Techgrumps 3.42:: Three thefts don't make a right

WEBVTT
00:00:00.000 --> 00:00:03.540
Welcome to Tech Grumps 3 .42. Tonight is Sunday
00:00:03.540 --> 00:00:07.240
the 12th of July and once again gathered to grump
00:00:07.240 --> 00:00:13.419
at you are Ian. Hello, I am Ian and I am recording
00:00:13.419 --> 00:00:18.000
from Manchester where I live. David. Hello, I'm
00:00:18.000 --> 00:00:20.440
David Eastman. I'm very hot in London recording
00:00:20.440 --> 00:00:25.589
where I live. Wendy. I'm Wendy Grossman. I'm
00:00:25.589 --> 00:00:29.410
also based in London, and my website is at pelicancrossing
00:00:29.410 --> 00:00:33.890
.net. And myself, Ryan Alexander, also based
00:00:33.890 --> 00:00:36.490
in London, desperately trying to beat the heat.
00:00:40.560 --> 00:00:44.039
Okay, and as has become our unfortunate tradition,
00:00:44.259 --> 00:00:47.579
we shall start off Tech Grumps with the AI bucket
00:00:47.579 --> 00:00:50.820
of fun. So we have a collection of stories about
00:00:50.820 --> 00:00:53.259
AI that we've gathered for your edification that
00:00:53.259 --> 00:00:54.619
we're just going to all talk about. I'm going
00:00:54.619 --> 00:00:56.899
to read off the stories and everyone can just
00:00:56.899 --> 00:00:58.880
jump in after I've told about all the different
00:00:58.880 --> 00:01:01.299
stories and draw whatever points they care to
00:01:01.299 --> 00:01:05.799
draw. We have a Munich court decision on AI overviews
00:01:05.799 --> 00:01:09.120
from Google. We have an article about YouTube
00:01:09.120 --> 00:01:12.719
being cooked. people using AI to generate their
00:01:12.719 --> 00:01:17.099
own videos, also without permission. We have
00:01:17.099 --> 00:01:20.879
SpaceX going public on June 12th. I'm sure none
00:01:20.879 --> 00:01:24.680
of us heard about that. The opening price, $135.
00:01:24.959 --> 00:01:28.620
Lots of details on how that went. We have code
00:01:28.620 --> 00:01:32.579
reveals. We have a story about smart glasses,
00:01:32.760 --> 00:01:35.459
meta smart glasses that can use face print tracking.
00:01:35.780 --> 00:01:38.909
So now we have... Facial recognition everywhere.
00:01:39.049 --> 00:01:41.810
Fantastic. We've got the last stakeholder gives
00:01:41.810 --> 00:01:44.730
up on gigantic data center in Northern Virginia
00:01:44.730 --> 00:01:47.090
after years of litigation over land rights. So
00:01:47.090 --> 00:01:49.590
a little bit of, you know, a refreshing moment
00:01:49.590 --> 00:01:51.870
of nice news in the middle of the apocalypse.
00:01:52.209 --> 00:01:55.709
We have another story on AI cheating at a scandal
00:01:55.709 --> 00:01:58.469
at Brown University. We have a story on Waymo's
00:01:58.469 --> 00:02:01.609
running out of power in San Francisco on the
00:02:01.609 --> 00:02:04.549
4th of July congestion. And last but not least,
00:02:04.609 --> 00:02:07.170
we have a story about protests against data centers
00:02:07.170 --> 00:02:10.909
for high -frequency trading in Brick Lane, our
00:02:10.909 --> 00:02:15.229
own hometown of London. Who wants to launch off
00:02:15.229 --> 00:02:17.550
first? Can I just say, it's not my hometown of
00:02:17.550 --> 00:02:20.889
London, but I get the point. So I want to jump
00:02:20.889 --> 00:02:25.710
in on this. Okay, so the YouTube is cooked. I
00:02:25.710 --> 00:02:31.340
understand. So I will post. A bunch of links.
00:02:31.560 --> 00:02:33.139
But I get that people are like, well, YouTube
00:02:33.139 --> 00:02:38.680
is just screwed. But I think they don't understand
00:02:38.680 --> 00:02:44.259
how bad it is. Literally, there are scrapers
00:02:44.259 --> 00:02:48.620
who are taking videos and generating it as fast
00:02:48.620 --> 00:02:52.219
as they can to kind of chime in on the clicks.
00:02:52.500 --> 00:02:55.280
And it is getting ridiculous. I understand that
00:02:55.280 --> 00:02:59.710
lots of... kind of content creators or i'll say
00:02:59.710 --> 00:03:03.030
producers are are just kind of leaving youtube
00:03:03.030 --> 00:03:06.270
and it is a pretty serious problem and i and
00:03:06.270 --> 00:03:10.169
i'm as someone who's in the fediverse and that
00:03:10.169 --> 00:03:13.189
kind of thing i do think it's i don't really
00:03:13.189 --> 00:03:16.210
i do care about youtube because i do watch a
00:03:16.210 --> 00:03:20.969
lot too but i just think this bad news full stop
00:03:20.969 --> 00:03:25.449
for A lot of people who do create videos and
00:03:25.449 --> 00:03:28.810
even people who create audio podcasts. YouTube
00:03:28.810 --> 00:03:32.189
and Google have made a lot of efforts to try
00:03:32.189 --> 00:03:36.169
and get people to create stuff for YouTube, which
00:03:36.169 --> 00:03:39.129
were podcasts. So I wouldn't be surprised if
00:03:39.129 --> 00:03:41.229
someone's listening to this and it's actually
00:03:41.229 --> 00:03:45.310
being generated by or regenerated by a large
00:03:45.310 --> 00:03:48.520
language model. And it's not us. Is it a problem
00:03:48.520 --> 00:03:50.759
for the BBC as well? Because they're talking
00:03:50.759 --> 00:03:54.039
about putting more stuff on social media. Yes,
00:03:54.039 --> 00:04:00.319
I would suggest yes, it would be. I think it's
00:04:00.319 --> 00:04:03.120
also Channel 4 and others who have kind of started
00:04:03.120 --> 00:04:07.319
to use YouTube as their means of delivering content
00:04:07.319 --> 00:04:12.300
beyond transmission. And so I do feel that they
00:04:12.300 --> 00:04:15.479
are also going to get run straight into this
00:04:15.479 --> 00:04:20.829
brick wall. interesting my brain just went oh
00:04:20.829 --> 00:04:24.209
it's a flash mob but it's made up of video content
00:04:24.209 --> 00:04:30.389
like it it's it's an interesting pattern similar
00:04:30.389 --> 00:04:32.569
to flash mobs it's just that people are going
00:04:32.569 --> 00:04:34.110
oh is this the thing that everyone's talking
00:04:34.110 --> 00:04:37.310
about here's 20 like i can i can spin up 20 000
00:04:37.310 --> 00:04:39.470
videos about it and all together they'll get
00:04:39.470 --> 00:04:44.670
enough clicks right yeah yeah yeah And that's
00:04:44.670 --> 00:04:47.269
just for people who have the pure motivation
00:04:47.269 --> 00:04:53.069
of just wanting to click from and get money from
00:04:53.069 --> 00:04:56.129
the clicks. What if you have a bad actor who's
00:04:56.129 --> 00:04:59.009
just like, I would like for there to be a hundred
00:04:59.009 --> 00:05:01.269
versions of this story out there, only one of
00:05:01.269 --> 00:05:02.670
which is the one that I care about and everyone
00:05:02.670 --> 00:05:04.350
else is actually not that big of a deal to me,
00:05:04.410 --> 00:05:06.709
and I have enough money that I can put out a
00:05:06.709 --> 00:05:10.009
hundred thousand. I think that's a really good
00:05:10.009 --> 00:05:14.939
point because... I know, was it Hillary when
00:05:14.939 --> 00:05:16.920
they slowed down her clip so it sounded like
00:05:16.920 --> 00:05:21.699
she was stirring her words? Yeah, they made her
00:05:21.699 --> 00:05:25.519
sound like she was drunk. Yeah, and I, you know,
00:05:25.560 --> 00:05:28.860
it wouldn't take much or, once again, you try
00:05:28.860 --> 00:05:32.620
out loads of them and it's crowded and then those
00:05:32.620 --> 00:05:35.259
have got things that are slowed down deliberately
00:05:35.259 --> 00:05:39.060
and that's a realistic thing that might happen.
00:05:39.379 --> 00:05:42.750
Anyway. That's one I definitely want to dig into.
00:05:45.670 --> 00:05:56.490
Other ones. SpaceX going public. Well, you know,
00:05:56.509 --> 00:05:59.870
the bad thing there is that they've changed the
00:05:59.870 --> 00:06:03.970
rules sort of for SpaceX benefit. And so it's
00:06:03.970 --> 00:06:07.509
being added. It was added to one of the indexes
00:06:07.509 --> 00:06:09.889
much earlier than it should have been. And the
00:06:09.889 --> 00:06:12.370
insiders are being allowed to sell shares earlier
00:06:12.370 --> 00:06:16.310
than they would have been normally. And, you
00:06:16.310 --> 00:06:20.769
know, basically all decisions favoring Elon Musk
00:06:20.769 --> 00:06:23.930
over ordinary investors. There was a similar
00:06:23.930 --> 00:06:26.009
story that I should have put in here, but then
00:06:26.009 --> 00:06:28.329
I don't remember it just now. There was a story
00:06:28.329 --> 00:06:31.769
that went past about apparently the. the collision
00:06:31.769 --> 00:06:36.389
of the housing market in san francisco with the
00:06:36.389 --> 00:06:38.850
privileged few who are having pre -access to
00:06:38.850 --> 00:06:42.050
anthropic and open ai shares and there are people
00:06:42.050 --> 00:06:44.610
who are putting their houses on the market but
00:06:44.610 --> 00:06:47.449
only if you are going to give them anthropic
00:06:47.449 --> 00:06:53.290
shares for them huh that's crazy yeah it is great
00:06:53.290 --> 00:06:55.889
well remember there was a period when people
00:06:55.889 --> 00:06:58.149
were taking out mortgages to buy bitcoin too
00:06:58.889 --> 00:07:01.850
I mean, this is a version of that. What? Yeah.
