebombing of open AI. Yeah, I mean, the reports came
5:09
out and I think the the guy has already been charged with
5:12
attempted murder for throwing essentially A Molotov cocktail,
5:17
I believe, at Sam Sam Altman's house in San Francisco.
5:20
And this is to a certain extent, you know, building with people's
5:26
frustrations with AI and information, whether it's good
5:29
or bad or misinformation that they're getting.
5:31
And then apparently he went down to the offices of open AI and
5:35
threatened to set the building on fire.
5:37
So, so these are all things that that are there.
5:42
But and interestingly enough, a lot of these AI executives are
5:47
building bunkers in their homes or vacation homes, which is kind
5:52
of curious in terms of, you know, why those things are are
5:55
necessary. But it's unfortunate that his
5:58
family had to go through that. But this is kind of indicative
6:01
of the frustration people are feeling over the news that
6:06
constantly is coming out about how AI is going to take
6:08
everyone's jobs. Yeah, I I found it interesting
6:11
that this this individual was upset enough.
6:13
And then obviously there are levels of mental health that can
6:17
can correlate with these, these types of actions, but was upset
6:22
enough to have travelled all the way from Texas to do this.
6:25
And, and then in short order, there were shots fired in
6:29
Altman's neighborhood and, and those shots also hit his house.
6:32
Separate. Separate incident.
6:34
Yeah. And I don't know if that I don't
6:35
believe I've seen anything that indicates that was targeted at
6:38
him, but it's just the convergence of horrible things
6:42
that happened at or near his house during during the past few
6:45
days. Yeah.
6:48
I guess parallel to AI and kind of the magnitude of the changes
6:56
that that really caught me unaware.
7:00
I'll share with you that just before we were to record today,
7:05
the team that does IT work related to our firm was was here
7:11
and he was really stunned by Anthropics Mythos.
7:20
It was something that I quite honestly I had not heard of, but
7:24
within the last week a a quad adjacent product Mythos had
7:30
uncovered numerous 0 day exploits.
7:36
And for those who, who, who are not familiar, when operating
7:40
systems come out, there are sometimes ways that one can
7:45
penetrate the operating system and, and get through security.
7:49
I think the reason our IT person was so concerned about it and,
7:54
and in fact, he commented, you know, mark your calendar.
7:56
This is the week where security changed that we really going
8:00
forward are going to reach a point where it will be near
8:04
impossible to lock down the information that we, we have
8:09
even with two factor authentication.
8:11
This is now my my monologue. My question to him was you do
8:16
our IT work and you you've been doing this a long time.
8:19
What recommendations do you have as as far as a business is
8:22
concerned? And and paraphrasing when he and
8:24
us and said pray things are going to shift and that there
8:27
will be very little that we can do about it.
8:29
When you look at this issue, right, I mean, before when you
8:33
had the, you know, all the encryption that was developed
8:37
over the years, you know, I always think about like a
8:40
hackers to some kids sitting in a basement and you know, create
8:44
coming up with creative ways to try to hack a system.
8:48
And now the problem is that when you have supercomputing
8:52
involved, the number of calculations of this thing can
8:55
do in a millisecond is something that essentially is probably the
9:01
factor of trillions of times faster than we were able to do
9:04
before. And that's what encryption
9:07
really was based on, is the fact that it would take him, you
9:10
know, years and decades to break it if they they were able to do
9:14
it. And now you've got machines that
9:16
can do it very quickly. And the question is how
9:19
vulnerable are the systems that are out there and how is that
9:24
going to impact the big, you know, the big picture?
9:26
And with this anthropic product, Methos, apparently it wasn't a
9:30
closed system, like it had no Internet access.
9:34
And somehow it managed to figure out and exploit the, you know,
9:40
whatever ports that were available to it, access the
9:45
Internet and then go out and essentially post on a, you know,
9:51
chat bot platform, which then caught the attention of security
9:57
researchers. And that's how they figured out
9:59
what had happened. And this is something with that
10:02
Anthropic, you know, to their credit, has raised the flag on
10:05
and said, hey, look, this is now where we're getting to.
10:08
And all these companies need to be aware of what they're about
10:10
to do. To follow up on that piece and
10:13
Tropic did not release the product to the general public.
10:16
They gave it to the software manufacturers to help find these
10:21
vulnerabilities so that it's the software companies who who are
10:25
aware of them and can fix them. But some of the vulnerabilities,
10:28
I think 1 was from a 27 year old operating system that had just
10:32
never been found by any of the the hackers out there.
10:35
So what What the raw magnitude of computing power has been able
10:40
to to turn up is is really quite interesting.
10:44
Yeah, and interestingly enough, the Alphabet or Google CEO was
10:49
a, you know, in a segment of 60 minutes kind of talking about
10:53
how the US must be responsible in the development of AI.
10:57
So we'll see how companies like Google, who are the leaders in
11:00
this industry, are going to lead in trying to make this a more
11:06
ethical or responsible growth as opposed to just unleashing this
11:12
and hoping that things go well. I appreciate that.
11:15
I, I think the challenges and, and we've seen this with other
11:18
AI products, but with, with Mythos, yet again, they, they
11:24
don't know exactly how it is doing what it does it.
11:27
It's almost as if we are reaching the point where we,
11:30
from a technological standpoint, analogue the human body, you
11:36
know, doctors have been working for millennia on the human body
11:41
and we are still kind of kind of at a very base level in terms of
11:46
understanding how it works. Now we have software AI tools
11:52
that are beyond our capacity of understanding how they are doing
11:57
what they do. Yeah.
11:58
And I mean that's goes to the analogy I think we talked about
12:01
was, you know, if there's a black box that you don't really
12:07
know what's in it or what it's capable of it.
12:09
But we're keeping, you know, keeping it fed and basically
12:14
giving it more and more information and energy to
12:18
develop. And once that supercomputing
12:22
pieces in there, if it's smarter than us and if it's able to
12:27
access resources in a way where no single human being on its own
12:30
can do, then obviously that's going to be an issue in terms of
12:35
how do we control what it does and and how it reacts to things
12:40
as it as it grows and gets more and more powerful.
12:44
I'm getting an image of the plant from a little Shop of
12:47
Horrors from from way back in the day.
12:49
Martin, Red Brown. We're now aging our dating
12:53
ourselves, I should say. When I make reference to movies
12:56
that have been filmed before, a lot, lot of our listeners have
12:58
been born. That's that is an issue.
12:59
But it is it is an oldie and a good one.
13:03
Fortunately, I've made our kids watch the, the, you know, the
13:06
movies I watched as a kid just to, you know, get them to
13:09
understand the humor as it comes out.
13:13
In a a parallel vein, product liability as it relates to AI,
13:17
you brought up the dancing robot and the dancing robot fiasco,
13:20
which we'll put a link to in in the show because it is
13:25
stunningly amusing and a little frightening if it it had gotten
13:29
out of control in a different way.
13:30
But but I'd be interested in your thoughts on that on the
13:32
product liability side. So just to give the audience
13:35
context. So in the last couple of weeks,
13:37
there's a video that came out of a robot that was, you know, a
13:40
little robot in a restaurant that dances around and
13:42
entertains the customers. And the robot had a malfunction
13:46
and started kind of beating on a table.
13:49
And this one of the poor servers whose look on her face is like,
13:53
I don't get paid enough for this stuff, is trying to hold this
13:57
thing back. And it keeps, you know, dancing
13:59
and hitting the table. And it, it made me think like,
14:03
all right, you've got these robots.
14:04
And there have been sightings of robots walking around different
14:07
cities around the United States. And China did an incredible
14:10
demonstration of the robots. I don't know if you saw that,
14:14
but these robots were basically performing like Shaolin monks,
14:19
had some sort of a, you know, a country function and their
14:23
ability to do backflips off walls and balance was just
14:27
phenomenal. So the question is there right
14:30
when you've got a robot and if that robot malfunctions and
14:33
punches somebody in the face or you know, hurts a child or
14:36
something like that, the product liability consequences that it's
14:41
going to have. And then you know, what about
14:44
the business? Is the business liable for the
14:46
robots actions? Is the manufacturer?
14:49
What kind of a document did the business sign to purchase that
14:53
robot? It's going to be an interesting
14:56
onion to peel back as these things become, you know, part of
15:00
our lives. And I don't know if you heard
15:03
Elon Musk talk about his trillion dollar pay package, but
15:06
one of the things he said was I need essentially 25% of Tesla's
15:10
stock because I'm building a robot army.