00:07:03.750 --> 00:07:05.430
Never heard that. Oh, my gosh. Yeah, they're
00:07:05.430 --> 00:07:09.230
all rich. No, probably not. Probably some of
00:07:09.230 --> 00:07:11.589
them lost their houses. Yeah, it depends what
00:07:11.589 --> 00:07:13.410
they did. Bitcoin goes up and down a lot, and
00:07:13.410 --> 00:07:16.029
if you bought at the wrong moment, you'd... Oh,
00:07:16.029 --> 00:07:19.930
yeah. It's the timing. I never heard... Okay.
00:07:20.490 --> 00:07:24.189
That's crazy. It's an interesting idea that...
00:07:24.189 --> 00:07:31.160
My thought is... we're edge like it's as if bitcoin
00:07:31.160 --> 00:07:33.199
wasn't bad enough we're edging into like the
00:07:33.199 --> 00:07:35.899
problem with bitcoin is that anybody could buy
00:07:35.899 --> 00:07:38.259
it if they had a computer and the right money
00:07:38.259 --> 00:07:40.819
to put in now we have things like anthropic shares
00:07:40.819 --> 00:07:43.040
and these shares which people are starting to
00:07:43.040 --> 00:07:47.860
do trading on and it's like money but only very
00:07:47.860 --> 00:07:52.279
special people get to have it yeah i was about
00:07:52.279 --> 00:07:55.980
10 years ago i think i was at a a conference
00:07:56.860 --> 00:08:02.300
That year was being held at Stanford. And one
00:08:02.300 --> 00:08:04.720
of the business professors said to me that we
00:08:04.720 --> 00:08:07.319
were reaching a point where ordinary investors
00:08:07.319 --> 00:08:09.939
were not going to be able to make serious money
00:08:09.939 --> 00:08:15.600
by buying Amazon when it goes public, which quite
00:08:15.600 --> 00:08:19.540
a few people did, because the insiders were keeping
00:08:19.540 --> 00:08:23.379
the shares to themselves for much longer. And
00:08:23.379 --> 00:08:25.439
so when they finally went public, all the big...
00:08:25.759 --> 00:08:30.399
growth had already happened now nvidia has happened
00:08:30.399 --> 00:08:32.240
since then and you could you could have made
00:08:32.240 --> 00:08:34.500
quite a bit of an off nvidia if you'd bought
00:08:34.500 --> 00:08:40.919
it in 2018 or 2019 but in general you know he's
00:08:40.919 --> 00:08:44.360
not every time i see look at every time i see
00:08:44.360 --> 00:08:47.159
an ipo and i see when they went public as compared
00:08:47.159 --> 00:08:49.100
to when they would have gone public say 20 years
00:08:49.100 --> 00:08:52.179
ago you know it's really noticeable that that
00:08:52.179 --> 00:08:55.750
a lot of the profits are being kept off the market.
00:08:56.269 --> 00:09:00.809
And private equity makes that worse. I think
00:09:00.809 --> 00:09:02.429
I can do a clever segue into some of the other
00:09:02.429 --> 00:09:05.809
points about this, because we've got out of the
00:09:05.809 --> 00:09:08.590
little bucket that we have here, there's a common
00:09:08.590 --> 00:09:10.990
thread between some of them, which is the Munich
00:09:10.990 --> 00:09:16.549
court decision, the smart glasses, the cheating
00:09:16.549 --> 00:09:20.129
scandal, but also like there's this thread of
00:09:20.129 --> 00:09:25.360
holding the companies responsible. So that's
00:09:25.360 --> 00:09:27.139
one of the things that I see. I keep thinking
00:09:27.139 --> 00:09:29.240
to myself, in a lot of these cases, these valuations
00:09:29.240 --> 00:09:31.200
are just waiting for one really, really viable
00:09:31.200 --> 00:09:33.299
class action lawsuit to make the stocks tank.
00:09:35.759 --> 00:09:40.860
Like with Waymo's running out of power. They're
00:09:40.860 --> 00:09:41.940
getting with it because they're kind of going,
00:09:42.019 --> 00:09:44.659
you know. Oh, but it's still somebody else's
00:09:44.659 --> 00:09:46.620
fault if something goes wrong. Actually, Waymo
00:09:46.620 --> 00:09:48.980
has had a lot of news stories in the last month
00:09:48.980 --> 00:09:52.059
or two. One of them, they've recalled 3 ,900
00:09:52.059 --> 00:09:54.279
cars because they kept driving into construction
00:09:54.279 --> 00:09:57.480
zones. There have been issues with them blocking
00:09:57.480 --> 00:10:02.120
fire and ambulance people. Tesla's got problems,
00:10:02.279 --> 00:10:05.580
too. People keep turning on full self -drive,
00:10:05.720 --> 00:10:10.200
and then one car drove at 100. miles an hour
00:10:10.200 --> 00:10:12.519
or something into a house and killed a 76 year
00:10:12.519 --> 00:10:15.279
old woman have you heard about the hack to get
00:10:15.279 --> 00:10:17.740
over the you must be at the wheel self -drive
00:10:17.740 --> 00:10:21.460
no i i did hear about some guy who was driving
00:10:21.460 --> 00:10:23.519
along the highway with kids in the back of the
00:10:23.519 --> 00:10:27.659
car and had gone to sleep no so the there's this
00:10:27.659 --> 00:10:31.159
and i hope you can hear the quotation marks that
00:10:31.159 --> 00:10:33.259
i'm making with my fingers safety feature on
00:10:33.259 --> 00:10:36.940
the teslas which is that Even when you're in
00:10:36.940 --> 00:10:39.360
self -drive mode, you have to be present at the
00:10:39.360 --> 00:10:43.100
wheel. And so what people have done is they've
00:10:43.100 --> 00:10:45.320
taken little passport photos of themselves and
00:10:45.320 --> 00:10:47.759
dangled them on a little wire in front of the
00:10:47.759 --> 00:10:50.120
internal camera, which is enough to fool them.
00:10:50.539 --> 00:10:56.399
Yep. Well, oh, my God. Like I say, you know,
00:10:56.399 --> 00:10:58.500
if you don't want to drive, you know, lobby for
00:10:58.500 --> 00:11:02.279
better public transport. Oh, and have to ride
00:11:02.279 --> 00:11:06.330
with the hoi polloi. When I have money, why would
00:11:06.330 --> 00:11:09.049
I do that? I don't actually think that's, I'm
00:11:09.049 --> 00:11:12.309
not sure that that's really the issue. I mean,
00:11:12.309 --> 00:11:14.590
a lot of people in the States where these things
00:11:14.590 --> 00:11:16.590
are happening don't really have a lot of choice.
00:11:16.710 --> 00:11:18.850
Oh, but Waymo Austin was in trouble in London
00:11:18.850 --> 00:11:21.029
because apparently they've been running test
00:11:21.029 --> 00:11:23.990
cars around East London. Oh, yeah, I've seen
00:11:23.990 --> 00:11:27.870
them. There was a story about this series of
00:11:27.870 --> 00:11:30.250
cars that got stuck in a small cul -de -sac.
00:11:30.960 --> 00:11:32.679
And it's like 3 o 'clock in the morning, and
00:11:32.679 --> 00:11:34.259
they make all these noises because they're having
00:11:34.259 --> 00:11:37.139
to back up to try to get out. And it was waking
00:11:37.139 --> 00:11:41.860
up all the residents. Yeah. Speaking of problems
00:11:41.860 --> 00:11:45.299
in London, so there's a story in here about the
00:11:45.299 --> 00:11:51.480
data center in Brick Lane. And so I know the
00:11:51.480 --> 00:11:53.159
location. Actually, David, I think we've been
00:11:53.159 --> 00:11:54.860
in that location. They're looking at – it's the
00:11:54.860 --> 00:11:57.519
Truman Brewery, the old Truman Brewery. Oh, yeah.
00:11:57.759 --> 00:12:01.100
Everyone's been there, right? Yeah. Don't get
00:12:01.100 --> 00:12:02.539
me wrong, it's a nice space, but I'm looking
00:12:02.539 --> 00:12:05.320
at that going, fucking data center? That's way
00:12:05.320 --> 00:12:08.120
too small. Because the stories I'm hearing about
00:12:08.120 --> 00:12:11.059
data centers is that they are enormous. And I'm
00:12:11.059 --> 00:12:14.320
going, I mean, basically my brain's going, what
00:12:14.320 --> 00:12:17.620
would you need that is unique to make the Truman
00:12:17.620 --> 00:12:21.220
Center, like the Truman Brewery, a data center?