15:14
And I want to make sure nobody wrestles the control of that
15:16
army away from me. When I think about people that I
15:20
want in control of robot army, I cannot think of anyone else who
15:23
makes me feel more well maybe, maybe, maybe our commander in
15:25
chief. But but that those two I think
15:29
would be the world 2nd 1/2. Yeah, so you know, it it, it'll
15:32
be interesting to see how this happens.
15:35
And then, you know, you've got, you know, cars driving on the
15:38
wrong side of the streets and, you know, hitting pedestrians or
15:44
you have these delivery robots now just blocking fire trucks,
15:48
right? The fire truck doesn't get to
15:50
the person's home. They, you know, somebody dies in
15:53
that fire, you've got potential liability there.
15:57
So there's a lot of things that could happen.
16:00
And sometimes I see the the stuff that's occurring and you
16:03
think, well, you can't really make this up.
16:06
You know, it's just amazing that these things are occurring now.
16:10
Yeah, On the law front, I went to a wonderful Bar Association
16:14
of San Francisco event that talked about AI and the law.
16:19
And one of the areas that that they focused on were some issues
16:25
related to some new cases that have come out.
16:27
Law is trying to keep up with AI on the evidence front, which is
16:32
always an interesting, can law move as fast as tech?
16:36
Usually not. And so from an evidence
16:39
perspective, looking at some of the rules are interesting.
16:42
The rulings that certainly caught my attention in terms of
16:48
being cautious with what we do and what we advise our clients
16:52
to do and what Imbro pers do. And I'll just give a brief
16:56
summary of these and again, we'll, we'll, we'll link to
16:59
these in in the show notes. United States versus Hebner was
17:02
a case where a criminal defendant took areas that the
17:09
defendant had questions about plugged into AI and ultimately
17:13
this was located when a device was seized.
17:16
And the judge determined that the questions that the criminal
17:19
defendant were asking about his own potential criminal liability
17:24
were things that were were discoverable and, and could be
17:27
shown to a jury. The piece there that I thought
17:30
and, and, and Sherrod, I would be interested in in your
17:33
thoughts in terms of how that that filters into our practice.
17:36
I'm guessing that you have had clients who have come to you and
17:40
with their own research like thank, thanks for your demand
17:44
package or your mediation brief or whatever it is.
17:47
But I've done some research and it tells me that my case is
17:50
worth this or that we're going to win because of why has that
17:53
come up for you? It did recently in a interesting
17:56
context and that was that we'd already settled the case and the
18:00
settlement agreement was sent to the client.
18:02
And then we got feedback. And one of the pieces of
18:06
feedback was I want these changes made to make the
18:10
agreement conform to the standards that are acceptable in
18:14
the industry, which were not things that I had seen and, you
18:18
know, over, you know, 20 some odd years of practice.
18:22
And the client also, you know, had an idea that the funds
18:28
should be paid before the settlement agreement is signed.
18:31
And again, I've done both defense plaintiffs and mediation
18:35
work as a mediator. And again, if you're more power
18:38
to you if you can pull that off, but I've never seen that done.
18:42
So those are things that they genuinely believe that this was
18:47
the way to do it. And, you know, it got to a point
18:49
where I had to basically say, look, you have to trust me.
18:52
They're not going to somehow do something shady.
18:56
And you know, I've dealt with these lawyers in this
18:59
particular, you know, insurance carrier and I, I've never had an
19:03
issue. I don't know anybody else who's
19:04
had an issue, but it creates legitimate concerns.
19:07
And in fact, you know, as a side note, which is interesting,
19:11
there was a real estate agent in New York who posted that he had
19:14
some $50 million deal going. And both the seller and the
19:18
buyer went to open AI ChatGPT and said, hey, am I getting
19:24
overpaying or not paying enough or not receiving enough?
19:28
And it gave them both opposite information, like it told the
19:32
buyer you're overpaying, it told the seller you're not getting
19:35
enough money. So it just kind of goes to show
19:37
about the lack of consistency. And again, going back to one of
19:41
our early episodes, these systems are people pleasers.
19:44
And how you ask the questions could cause the answer to be
19:48
very different than if you were to do it in a more, you know, in
19:54
a different way I should say. Part of the reason I brought
19:58
that up as a foundation piece is I I think.