00:12:21.679 --> 00:12:24.360
Is it just you'd need ridiculously extensive
00:12:24.360 --> 00:12:27.139
power supplies and internet going in and out
00:12:27.139 --> 00:12:30.190
of it? Well, as I understand it, the purpose
00:12:30.190 --> 00:12:33.570
is to enable high, to help high frequency trading.
00:12:34.669 --> 00:12:38.509
So apparently the location matters. Let me unmute.
00:12:38.710 --> 00:12:40.429
Yeah, I mean, that's exactly, it's just high
00:12:40.429 --> 00:12:41.909
frequency trading. That's the only interesting
00:12:41.909 --> 00:12:43.590
it being anywhere near the centre of London.
00:12:44.889 --> 00:12:47.409
I mean, obviously the size is important because
00:12:47.409 --> 00:12:49.330
otherwise the heating becomes a problem. So there's
00:12:49.330 --> 00:12:52.269
always a concern about separating out the surface.
00:12:52.370 --> 00:12:54.590
Otherwise the general temperature just rises.
00:12:55.289 --> 00:12:59.240
Hence you need the size. I mean, the physical
00:12:59.240 --> 00:13:03.320
area, sorry, not necessarily the size. How many
00:13:03.320 --> 00:13:05.000
neighbors would like to be more baked? I mean,
00:13:05.019 --> 00:13:10.600
there are some famous bakeries. Yeah. Yeah. Or
00:13:10.600 --> 00:13:12.700
have a swimming pool. I mean, that's the other
00:13:12.700 --> 00:13:17.240
thing, you know. Well, what the council and the
00:13:17.240 --> 00:13:20.519
local residents want is affordable housing. Yeah.
00:13:20.940 --> 00:13:23.519
Well, everybody wants that. We're not going to
00:13:23.519 --> 00:13:26.139
get it. Well, I don't know what it is about London,
00:13:26.220 --> 00:13:30.990
but I... I'm astonished that anyone at this point
00:13:30.990 --> 00:13:34.730
believes when anybody says, especially in London,
00:13:34.850 --> 00:13:36.370
that they're going to build affordable housing
00:13:36.370 --> 00:13:39.450
because time and time again, it's like there's
00:13:39.450 --> 00:13:42.809
got to be a couple of consulting firms out there
00:13:42.809 --> 00:13:45.909
whose entire business is once you have sold your
00:13:45.909 --> 00:13:48.169
building on the basis of affordable public housing,
00:13:48.289 --> 00:13:50.809
we show you all the ways to get out of doing
00:13:50.809 --> 00:13:54.820
it. There definitely are, yeah. Actually, the
00:13:54.820 --> 00:13:57.820
strange one is, given the new Prime Minister
00:13:57.820 --> 00:14:00.659
Andy Burden, they're looking at the famous Manchester
00:14:00.659 --> 00:14:02.399
model, Manchesterism, which actually involves
00:14:02.399 --> 00:14:06.299
doing lots of buildings, famously, normally without
00:14:06.299 --> 00:14:07.779
any social housing. In fact, they've recently
00:14:07.779 --> 00:14:10.740
said if we had to include the numbers of social
00:14:10.740 --> 00:14:16.139
or, you know, achievably buying, buyable houses,
00:14:16.220 --> 00:14:18.419
I mean, that's the best you could put it, they
00:14:18.419 --> 00:14:20.519
wouldn't have built the amount of houses they
00:14:20.519 --> 00:14:23.159
had done. So we're not going to get anywhere
00:14:23.159 --> 00:14:32.409
near it. That's just not happening. There's so
00:14:32.409 --> 00:14:36.889
much I want to say about Andy Burnham, but I
00:14:36.889 --> 00:14:42.649
will leave it for another show. A political podcast,
00:14:42.950 --> 00:14:46.370
perhaps. That's a good idea. We should have political
00:14:46.370 --> 00:14:52.090
grumps or something. Didn't face special? There
00:14:52.090 --> 00:14:54.570
have been so many bin face podcasts. It's just
00:14:54.570 --> 00:14:58.710
been amazing. And he's been on all of them. Yeah,
00:14:58.750 --> 00:15:01.169
yeah. He's been on quite a lot of them. He's
00:15:01.169 --> 00:15:04.490
done quite well. But they haven't been like specially
00:15:04.490 --> 00:15:06.389
made ones. He's just been on all the podcasts
00:15:06.389 --> 00:15:16.669
and they're all giving me life. Recycling. and
00:15:16.669 --> 00:15:19.649
that's your ai roundup up next we have a story
00:15:19.649 --> 00:15:22.870
that wendy brought to us from rest of world on
00:15:22.870 --> 00:15:26.690
ai analytics being used at the world cup this
00:15:26.690 --> 00:15:28.909
is kind of ironic because i'm actually a football
00:15:28.909 --> 00:15:31.649
free zone i have no idea who's playing or who
00:15:31.649 --> 00:15:33.629
is still in the competition and i don't care
00:15:33.629 --> 00:15:36.830
but this story was interesting to me because
00:15:36.830 --> 00:15:40.590
so much of ai is humans all the way down and
00:15:40.590 --> 00:15:44.559
it turns out that football is no exception Data
00:15:44.559 --> 00:15:46.860
analytics is really important in football. Who
00:15:46.860 --> 00:15:52.899
knew? And, you know, every part, it seems like,
00:15:52.960 --> 00:15:56.419
according to Rest of World, the current edition
00:15:56.419 --> 00:15:58.860
of the FIFA World Cup features a sensor fitted
00:15:58.860 --> 00:16:01.840
ball, real time tracking, artificial intelligence
00:16:01.840 --> 00:16:05.399
assisted offside calls, and an AI assistant for
00:16:05.399 --> 00:16:09.429
each of the 48 teams. And behind these innovations,
00:16:09.649 --> 00:16:11.929
and this is the part that really drew my attention,
00:16:12.070 --> 00:16:14.529
are data workers in countries including India,
00:16:14.690 --> 00:16:16.570
Cambodia, and the Philippines, who are essential
00:16:16.570 --> 00:16:19.649
for the many AI tools in play. In other words,
00:16:19.669 --> 00:16:22.029
what we've got is a bunch of workers all over
00:16:22.029 --> 00:16:24.509
the global south who are doing things like labeling
00:16:24.509 --> 00:16:28.730
video clips and labeling data and figuring out,
00:16:28.750 --> 00:16:31.049
you know, doing analytics and all the rest of
00:16:31.049 --> 00:16:36.580
it. You know, it's just very typical that what
00:16:36.580 --> 00:16:39.860
they claim is AI always turns out to be a bunch
00:16:39.860 --> 00:16:43.019
of humans in the background, you know, doing
00:16:43.019 --> 00:16:46.000
all sort of all the work. The good news is all
00:16:46.000 --> 00:16:47.840
those out -of -work Amazon Fresh people have
00:16:47.840 --> 00:16:51.840
found a new venue. That's right. Amazon is discontinuing
00:16:51.840 --> 00:16:56.980
Mechanical Turk, isn't it? Mechanical Turk. Yeah,
00:16:56.980 --> 00:16:59.360
well, there's a lot of there is an awful lot.
00:16:59.399 --> 00:17:01.700
I mean, I got even I was a bit surprised when
00:17:01.700 --> 00:17:04.059
they were sort of American statistics. Somebody
00:17:04.059 --> 00:17:06.900
had done a free kick and they'd said the distance
00:17:06.900 --> 00:17:10.579
of the distance of the speed of the ball, which
00:17:10.579 --> 00:17:12.480
you can get because there's a detector on the
00:17:12.480 --> 00:17:15.099
ball, but also the spin on the ball. And you
00:17:15.099 --> 00:17:17.660
definitely you definitely, of course, need quite
00:17:17.660 --> 00:17:21.279
a careful, you know, you need the two senses
00:17:21.279 --> 00:17:23.599
on it and get the spin. But the other thing is.
00:17:24.170 --> 00:17:26.269
Nobody playing football gives a hoot about the
00:17:26.269 --> 00:17:28.710
spin on the ball. It sounds a bit American, but
00:17:28.710 --> 00:17:31.750
to be honest, there's an awful lot. I doubt that
00:17:31.750 --> 00:17:33.930
that's true. I doubt that that's true because
00:17:33.930 --> 00:17:36.390
at least in tennis, if the ball has spin on it,
00:17:36.410 --> 00:17:39.250
it changes how it bounces. Well, I mean, the
00:17:39.250 --> 00:17:41.309
reason that you would look at the spin is that
00:17:41.309 --> 00:17:43.170
like, yeah, so you can swerve the ball, but you
00:17:43.170 --> 00:17:46.109
can physically see it. I mean, and it's like,
00:17:46.150 --> 00:17:48.069
it's good to know that you can do it, but unless
00:17:48.069 --> 00:17:49.369
you can do it, you're not going to be at the
00:17:49.369 --> 00:17:53.319
World Cup taking a free kick. But the data is
00:17:53.319 --> 00:17:56.019
not just going to players. The analytics go to
00:17:56.019 --> 00:17:58.619
broadcasters and commentators and so forth. It
00:17:58.619 --> 00:18:01.240
doesn't go to players at all. It's broadcasted,
00:18:01.240 --> 00:18:03.140
but most of it goes to the betting industry.