20:01
Most of us in practice are starting to see clients or
20:03
potential clients are using AI as a way to get more insight
20:09
into the mystery that is awe. And, and I, I don't begrudge a
20:14
client or potential client for doing that.
20:17
Where, where this particular case, USP Heppner caused me
20:20
concern is we know that one of the things that shows up in a
20:26
lot of our cases are discovery of social media.
20:28
It's just kind of standard, you know, it shows up in
20:31
interrogatories and requests for production etcetera.
20:34
The next evolution of this very likely seems to be a demand for
20:40
all conversations case related with your AI tools because
20:45
arguably under USV Hebner that they would be discoverable.
20:49
Well, I mean, but USV Hebner, in my opinion, was slightly
20:52
different, right? It's basically a criminal
20:55
defendant putting this stuff in there and then that computer
20:58
getting seized or by a actual warrant, not an administrative
21:04
warrant, a real warrant. So that was a little bit
21:07
different in my opinion. But at the same time, it is
21:11
something that's concerning all around because it cuts both
21:14
ways, right? If the defense asked for that,
21:17
we may ask for the defendants, you know, ChatGPT conversations,
21:22
which could open up a whole can of worms that they don't want to
21:25
go down. And so I think it bears, you
21:31
know, saying that all of our clients should be cautious in
21:34
terms of what they put because on social media, nobody puts
21:38
down that, hey, I'm in pain in my back and I can't move today.
21:42
They're always putting on a smile and that sort of thing.
21:45
When in fact, in, you know, in a chat bot, it may be very
21:48
different because they're seeing that chat bot as a confidant and
21:51
they may be more open about what they're going through.
21:54
So who knows how that's going to play out.
21:56
But you know, since this came out, I try to tell our clients,
22:00
look, you know, don't, don't go doing something like this
22:04
because it could have, you know, consequences.
22:07
That was the point that I was ultimately trying to get to is
22:11
while it is uncharted territory, it could very well be an area
22:15
that becomes a, a nation discovery region where people,
22:22
people's questions because it's not attorney work product.
22:26
It's a it's a client having a conversation with a third party
22:30
tool. And I wonder if, what if the
22:32
what if the client was given a login or an account by the
22:36
attorney? The Heppner case may have
22:38
written very differently because that may have been seen
22:43
differently by the court, but that issue was not present.
22:47
It was somebody doing it wholly separately and going in and
22:51
putting that information in before they even provided the
22:54
information to their own attorney.
22:57
That's a perfect way to kind of get into more the Gilbarco,
23:02
which is in Eastern District of Michigan evidentiary ruling
23:06
federal court case. And in that case the the
23:10
plaintiff was a pro SE plaintiff.
23:12
The pro SE plaintiff was using ChatGPT to draft legal material.
23:16
And in in that case, the court found, look, you know, the
23:21
prostate litigant using chat to help come up with legal
23:24
material. That's a tool that's, that's not
23:27
something that is discoverable. And and so that that does go to
23:32
potentially your, your, your notion that if it's within the
23:35
confines of litigation there, there is production for it.
23:39
Well, and I wonder if that argument was made in Heppner,
23:42
that at the time he was doing this, he was pro SE and
23:44
therefore he was simply using a tool as opposed to, you know,
23:49
somebody was represented by an attorney.
23:50
I, I, and I don't know the answer to that, but I don't.
23:53
And it would have, it would be interesting to, you know, figure
23:56
out if that would have made a difference in the judge's
23:58
ruling. My loose memory of Heppner is
24:02
that he was a fairly high this is a white collar fraud.
24:05
Fairly high up executive had it an army of lawyers so the pro
24:10
southeast side probably would have been a hard hard argument
24:12
to me. Who?
24:13
Clearly he did not trust because he wanted to open AI.
24:20
The corollary thing that came up for me was the issue of AI
24:25
transcription versus AI note taking.
24:27
And, and I raised this because so many people use Zoom or Zoom
24:33
variant meetings these days and there are so many AI
24:37
transcription tools, recording tools and note taking tools.
24:41
When one is having a, say, Zoom conversation with a client, you
24:45
know, obviously that that conversation would be
24:47
privileged. But if you're doing that with a,
24:50
a non client, say an expert witness or general witness to,
24:55
to a case, then you have an issue.