00:18:03.539 --> 00:18:07.259
I was about to say that. Yeah. Okay, if you go
00:18:07.259 --> 00:18:09.539
into any of the betting shops now and say, oh,
00:18:09.619 --> 00:18:12.920
I want to bet on a match, it looks like suddenly
00:18:12.920 --> 00:18:14.920
you're running a bank. The amount of statistics,
00:18:15.400 --> 00:18:17.319
the amount of things you can bet on, you know,
00:18:17.339 --> 00:18:19.650
it used to be. correct score and half -time,
00:18:19.690 --> 00:18:23.109
full -time results. Now there's a list that goes
00:18:23.109 --> 00:18:25.109
off. I mean, I normally bet during the World
00:18:25.109 --> 00:18:27.490
Cup and I never bet online because I just know
00:18:27.490 --> 00:18:29.470
I'm never going to do that. But if you then just
00:18:29.470 --> 00:18:33.109
go into a... It'll be the same online. Go into
00:18:33.109 --> 00:18:36.190
a bookie and then have a look at the list and
00:18:36.190 --> 00:18:38.490
it's off. Literally, you have to scroll through
00:18:38.490 --> 00:18:41.089
it because it's just basically data. They make
00:18:41.089 --> 00:18:42.930
you think you're running a bank by having all
00:18:42.930 --> 00:18:45.450
the different statistics you can bet on. And
00:18:45.450 --> 00:18:48.900
that... our betting as for what you know is happening
00:18:48.900 --> 00:18:51.880
in the states of calci and so forth and there
00:18:51.880 --> 00:18:54.839
and there's stuff god knows because i even want
00:18:54.839 --> 00:18:59.660
to go there what's really disturbing is during
00:18:59.660 --> 00:19:03.119
the french open tennis calci was actually a sponsor
00:19:03.119 --> 00:19:09.259
of the broadcast oh yeah quoting statistics and
00:19:09.259 --> 00:19:11.519
they had crawls all the time showing their calci's
00:19:11.519 --> 00:19:17.170
odds I wonder at the role that the U .S. influence
00:19:17.170 --> 00:19:19.450
has on this, because I can think of a very I
00:19:19.450 --> 00:19:21.609
can think of a straightforward reason why they
00:19:21.609 --> 00:19:23.849
want to be spamming all those stats all the time.
00:19:23.910 --> 00:19:28.089
And it's the 24 -7 news culture over there. They
00:19:28.089 --> 00:19:30.809
can't stand a silent moment and they want to
00:19:30.809 --> 00:19:33.210
just be spewing data at you all the time. No,
00:19:33.329 --> 00:19:36.549
Bet365 has been going in Europe and tennis for
00:19:36.549 --> 00:19:40.750
a long time now. Fair. There's a load of YouTube
00:19:40.750 --> 00:19:43.930
channels that just do betting. I mean, there
00:19:43.930 --> 00:19:45.869
must have been. I mean, I was sort of interested
00:19:45.869 --> 00:19:48.029
in one, and then as YouTube showed me, about
00:19:48.029 --> 00:19:52.210
100. Most of them are American. They have a completely
00:19:52.210 --> 00:19:54.450
different statistical range that we don't do.
00:19:54.490 --> 00:19:56.430
They do this Asian handicap system that you'll
00:19:56.430 --> 00:19:58.069
hear in other sports that doesn't necessarily
00:19:58.069 --> 00:20:00.309
match football. I wonder if they've actually
00:20:00.309 --> 00:20:02.890
thought about, when they're sending all the stuff
00:20:02.890 --> 00:20:04.589
out to the Global South, I wonder if anyone has
00:20:04.589 --> 00:20:06.750
actually thought about the fact that the Global
00:20:06.750 --> 00:20:11.210
South also gives a shit about football. For some
00:20:11.210 --> 00:20:13.829
reason, the Indian AI data centers have been
00:20:13.829 --> 00:20:18.170
returning nothing but lack of penalties on any
00:20:18.170 --> 00:20:21.369
call that relates to the Indian team. In fact,
00:20:21.470 --> 00:20:26.269
the data labelers themselves, they're often quite
00:20:26.269 --> 00:20:28.529
passionate fans. It says this in the article,
00:20:28.630 --> 00:20:33.470
and many of them also play football. I mean,
00:20:33.490 --> 00:20:36.809
this has been true for a long time. I mean, IBM
00:20:36.809 --> 00:20:39.099
started doing statistics. at wimbledon in the
00:20:39.099 --> 00:20:42.519
early 90s and they have somebody a court side
00:20:42.519 --> 00:20:44.539
you know sort of marking down all this stuff
00:20:44.539 --> 00:20:47.460
about each point and and those guys have to be
00:20:47.460 --> 00:20:50.279
like county level tennis players or or better
00:20:50.279 --> 00:20:53.220
yeah you can certainly if you're doing labeling
00:20:53.220 --> 00:20:55.420
you need to understand the statistics because
00:20:55.420 --> 00:20:57.000
they are i mean the football related one the
00:20:57.000 --> 00:20:59.980
football related is like successful tackles six
00:20:59.980 --> 00:21:02.880
said successful counter pressure which you actually
00:21:02.880 --> 00:21:04.380
have to think about what that actually means
00:21:04.380 --> 00:21:07.700
for it you know a doll is easy enough but think
00:21:07.700 --> 00:21:09.960
how quickly did you stop the opposition getting
00:21:09.960 --> 00:21:11.740
the ball when you just gave it away that requires
00:21:11.740 --> 00:21:13.259
somebody who actually understands football that
00:21:13.259 --> 00:21:16.160
stat is stats are taken for each match of which
00:21:16.160 --> 00:21:19.140
there are an awful lot yeah the equivalent in
00:21:19.140 --> 00:21:21.140
tennis is being able to tell the difference between
00:21:21.140 --> 00:21:24.279
a forced error and an unforced error yes yes
00:21:24.839 --> 00:21:28.759
Exactly. Yes. Yes. Missed referee mistakes or
00:21:28.759 --> 00:21:30.680
so recorded, which is quite interesting because
00:21:30.680 --> 00:21:32.579
you can have a there's different books on referee
00:21:32.579 --> 00:21:35.880
errors. And this was before Trump decided what
00:21:35.880 --> 00:21:38.059
was and wasn't correct. You can still have people
00:21:38.059 --> 00:21:40.559
doing that. So, yes, there are a lot of statistics,
00:21:40.640 --> 00:21:43.619
but most of it ends up to be relevant specifically
00:21:43.619 --> 00:21:46.880
to gambling. But actually, the interesting thing
00:21:46.880 --> 00:21:49.160
is that legal sports betting is actually quite
00:21:49.160 --> 00:21:53.220
new in the States. Yes, I mean, that's why I
00:21:53.220 --> 00:21:56.140
was talking about the UK sports betting the league,
00:21:56.180 --> 00:21:57.779
which is the legal side, of course, and it's
00:21:57.779 --> 00:22:00.319
been around for some time. I wonder if we've
00:22:00.319 --> 00:22:05.880
had... We were talking about Kelshi and Paramutual
00:22:05.880 --> 00:22:09.880
and this other form of betting. I wonder if at
00:22:09.880 --> 00:22:12.960
this point... Prediction markets. I wonder if
00:22:12.960 --> 00:22:18.119
at this point I can see very easily the prediction
00:22:18.119 --> 00:22:22.279
markets suddenly becoming a anti -sports gambling
00:22:22.279 --> 00:22:26.119
advocate because they see it as muscling in on
00:22:26.119 --> 00:22:30.980
their turf. It's a little difficult because it's
00:22:30.980 --> 00:22:32.619
been there for a long time, but they might look
00:22:32.619 --> 00:22:36.319
at it like that. Certainly not in the UK, but
00:22:36.319 --> 00:22:38.059
in the US where there's always been a struggle
00:22:38.059 --> 00:22:40.279
for gambling and the ways they've gotten around
00:22:40.279 --> 00:22:45.960
it. Oh, that's an interesting thought. I was
00:22:45.960 --> 00:22:50.579
just going to... So there's a podcast I listen
00:22:50.579 --> 00:22:57.480
to by the FT called Tectonic, and it did actually
00:22:57.480 --> 00:23:01.359
cover this. But the interesting angle, and I
00:23:01.359 --> 00:23:03.960
know it's a little bit different from the story
00:23:03.960 --> 00:23:07.519
that Wendy originally brought up, but there's
00:23:07.519 --> 00:23:13.319
a real question of how much does this ruin sports
00:23:13.319 --> 00:23:17.190
for them? Or was it for... The fans who just
00:23:17.190 --> 00:23:20.190
like, like David, who just want to potentially
00:23:20.190 --> 00:23:22.269
put a bet, you know, they just want the kind
00:23:22.269 --> 00:23:24.509
of like, they just want the final score. They
00:23:24.509 --> 00:23:26.970
don't want all of these kind of, all of these
00:23:26.970 --> 00:23:30.609
statistics and stats to, because it's just, you
00:23:30.609 --> 00:23:34.450
know, like I've seen, especially in sort of the
00:23:34.450 --> 00:23:38.549
American coverage, you know, the kind of polymarket
00:23:38.549 --> 00:23:42.329
kind of couchy. kind of stuff comes up on about
00:23:42.329 --> 00:23:45.069
at the bottom of the screen but also just like
00:23:45.069 --> 00:23:48.789
you know how much spin all that stuff is isn't
00:23:48.789 --> 00:23:52.990
is it that interesting i mean i i get it no it's
00:23:52.990 --> 00:23:55.210
not interesting yeah it is exactly it's like
00:23:55.210 --> 00:23:59.569
how much does this actually take away um actually
00:23:59.569 --> 00:24:03.529
being in the game or like in the game even if
00:24:03.529 --> 00:24:05.970
you're watching the game i wonder because you
00:24:05.970 --> 00:24:09.509
know and i i look at I imagine this scenario
00:24:09.509 --> 00:24:11.730
where people are just in the game. They're on
00:24:11.730 --> 00:24:14.710
the front. I feel like basketball, right? You're
00:24:14.710 --> 00:24:19.509
talking about a subset of the people who enjoy
00:24:19.509 --> 00:24:22.690
the game who gain that enjoyment through betting
00:24:22.690 --> 00:24:25.730
on it. So first of all, it's really important
00:24:25.730 --> 00:24:28.690
to note that people who don't bet on these games
00:24:28.690 --> 00:24:34.319
are really not impacted. by everyone else in
00:24:34.319 --> 00:24:37.759
the game. I'm not even sure yet. Like, no more
00:24:37.759 --> 00:24:39.880
so than anything else. Now, we can talk about
00:24:39.880 --> 00:24:42.119
the experience of the people who are betting,
00:24:42.279 --> 00:24:45.299
who are like, oh, like, but here's the thing.
00:24:45.319 --> 00:24:48.359
At this point, this isn't even, like, maybe it's
00:24:48.359 --> 00:24:50.980
assisted by AA, but what's really happening is,
00:24:50.980 --> 00:24:54.019
for me, when I look at that, I think they're
00:24:54.019 --> 00:24:56.619
just chasing the long tail of the people who
00:24:56.619 --> 00:25:00.900
are already upset. so obsessively betting that
00:25:00.900 --> 00:25:02.839
they will be encouraged to bet more with more
00:25:02.839 --> 00:25:06.680
statistics and maybe like maybe there's a small
00:25:06.680 --> 00:25:08.279
fraction of people who'd be like oh it's just
00:25:08.279 --> 00:25:10.720
not fun anymore now that i get 14 pages and don't
00:25:10.720 --> 00:25:12.859
just they won they lost i don't think anyone's
00:25:12.859 --> 00:25:14.740
stopping them from doing bets on they won they
00:25:14.740 --> 00:25:16.960
lost or whatever the spread is oh no of course
00:25:16.960 --> 00:25:19.299
it's just more it's not yeah it's just it's just
00:25:19.299 --> 00:25:24.720
more in tennis it actually has been quite damaging
00:25:24.720 --> 00:25:28.670
because there are lots of very small tournaments
00:25:28.670 --> 00:25:32.410
where the players get paid very little and you
00:25:32.410 --> 00:25:34.769
can bet on all these crazy things like the next
00:25:34.769 --> 00:25:37.730
point whether somebody wins a set whether somebody
00:25:37.730 --> 00:25:40.630
breaks serve here and so forth and it has encouraged
00:25:40.630 --> 00:25:43.329
two things one is a lot of cheating at bets uh
00:25:43.329 --> 00:25:46.690
you know they basically had to ban laptops and
00:25:46.690 --> 00:25:49.410
and they're very funny about mobile phones because
00:25:49.410 --> 00:25:52.109
people can people can bet at courtside in that
00:25:52.109 --> 00:25:56.170
tiny split second between the ball ending the
00:25:56.170 --> 00:26:00.230
point and the umpire calling the score. And that's
00:26:00.230 --> 00:26:02.910
one aspect. But the other aspect is that these
00:26:02.910 --> 00:26:05.369
players who are paid very little are really,
00:26:05.369 --> 00:26:08.369
really vulnerable to being offered bribes to,
00:26:08.410 --> 00:26:12.849
to fix a match. Yeah. And that, that has been
00:26:12.849 --> 00:26:18.029
really, really bad. You could do the equivalent
00:26:18.029 --> 00:26:21.190
of high -speed trading on sporting events by
00:26:21.190 --> 00:26:23.529
setting up cameras with fast internet links and
00:26:23.529 --> 00:26:25.769
analytics to be like... Because when you said
00:26:25.769 --> 00:26:27.829
that, I was like, that sounds just like high
00:26:27.829 --> 00:26:30.630
-speed trading. Oh, crap. It is high -speed trading.
00:26:30.710 --> 00:26:35.750
Oh, yes. Yeah. Okay, so let me just take a step
00:26:35.750 --> 00:26:40.930
back. And I agree with Wendy. So the example
00:26:40.930 --> 00:26:44.329
they use is VAR. I know everyone gets upset about
00:26:44.329 --> 00:26:48.740
VAR, but... with var you get this even if you're
00:26:48.740 --> 00:26:52.420
there watching you out of four tap grumps do
00:26:52.420 --> 00:26:55.880
not give a shit about avr no no what is it video
00:26:55.880 --> 00:27:00.539
assisted refereeing so yeah sorry to know if
00:27:00.539 --> 00:27:03.240
it if it definitely so similar to like tennis
00:27:03.240 --> 00:27:06.319
where did it go in or did it not go in yes the
00:27:06.319 --> 00:27:08.940
referee the female i guess the difference between
00:27:08.940 --> 00:27:12.920
tennis and football the referee can be overruled
00:27:13.500 --> 00:27:16.799
by VAR, which I don't think that's true in Wimbledon
00:27:16.799 --> 00:27:19.500
or in tennis. It's the other way around in Wimbledon.
00:27:19.640 --> 00:27:22.700
Tennis has literally just changed to get rid
00:27:22.700 --> 00:27:27.819
of line judges. And so now it's just electronic
00:27:27.819 --> 00:27:32.720
line calling plus. Okay. Tennis is years ahead
00:27:32.720 --> 00:27:35.160
of football. I mean, football only recently,
00:27:35.420 --> 00:27:39.579
you know, football only recently. Oh, it's actually
00:27:39.579 --> 00:27:45.349
a goal, you know. Only now the referees can check
00:27:45.349 --> 00:27:48.170
that a goal was actually scored. Beforehand,
00:27:48.170 --> 00:27:51.230
we had the famous, the balls crossed the line,
00:27:51.369 --> 00:27:53.750
but the goalkeeper quickly padded it out and
00:27:53.750 --> 00:27:55.230
everyone knows it's a goal except the referee.
00:27:56.190 --> 00:27:58.869
So most sports are actually ahead of football
00:27:58.869 --> 00:28:02.089
in terms of technology. Okay, so the point I'm
00:28:02.089 --> 00:28:04.930
making, sorry. If you're there and you're just
00:28:04.930 --> 00:28:09.509
watching the game. Now, it's a bunch of stuff
00:28:09.509 --> 00:28:11.730
that comes up when they're waiting for the VAR
00:28:11.730 --> 00:28:14.650
decision about statistics and stuff like that,
00:28:14.789 --> 00:28:17.970
and they do come up, and then you're forced to...
00:28:17.970 --> 00:28:19.990
I've not been to a game, so maybe someone else
00:28:19.990 --> 00:28:22.289
is like, nope, that's not true, but there seems
00:28:22.289 --> 00:28:24.569
to be a bunch of things that do come up when
00:28:24.569 --> 00:28:27.670
you're waiting. When you say when you were there,
00:28:27.890 --> 00:28:29.890
do you mean when you're sitting in front of the
00:28:29.890 --> 00:28:31.930
telly watching it, or when you're at the game?
00:28:33.049 --> 00:28:37.829
I gather... when you're watching it on live okay
00:28:37.829 --> 00:28:40.150
that's probably an important thing to say at
00:28:40.150 --> 00:28:41.789
this point because when you're there usually
00:28:41.789 --> 00:28:44.369
means people who are at the stadium yeah okay
00:28:44.369 --> 00:28:46.970
so i also i think that the idea is that that
00:28:46.970 --> 00:28:49.670
on the big screens there are also statistics
00:28:49.670 --> 00:28:52.869
as well that come up yeah they're showing camera
00:28:52.869 --> 00:28:54.950
angles and statistics not necessarily the ones
00:28:54.950 --> 00:28:58.450
that the ar studio in wherever it is down the
00:28:58.450 --> 00:29:03.910
M4. What I'm trying to get to is that they're
00:29:03.910 --> 00:29:08.049
starting to creep into the experience of watching
00:29:08.049 --> 00:29:10.630
the sport that you like. That's what I'm trying
00:29:10.630 --> 00:29:13.490
to get at. And that's what the FT piece is about.
00:29:14.150 --> 00:29:15.769
Yeah, yeah, yeah. No, it's definitely a fact.
00:29:16.009 --> 00:29:20.369
I mean, the idea is we're going, we want an accurate
00:29:20.369 --> 00:29:22.450
result because we've known for years that...