24:58
You know, from a transcription perspective, it's it's not
25:00
privileged. A question is if you're
25:04
directing the AI and what's important for you in terms of
25:07
the note taking, is that your work product, attorney work
25:10
product or is it not? Well, I guess depends on if I'm
25:15
asking the other side to produce it or if they're asking me to
25:18
produce it, right? Well, well play.
25:23
Yeah, no, but I mean, look, these are all issues that we're
25:26
we're all, you know, looking at and facing.
25:28
I mean, to be honest, I've, I've been very hesitant to use
25:32
transcription tools because I'm just not comfortable that I've
25:36
done a deep enough dive to understand what's happening with
25:39
that information. I mean, from your voice being
25:42
used by the companies to, you know, God knows what else.
25:47
I mean, you know, there's the reports of people getting calls
25:52
from what they think are their kids and the voices are being
25:56
mimicked. So these are very serious issues
26:00
that we're dealing with. So again, I haven't used those
26:03
tools. So I I really can't speak that
26:05
much about it in terms of what it means in the long run, but
26:09
it's something that we should all be very cautious about.
26:12
Well, we're in the area of transcription and what are the
26:14
other cases in terms of, of on, on the the liability side that
26:18
you sent over there I thought was fascinating was use of AI
26:22
transcription in the medical environment without consent.
26:25
Yeah. So apparently recently Sutter
26:27
Health got sued because they were using an AI system to
26:32
record the conversations that their docs were having with the
26:36
patients during the consult. And so the lawsuit I believe is
26:43
alleging that there was not proper consent by the patients
26:47
to allow this to be done, which is interesting because, you
26:52
know, I've had like one or two doctors encounters where they
26:56
specifically ask you, is it okay for us to do this?
27:00
And so it'll be interesting to see how this case plays out as
27:03
to whether or not Sutters system ever properly notified the the
27:09
the patients that their system was actually recording those
27:12
conversations and then transcribing them for the
27:15
doctors, which I assume are going into the notes that
27:19
they're taking in the chart. Like you, I've, I've had a
27:23
similar doctor's visits where the, the doctor has said a
27:26
variant, what I would consider to be a very well written
27:29
script, which, which was as part of our process to be able to
27:35
focus on you instead of focusing on the machine.
27:37
We're now using an AI transcription tool.
27:40
Is it OK with you? If, if this is if our
27:42
conversation and, and the the session here is, is, is
27:45
transcribed to which I, I always have said yes.
27:48
I mean, it seems like I'd rather have my doctor focusing on me
27:51
than taking notes. Sure.
27:52
So I'm, I'm, I'm wondering if this was early in the setter's
27:55
process and they, they made some errors or if there's a, another
28:00
layer in terms of, because one of the, one of the arguments as
28:03
you, as you drill down into the case a little bit is that this
28:06
information is going to a third party vendor that is outside of
28:10
the Sutter system. And I and I think that that is
28:12
an an additional breach of privacy that is a concern for
28:15
the class. Well, and I guess it, it goes to
28:18
the informed consent piece of it, right, That if they're
28:21
telling the patient, hey, we're recording this and it's being
28:24
transcribed, Are they truly this disclosing exactly what's
28:29
happening? Because quite frankly, the
28:31
doctors don't know exactly what's happening, right?
28:34
It's maybe the management at Sutter who's had discussions
28:37
with this AI company that says, oh, this is a great idea, but do
28:41
they really understand what's happening on the back end in
28:44
order to get that piece of information into the chart?
28:48
And I think one of the things that a lot of people are unaware
28:52
of is the human feedback, which is a lot of the AI products are
28:59
sending this information typically to underpaid,
29:03
overworked individuals in developing countries where the
29:09
there's human feedback going on to confirm that the AI is
29:13
getting it right. Which means that that there are
29:15
there are humans out there who are seeing pictures of your
29:18
stuff or hearing audio of your things.
29:22
And are they likely to tie it back directly to you and say
29:25
you're diagnosed with your doctor?
29:27
Hopefully not, but but perhaps. And I think that that is 1 area
29:30
that that you're right, the doctors don't know and the
29:32
patients don't know. Well, and that was underscored
29:35
in AI think it was a congressional hearing with a
29:37
Waymo executive who confirmed that the people that intervene
29:42
when that Waymo was sitting in the middle of the intersection
29:44
not knowing what to do are essentially a low paid
29:49
individuals in another country. I think it was in the
29:52
Philippines that are taking over and then trying to maneuver that
29:56
car out of intersection or, you know, fix whatever it's doing
30:00
that's not proper. So that that is a dirty little
30:05
secret of AI. And in fact, you know, we talked
30:08
about the American Arbitration Association having their AI
30:11
arbitrator. But when I went and read their
30:13
website, it says, is this all AI?