00:29:22.799 --> 00:29:25.380
you know offside decisions are random but people
00:29:25.380 --> 00:29:27.579
are saying yeah but if a human doesn't make it
00:29:27.579 --> 00:29:29.319
then it's just going to be a bunch of humans
00:29:29.319 --> 00:29:32.180
in a computer and a long pause and now people
00:29:32.180 --> 00:29:34.460
are saying oh we thought vao would be great and
00:29:34.460 --> 00:29:35.819
then again no maybe it's not so good because
00:29:35.819 --> 00:29:37.279
you have to sit there three minutes while nothing
00:29:37.279 --> 00:29:40.460
happens that's like that's not a very it's not
00:29:40.460 --> 00:29:43.599
a very continuous sport is it well three minutes
00:29:43.599 --> 00:29:48.819
they call it a hydration break Some years ago,
00:29:48.900 --> 00:29:51.240
a couple of professors in Wales wrote a book
00:29:51.240 --> 00:29:53.440
about electronic line calling in tennis. And
00:29:53.440 --> 00:29:56.900
I think some technology in cricket as well. And
00:29:56.900 --> 00:29:59.039
the interesting thing about what they said was
00:29:59.039 --> 00:30:01.900
that it isn't that people really want accuracy.
00:30:02.119 --> 00:30:05.700
What they want is fairness. Yeah. That's what
00:30:05.700 --> 00:30:08.660
they say they want. Yes. Or they feel they want.
00:30:08.660 --> 00:30:11.200
That's what people get really exercised about.
00:30:13.289 --> 00:30:16.170
Yes, but not at the time. I think the idea is
00:30:16.170 --> 00:30:18.309
before and afterwards they say they want fairness.
00:30:18.410 --> 00:30:20.990
As you said, every sport you want fairness. But
00:30:20.990 --> 00:30:24.210
when it actually interferes with watching the
00:30:24.210 --> 00:30:26.630
sport, it does so. Even in cricket, it's annoying
00:30:26.630 --> 00:30:29.549
when they have to have a replay. So there's a
00:30:29.549 --> 00:30:32.329
Hawkeye. It's the same in cricket to test if
00:30:32.329 --> 00:30:35.109
it's leg before wicket. And that does annoy people.
00:30:35.289 --> 00:30:37.670
But before and afterwards, they say they want
00:30:37.670 --> 00:30:40.640
it to be fair until... You know, a human makes
00:30:40.640 --> 00:30:42.380
one decision, then the human has to stop. They
00:30:42.380 --> 00:30:45.099
have to ask, there's a cause, and then the thing
00:30:45.099 --> 00:30:46.859
gets reversed. Actually, it can be quite annoying,
00:30:47.039 --> 00:30:50.920
even if it is correct. And this is what we're
00:30:50.920 --> 00:30:56.819
finding. And next up on our hit parade is Ian
00:30:56.819 --> 00:30:59.799
talking about locally hosted AI systems. What
00:30:59.799 --> 00:31:04.940
do you got for us, Ian? I have been... I'm still
00:31:04.940 --> 00:31:09.410
kind of in between... I'm not a booster and I'm
00:31:09.410 --> 00:31:11.369
not a hater. You've brought the enemy into your
00:31:11.369 --> 00:31:14.789
house, Ian. How dare you? I have. I have. I have
00:31:14.789 --> 00:31:19.569
got it under in a cage. Right. So I've been thinking
00:31:19.569 --> 00:31:22.549
about this for a long, long while. And as someone
00:31:22.549 --> 00:31:26.930
who does host their own websites and other things,
00:31:27.190 --> 00:31:28.970
even though my peer tube is currently broken,
00:31:29.589 --> 00:31:33.269
I kind of started thinking, well, I've been hearing
00:31:33.269 --> 00:31:38.400
about local elements. And I decided I had a long
00:31:38.400 --> 00:31:42.480
list of things to do, including setting up a
00:31:42.480 --> 00:31:47.220
local LLM. So I set up one called, what's it
00:31:47.220 --> 00:31:50.019
called? I can't remember the name. We do need
00:31:50.019 --> 00:31:52.599
to know the model, Ian. We absolutely need to
00:31:52.599 --> 00:31:54.940
know the model. Well, not the model. I'm talking
00:31:54.940 --> 00:31:58.160
about the, oh yeah, it's called, right, the container,
00:31:58.359 --> 00:32:02.539
I guess, is called LM Studio, right? So I'm using
00:32:02.539 --> 00:32:05.220
that. I did use Ulama and stuff like that, and
00:32:05.220 --> 00:32:07.960
it didn't quite. work for me right so i i set
00:32:07.960 --> 00:32:11.039
up it was pretty straightforward um i'm running
00:32:11.039 --> 00:32:15.500
linux so it was pretty straight forward and from
00:32:15.500 --> 00:32:20.200
there you can then install whatever um whatever
00:32:20.200 --> 00:32:22.019
large language models you want it gives you a
00:32:22.019 --> 00:32:24.819
list and it also says it knows obviously what
00:32:24.819 --> 00:32:27.980
your machine is and what the capabilities of
00:32:27.980 --> 00:32:30.920
your machine are and so it kind of hey you can
00:32:30.920 --> 00:32:36.569
install the the gemma which is by google Yeah,
00:32:36.569 --> 00:32:38.650
it's one of the open -weight models, yeah. Yeah,
00:32:38.750 --> 00:32:41.329
so you could do that, but you probably don't
00:32:41.329 --> 00:32:44.170
want to install the, I don't know, I'll make
00:32:44.170 --> 00:32:49.430
up a number, the 4096 gigabyte version, you know,
00:32:49.450 --> 00:32:50.990
because that would be stupid because you haven't
00:32:50.990 --> 00:32:54.650
got enough memory for that, okay? But there's
00:32:54.650 --> 00:32:56.430
loads of other ones in between, and there's lots
00:32:56.430 --> 00:33:00.210
of small models. And so I was able to... i first
00:33:00.210 --> 00:33:01.930
i actually started using gemma i started using
00:33:01.930 --> 00:33:05.950
um quinn i started using the ibm one and it's
00:33:05.950 --> 00:33:07.609
just you just install them they took a while
00:33:07.609 --> 00:33:09.130
to download i've got a gigabit connection so
00:33:09.130 --> 00:33:12.089
it's not too bad and then you can start just
00:33:12.089 --> 00:33:14.910
talking to it and start getting to do things
00:33:14.910 --> 00:33:19.730
there are some limitations um honestly on my
00:33:19.730 --> 00:33:24.069
framework which to fair has got um how much gig
00:33:24.069 --> 00:33:29.170
have i got 40 no i got 64 64 gig of memory It
00:33:29.170 --> 00:33:31.809
wasn't even getting anywhere close to 32, some
00:33:31.809 --> 00:33:33.849
of the smaller models. And it was going quite
00:33:33.849 --> 00:33:37.210
quickly. And I'm not running like an NVIDIA card.
00:33:37.470 --> 00:33:39.890
I'm just using it off the CPU. I'm using the
00:33:39.890 --> 00:33:46.269
Intel Arc GPU. It was pretty quick. One of the
00:33:46.269 --> 00:33:48.109
things that really struck me was that actually,
00:33:48.269 --> 00:33:51.910
there's a lot of criticism about how much speed
00:33:51.910 --> 00:33:54.009
you need, what kind of GPUs, all this stuff.
00:33:54.049 --> 00:33:57.730
I'm like, you know what? I'm the kind of person
00:33:57.730 --> 00:34:00.680
who used to wait. I'll say this to someone else.
00:34:00.920 --> 00:34:03.359
A long time ago, I used to render stuff on my
00:34:03.359 --> 00:34:08.119
old 16 -bit computer. I used to leave my computer
00:34:08.119 --> 00:34:15.260
on overnight for a 320 by 240 kind of scene.
00:34:15.820 --> 00:34:18.039
And I don't really mind. And these things are
00:34:18.039 --> 00:34:21.219
fast and they're really quick. And if you start
00:34:21.219 --> 00:34:24.840
to, it starts to help you to realize. don't do
00:34:24.840 --> 00:34:26.739
what you do with the large models where you type
00:34:26.739 --> 00:34:28.480
in a bunch of stuff you type in stuff that's
00:34:28.480 --> 00:34:31.980
direct and you you end up working out in between
00:34:31.980 --> 00:34:36.119
and actually it was pretty useful and i actually
00:34:36.119 --> 00:34:40.900
found it quite easy it worked pretty well um
00:34:40.900 --> 00:34:43.960
i really urge people to at least have a look
00:34:43.960 --> 00:34:47.920
at the or try some of the local model stuff because
00:34:47.920 --> 00:34:50.719
it's not going to solve a whole bunch of things
00:34:50.719 --> 00:34:54.090
the simple things that you probably do with um
00:34:54.090 --> 00:34:58.369
the the other large language models are actually
00:34:58.369 --> 00:35:00.250
pretty good now to be honest i'll be sorry to
00:35:00.250 --> 00:35:04.210
be very clear i am not saying it's a good this
00:35:04.210 --> 00:35:06.409
is good because it's still biases and you can
00:35:06.409 --> 00:35:09.030
actually start to find some of the biases in
00:35:09.030 --> 00:35:11.250
some of the large language models that you you
00:35:11.250 --> 00:35:13.849
use and you can see like for example i got it
00:35:13.849 --> 00:35:17.610
to do a very simple thing where it was i create
00:35:17.610 --> 00:35:21.300
an xml And then I created an XSLT for it. Sorry,
00:35:21.380 --> 00:35:23.239
this is a transformation for it. And I tried
00:35:23.239 --> 00:35:26.380
to get it to create the XML in Markdown. And
00:35:26.380 --> 00:35:28.920
in a Markdown file, and it basically had nothing
00:35:28.920 --> 00:35:32.139
in it. And it was like, okay, this is obviously
00:35:32.139 --> 00:35:34.639
wrong. Because it didn't understand what Markdown
00:35:34.639 --> 00:35:39.639
was. So I think this... In the simple Markdown,
00:35:39.639 --> 00:35:42.940
it didn't really understand what I was trying
00:35:42.940 --> 00:35:45.300
to say. Even though I made it very clear and
00:35:45.300 --> 00:35:49.119
easy, I had to really... do a bunch of things
00:35:49.119 --> 00:35:52.619
to go, this is a markdown file. This is an example
00:35:52.619 --> 00:35:54.940
of what I would like. I want you to do the same
00:35:54.940 --> 00:35:57.639
with this XML file, which I've given you. And
00:35:57.639 --> 00:36:01.800
it took a while, but it got there. And I found
00:36:01.800 --> 00:36:04.699
it really interesting that it all runs pretty
00:36:04.699 --> 00:36:10.000
straightforwardly on my framework for a team
00:36:10.000 --> 00:36:14.280
without any stress. And it wasn't even touching
00:36:14.280 --> 00:36:17.969
the top end of the memory. I want to make a point
00:36:17.969 --> 00:36:23.269
about this entire kind of area, which is that
00:36:23.269 --> 00:36:24.989
one, I think it's going to become more important.