30:15
No, there's humans that are going to review this.
30:19
Yeah, yeah. I think last in the area of
30:24
cases, I would love to know where where you spend your time
30:27
reading because you always bring them an interesting feed
30:29
forward. So this is the human only rule
30:33
related to copyright protection, which was upheld by the Supreme
30:36
Court. Staler V Perlmutter.
30:39
Do you want to unpack this for us?
30:41
So essentially, the Supreme Court declined to hear a
30:44
copyright case in which the lower court had essentially
30:47
ruled that work product generated by solely AI from
30:53
human input, meaning a human put in the prompt that resulted in
30:57
that work being generated by the AI, is not protected by
31:01
copyright. So there has to be some sort of
31:03
substantial human input in order to make it copyrightable, which
31:08
is interesting. And it opens up a whole host of
31:13
things in terms of how that could have ripple effects in
31:16
copyright and ownership and whose work product is it in
31:21
anyway? And I think there are safeguards
31:26
now that Amazon is putting in place because there are people
31:29
who are publishing like 100 books a day or something and
31:33
they're saying no, no, no, there's no way you're doing
31:35
that. You're basically creating these
31:38
with AI and you need to have something less prolific sort of
31:41
speak in order to put these books on our platform.
31:46
It will be fascinating to see 'cause I'm, I'm sure at some
31:48
point we're going to see a variant of a bright line test of
31:54
what constitutes human input. And I I could see that going a
32:00
variety of ways. What one argument that was
32:03
probably made in in some of the the the breathing related to
32:07
Thaller beat Perlmutter is that if you are designing an AI tool
32:12
and AI tools didn't just self replicate they they came from
32:16
somewhere humans. So isn't the product of an AI
32:24
thus human derived? Yeah, I mean, these are all
32:27
interesting questions. And the, you know, the next step
32:31
in the process in my mind is at what point does open AI come
32:35
with its hand out saying, hey, you know, give us our cut.
32:41
Yeah, right. Because you're going to have
32:42
people who are going to make money off of this.
32:44
And you know, the the big saying around tech now is that, oh, you
32:48
know, you know, a kid sitting in his or her living room can
32:53
become a billionaire overnight. Well, yes, but if they're doing
32:56
it on their platform, I can't imagine that open AI is not
33:00
attempting to monetize that at some point.
33:03
So it's not going to be free going forward it.
33:06
I would hate to disappoint them, but perhaps they will be treated
33:09
exactly the way the New York Times and Getty Images and all
33:11
the others that were scraped were treated, which was, you
33:16
know, point OOO 1 cent for the value of your intellectual
33:20
property, right. One thing that that I, I didn't
33:22
put in the outline, but I think is, is, is a wise thing to bring
33:25
up that that is related to what we just talked about.
33:28
And, and that is the blowback, societal blowback on what is
33:33
being described as AI slot. And, and, and you, whether you
33:36
talk about 100 books a day or there was a restaurant in Santa
33:40
Cruz that used AI to design A logo.
33:43
And it, it went from, you know, near 5 stars down to 1 because
33:50
of a very much an anti AI, anti putting out graphic designers
33:55
out of work. The idea that, that the material
33:58
that is being produced is identifiable and, and, and bad
34:03
and putting humans out of work, there is a risk associated with
34:06
that. And from a law perspective, that
34:11
may be less of an issue. I think employers as as we've
34:14
talked about, there is a desire to be more efficient in the work
34:18
that we do, not to solely rely on machines, but if there's a
34:21
way to get better briefing with less human hours while still
34:27
having the human input that there, there's not a downside to
34:30
that. But I think there is definitely
34:32
a risk in the greater public of an anti AI bias.
34:37
Yeah. And it was interesting.
34:38
I saw a snippet of an interview from, I believe it was the
34:42
president or the chief operating officer of Plant Air.