00:36:26.469 --> 00:36:28.969
Because first of all, you've heard the term frontier
00:36:28.969 --> 00:36:33.130
minus six. I've heard it. I didn't actually totally
00:36:33.130 --> 00:36:35.550
know what you mean. Frontier minus six is the
00:36:35.550 --> 00:36:38.250
models that are now six months old. And this
00:36:38.250 --> 00:36:40.190
is why I asked really important what the model
00:36:40.190 --> 00:36:45.119
is. But here's the thing. And we've already seen
00:36:45.119 --> 00:36:47.260
some of this. Actually, Simon Willison has done
00:36:47.260 --> 00:36:49.400
some work on this. He's got his little, basically,
00:36:49.840 --> 00:36:53.699
there are a whole bunch of problems that are
00:36:53.699 --> 00:36:57.119
associated with the frontier versions of these
00:36:57.119 --> 00:36:59.039
LLMs, which are the ones that are so big and
00:36:59.039 --> 00:37:00.800
so powerful. And they're currently the best performing
00:37:00.800 --> 00:37:03.260
and probably will always be the best performing
00:37:03.260 --> 00:37:08.539
due to their nature. But they can only be effectively
00:37:08.539 --> 00:37:13.599
run in the data centers. Yeah. So I have rejoined
00:37:13.599 --> 00:37:15.239
the world of the employed. And one of the things
00:37:15.239 --> 00:37:16.900
that we're talking about that we have to be mindful
00:37:16.900 --> 00:37:18.900
of in the work that we do, because sometimes
00:37:18.900 --> 00:37:22.639
it deals with PII, is like we cannot use Claude
00:37:22.639 --> 00:37:26.619
to handle anything that might touch UK -based
00:37:26.619 --> 00:37:29.139
PII, because we can't guarantee that Claude isn't
00:37:29.139 --> 00:37:31.000
going to send that PII back to the States to
00:37:31.000 --> 00:37:34.579
process it. Yeah. You might want to explain to
00:37:34.579 --> 00:37:38.219
people that PII is an American term. It's personally
00:37:38.219 --> 00:37:40.519
identifiable information. Over here, we just
00:37:40.519 --> 00:37:46.139
call it personal data. Or stuff. It depends if
00:37:46.139 --> 00:37:49.099
you're in banking, you'll use one term. The banks
00:37:49.099 --> 00:37:54.659
are definitely using PII. Besides that, here's
00:37:54.659 --> 00:37:56.860
the thing. Locally run models solve that problem.
00:37:57.579 --> 00:38:00.940
And the meta skill of understanding how to set
00:38:00.940 --> 00:38:05.079
up your machines to use local models, and it's
00:38:05.079 --> 00:38:07.059
particular to be able to swap them in and out
00:38:07.059 --> 00:38:09.460
easily, to have the interfaces that go, you know,
00:38:09.460 --> 00:38:10.940
because when another six months have passed,
00:38:11.320 --> 00:38:13.980
there'll be another round of locally runnable
00:38:13.980 --> 00:38:15.739
models that will be even better than the ones
00:38:15.739 --> 00:38:17.280
that we've got. Like it all just kind of keeps
00:38:17.280 --> 00:38:20.480
going down. It's turtles going down. And so that
00:38:20.480 --> 00:38:23.239
capacity to run these models locally. Now that's
00:38:23.239 --> 00:38:25.920
the technology side of it. The other thing that
00:38:25.920 --> 00:38:28.980
I'm finding really interesting is like, The variety
00:38:28.980 --> 00:38:31.039
of arguments, and there's quite a lot of them
00:38:31.039 --> 00:38:33.280
against AI, a lot of them are grounded in the
00:38:33.280 --> 00:38:36.119
environmental damage. What is the environmental
00:38:36.119 --> 00:38:37.960
damage if you're running it on your machine?
00:38:39.460 --> 00:38:42.139
Yeah, I mean, obviously, you know, the water
00:38:42.139 --> 00:38:44.760
table in your home might be ruined. Well, there's
00:38:44.760 --> 00:38:47.699
still the, like, but my box is the same box.
00:38:47.780 --> 00:38:50.199
Like, at that point, the environmental damage
00:38:50.199 --> 00:38:52.699
of running an LLM becomes the same environmental
00:38:52.699 --> 00:38:56.230
damage of me playing Factorio. Yeah, or GTA 5,
00:38:56.349 --> 00:38:59.969
good God. To some extent, because there's still
00:38:59.969 --> 00:39:03.469
the original training cost of these data. Yeah,
00:39:03.469 --> 00:39:05.510
I'm not going to, I can't get around, yeah, you're
00:39:05.510 --> 00:39:08.309
absolutely right. I think, I was thinking of
00:39:08.309 --> 00:39:10.670
that same thing. The actual cost, the actual
00:39:10.670 --> 00:39:15.769
environmental impact of my machine is the same
00:39:15.769 --> 00:39:19.809
as it is generally, but in like a bunch of things,
00:39:19.889 --> 00:39:24.519
but the cost to run and build those. large language
00:39:24.519 --> 00:39:26.920
models is still huge and there's nothing you
00:39:26.920 --> 00:39:30.000
can get around that. The training cost is big.
00:39:30.639 --> 00:39:32.719
Depending on how they do it, it can actually
00:39:32.719 --> 00:39:34.280
be a lesser training cost, particularly if they're
00:39:34.280 --> 00:39:35.840
doing something smart like grinding on an already
00:39:35.840 --> 00:39:38.440
existing LLM to get it down smaller. Oh no, they
00:39:38.440 --> 00:39:40.440
stole our data. Really? Did they steal your data?
00:39:41.139 --> 00:39:45.860
How dare they? How dare you steal the data that
00:39:45.860 --> 00:39:47.900
we rightfully stole on our own? But that's also
00:39:47.900 --> 00:39:51.079
the next other question. We stole it differently.
00:39:52.010 --> 00:39:54.269
But it's still derived from stolen data, which
00:39:54.269 --> 00:39:56.650
is one of the legitimate, also ethical problems
00:39:56.650 --> 00:39:59.309
with a lot of AIs as they exist, like because
00:39:59.309 --> 00:40:02.610
they were all trained on stolen data. And the
00:40:02.610 --> 00:40:05.170
fact that you've derived this data from stolen
00:40:05.170 --> 00:40:07.010
data, that you've stole the stolen data, you
00:40:07.010 --> 00:40:12.190
know, three thefts don't make a right. I just
00:40:12.190 --> 00:40:17.090
make a more smarter thief. But that's so like
00:40:17.090 --> 00:40:19.489
I but I. The thing I like about this is thinking
00:40:19.489 --> 00:40:20.949
we're going to have, it's going to be interesting
00:40:20.949 --> 00:40:22.949
to see how these discussions go around and this
00:40:22.949 --> 00:40:26.230
kind of, they start to bleed into this area of
00:40:26.230 --> 00:40:28.630
locally run LLMs and the ethos of locally running
00:40:28.630 --> 00:40:32.010
LLMs versus that. And I would be, I would not
00:40:32.010 --> 00:40:34.869
be surprised if we don't see a little bit of
00:40:34.869 --> 00:40:38.050
the common people who run their LLMs locally
00:40:38.050 --> 00:40:40.750
versus the big, nasty, awful corporate LLMs,
00:40:40.829 --> 00:40:45.769
you know, like basically, yeah. The time gap
00:40:45.769 --> 00:40:48.090
is important. I just underline that. Because
00:40:48.090 --> 00:40:49.710
of what you said about it doesn't understand
00:40:49.710 --> 00:40:52.550
Markdown, I haven't used an LLM that doesn't
00:40:52.550 --> 00:40:55.050
know Markdown better than me for six months.