34:46
He was saying something like, oh, AI is being misunderstood
34:48
because what they're, what we're seeing is if they can make a
34:52
company's more efficient, then would that results in having
34:56
extra shifts, which means hiring.
34:58
And that may be true, but what the, what's grabbing headlines
35:03
is look, you know, these tech companies are now laying off 4,
35:08
five, 10,000 people at a time. And well, if you're doing that,
35:13
how are you going to contribute to or control the narrative that
35:17
AI isn't taking people's jobs because it's taking your
35:20
people's jobs. So there, it's, it'll be
35:24
interesting to see how it plays out.
35:26
But I think for most people, especially attorneys, if you can
35:30
somehow enhance your abilities using AI and like you know,
35:35
we've talked about, our firms are not huge firms, but we can
35:38
stand up to firms that have thousands of lawyers.
35:42
And we can do that because we are figuring out efficient ways
35:46
of utilizing this tool so that we can increase and better
35:50
provide our clients services that, you know, maybe wouldn't
35:54
be available without those tools.
35:57
So those are all things that we have to consider.
35:59
But again, there is going to be certain portions of the
36:04
population who I think are going to get a hit with AI.
36:07
And, you know, with these robots coming on board, they're trying
36:10
to change and automate how everything is done.
36:15
I mean, I saw footage of a textile factory, I believe it
36:19
was in India. And these workers were made to
36:22
wear cameras on their foreheads so that they could record their
36:26
movements to program AI. So eventually a robot can
36:31
replicate those movements. So who knows what's going to
36:36
happen? There's so many directions that
36:39
this could go. But hopefully the, you know,
36:43
it'll be for the betterment of humanity as opposed to something
36:47
that's, you know, requiring people to go in now and hide in
36:50
their bunkers. A variant of, of, of what came
36:53
up earlier, you know, better minds than than mine.
36:57
You know, the, the billionaire tech founders, Elon Musk,
37:00
they've got this figured out. And I'm just putting their
37:02
faith, my faith in them, and it'll all be fine.
37:05
And for the record, this podcast stands for that.
37:08
You know it's going to be all roses and that AI is going to
37:12
enhance everything that we. Do well, you're laughing and and
37:17
and I think that there the future is very uncertain at this
37:22
point. And I think that there are a lot
37:24
of potential benefits and there are a lot of potential risks and
37:29
embracing the benefits and being high Wizard of the risks is, is
37:32
really the only way that we can can move forward.
37:35
None of us can control the future.
37:37
Right. And I think being informed is
37:40
the best way to calm a lot of your fears about this, right?
37:43
Is that OK, let's understand what's happening and let's see
37:48
how it's developing and what checks and balances are being
37:51
put in place. But ultimately, the idea that
37:54
this has to be an industry that's unregulated and
37:58
unchecked, someone has to think about it differently because we
38:03
know how that plays out when industries have no boundaries.
38:07
And I don't think we have faced a tool or a product that's this
38:13
powerful before, other than the atom bomb maybe, right?
38:18
So there has to be some safeguards put around it and the
38:24
owners of these products have to understand what do you want your
38:28
product to have longevity or do you want to just make the most
38:31
powerful thing that then just blows everything out?
38:34
And that's a a reality that they have to come to grips with.
38:38
Fortunately on the regulatory side and then, and this just
38:41
popped up in my my feed just now Open AI is coming out with a new
38:44
product called Skynet and supposed to, to be a perfect
38:46
regulatory process. And on that fine note, I I think
38:53
we should round off with you brought up the AI documentary,
38:57
which I will let you pronounce this time.
39:01
Apocalyptimist. Hopefully we'll check it out in
39:04
the next couple of weeks and report back on what it's
39:07
predicting. And what's interesting about
39:09
that documentary is that they did interview Sam Altman and the
39:14
CEO of Anthropic and a lot of people who are very much
39:17
involved in this industry. So we'll see what they have to
39:21
say. Very cool.
39:23
Anything else you think we should cover today?
39:25
No, I think we covered a tremendous amount of stuff in a
39:29
good amount of time. I would agree with you.
39:33
So thanks everyone for joining us.
39:35
Thanks Sharon for being being on as always.
39:39
Thanks for having me. Yeah, if you have anything that
39:42
you want to bring to our attention or you want us to
39:44
cover, e-mail us at podcast at Coopers dot law.
39:48
And for all of you doing justice out there, happy Hunter.