00:40:55.289 --> 00:40:57.929
I mean, because it's exactly that. Yes, once
00:40:57.929 --> 00:40:59.630
you get past the limit and you're running the
00:40:59.630 --> 00:41:01.670
bigger ones, they have no trouble with that at
00:41:01.670 --> 00:41:03.590
all. Was it not know Markdown, or did you simply
00:41:03.590 --> 00:41:05.710
ask it to expand its context window enough to
00:41:05.710 --> 00:41:07.090
the point where one of the things it dumped out
00:41:07.090 --> 00:41:09.130
was its knowledge of Markdown? Yeah, it might
00:41:09.130 --> 00:41:11.130
have bit rotted it off the left -hand side. I
00:41:11.130 --> 00:41:12.869
mean, that depends again on its window. Another
00:41:12.869 --> 00:41:16.630
problem. the smaller models i think i think okay
00:41:16.630 --> 00:41:18.809
so what i was trying i'm not going to try to
00:41:18.809 --> 00:41:22.010
do but um i i i think you're absolutely right
00:41:22.010 --> 00:41:27.309
i think there's going to be a space where it's
00:41:27.309 --> 00:41:29.710
i mean we see it already with some of the so
00:41:29.710 --> 00:41:33.050
like self -hosting and stuff like that um i mean
00:41:33.050 --> 00:41:36.329
even people like running vms in a cloud somewhere
00:41:36.329 --> 00:41:39.170
and they're running their own docker containers
00:41:39.170 --> 00:41:41.050
and stuff like that i think we will see that
00:41:41.050 --> 00:41:42.469
and i think there'll be a bunch of people who
00:41:42.469 --> 00:41:46.000
go well You know, what I'm doing is not that
00:41:46.000 --> 00:41:49.820
important or not that it doesn't affect what
00:41:49.820 --> 00:41:51.880
I think it doesn't affect. It's not affecting,
00:41:52.000 --> 00:41:59.159
sorry, not worrying much about privacy. But I
00:41:59.159 --> 00:42:01.099
think that there are a number of people who,
00:42:01.199 --> 00:42:04.480
like myself, who will just go, no way. And I'm
00:42:04.480 --> 00:42:06.300
happy to wait the six months. And I'm happy.
00:42:06.780 --> 00:42:09.440
This is why I was happy to wait and just kind
00:42:09.440 --> 00:42:11.940
of go and make a tea and come back and for it
00:42:11.940 --> 00:42:18.699
to be done. I was going to say, you might find
00:42:18.699 --> 00:42:20.519
in some corporate environments, I'm just going
00:42:20.519 --> 00:42:22.179
to have a tea and come back and get the solutions.
00:42:22.340 --> 00:42:27.400
That may not go so well. Obviously, it depends
00:42:27.400 --> 00:42:29.139
on your environment. If you're running your machine
00:42:29.139 --> 00:42:30.860
hot enough, you can use the waste heat to brew
00:42:30.860 --> 00:42:36.559
the tea. That would be efficient. Hey, guys,
00:42:36.599 --> 00:42:37.840
wait a minute. I'm just going to have some tea.
00:42:40.039 --> 00:42:43.460
If it helps, I do know of a cookbook that was
00:42:43.460 --> 00:42:46.519
designed for people who want to cook on the radiators
00:42:46.519 --> 00:42:49.260
in their cars. And you could probably adapt that
00:42:49.260 --> 00:42:53.119
to cook on your hot computer. I am certain that
00:42:53.119 --> 00:42:58.219
there was an old motorcycle -like accessory that
00:42:58.219 --> 00:43:01.360
was designed to clamp onto the tailpipe and let
00:43:01.360 --> 00:43:03.500
you cook food in much the same way off of the
00:43:03.500 --> 00:43:07.389
heat off the tailpipe. Put your sausage in, put
00:43:07.389 --> 00:43:10.130
your sausage in, go for a nice ride, have your
00:43:10.130 --> 00:43:12.630
lunch with the sausage already cooked. Well,
00:43:12.650 --> 00:43:15.650
the friend that I gave that cookbook to was a
00:43:15.650 --> 00:43:17.969
full time folk singer. So he was driving thousands
00:43:17.969 --> 00:43:20.949
of miles a year. And so he spent a lot of time
00:43:20.949 --> 00:43:23.489
in the car anyway. Oh, that sounds pretty cool.
00:43:24.829 --> 00:43:26.530
Oh, and but it wouldn't work so well in electric
00:43:26.530 --> 00:43:31.230
cars, though. Oh, yeah. Well. Not enough pollution,
00:43:31.389 --> 00:43:32.909
damn it. It's quieter and better for the environment,
00:43:32.989 --> 00:43:36.869
but where am I going to get my breakfast? Denny's.
00:43:36.949 --> 00:43:45.230
And for our last full story of the night, we
00:43:45.230 --> 00:43:48.690
have Wendy with a story about Waymos holding
00:43:48.690 --> 00:43:52.170
robo -taxis in a parking lot and some very bad
00:43:52.170 --> 00:43:55.690
consequences. Wendy? Well, the story is, and
00:43:55.690 --> 00:43:59.409
it's been all over social media, is that two
00:43:59.409 --> 00:44:04.690
15 -year -old boys were riding in a rented Waymo
00:44:04.690 --> 00:44:09.389
robo -taxi and were drinking alcohol and firing
00:44:09.389 --> 00:44:12.090
what are described as water beads. I'm not really
00:44:12.090 --> 00:44:14.170
sure what that is. They're obviously some kind
00:44:14.170 --> 00:44:16.889
of softish thing, but it's still not good to
00:44:16.889 --> 00:44:20.570
be firing them out the car windows. And Waymo,
00:44:20.769 --> 00:44:23.489
what seems to have happened is that someone at
00:44:23.489 --> 00:44:26.309
Waymo used the in -car cameras to see what they
00:44:26.309 --> 00:44:29.030
were doing and then pulled the car over into
00:44:29.030 --> 00:44:31.969
a parking lot and called the police. And there
00:44:31.969 --> 00:44:34.949
were like all sorts of reactions to this. One
00:44:34.949 --> 00:44:37.610
friend describes it as kidnapping. Another describes
00:44:37.610 --> 00:44:40.469
it as the teens getting what they should have
00:44:40.469 --> 00:44:42.190
gotten because they were doing something illegal.
00:44:43.230 --> 00:44:46.250
Other people, the Daily Mail called the car a
00:44:46.250 --> 00:44:49.489
tattletale. The Register called it a snitch.
00:44:49.630 --> 00:44:52.010
I mean, it's really, really interesting the way
00:44:52.010 --> 00:44:55.329
people have reacted to this. And there's like
00:44:55.329 --> 00:44:58.809
a couple of questions. I mean, I find the case
00:44:58.809 --> 00:45:02.090
fascinating just because it's clear that there
00:45:02.090 --> 00:45:03.929
are going to be lots more of these cases. And
00:45:03.929 --> 00:45:05.469
we're going to have to come up with some ground
00:45:05.469 --> 00:45:08.909
rules about when it's appropriate for the car
00:45:08.909 --> 00:45:13.179
company to be doing something. And when it's
00:45:13.179 --> 00:45:15.219
not, I mean, one question I have is why did they
00:45:15.219 --> 00:45:17.199
call the police and not their parents? I mean,
00:45:17.219 --> 00:45:20.119
the police showed up, you know, four police officers
00:45:20.119 --> 00:45:23.340
with guns drawn. Does not sound, you know, I
00:45:23.340 --> 00:45:26.019
think they're lucky they weren't killed. And
00:45:26.019 --> 00:45:29.340
but the other question that I find really interesting
00:45:29.340 --> 00:45:31.940
is how did they get the car in the first place?
00:45:33.079 --> 00:45:36.869
Because. Five, ten years ago, you go back and
00:45:36.869 --> 00:45:39.570
the story that everybody was talking about self
00:45:39.570 --> 00:45:41.389
-driving cars were going to make it possible
00:45:41.389 --> 00:45:44.809
for people who can't drive to ride in taxis.
00:45:46.010 --> 00:45:48.630
There was this fantasy that you would put your
00:45:48.630 --> 00:45:50.409
kids in the car and the car would take them to
00:45:50.409 --> 00:45:52.030
school and you wouldn't have to do it yourself.
00:45:52.849 --> 00:45:56.130
And blind people would be able to ride and it
00:45:56.130 --> 00:45:59.590
would open up car travel to blind people and
00:45:59.590 --> 00:46:04.289
so on. And apparently Waymo actually does have
00:46:04.289 --> 00:46:06.429
restrictions on who can rent their cars. And
00:46:06.429 --> 00:46:08.550
they're supposed to have tightened up on rentals
00:46:08.550 --> 00:46:10.630
to teens. But it's still very interesting to
00:46:10.630 --> 00:46:13.210
me that 10 years ago when people were sort of
00:46:13.210 --> 00:46:16.250
proposing this fantasy future, they didn't seem
00:46:16.250 --> 00:46:19.230
to realize that kids left alone in a rented car
00:46:19.230 --> 00:46:22.690
might find things to do that were not actually
00:46:22.690 --> 00:46:28.909
necessarily. It's not just kids. When I saw the
00:46:28.909 --> 00:46:30.909
story, my thought was. Well, of course they're
00:46:30.909 --> 00:46:33.329
going to. They have this problem with Airbnbs
00:46:33.329 --> 00:46:35.670
with adults where they're like, people will set
00:46:35.670 --> 00:46:38.030
up an Airbnb account. They'll use a limited,
00:46:38.289 --> 00:46:40.769
you know, the whole, let's get an Airbnb for
00:46:40.769 --> 00:46:42.789
the weekend and just trash it because who cares?
00:46:43.230 --> 00:46:45.769
And this is just a micro version of that instead
00:46:45.769 --> 00:46:50.889
of 15 college students grabbing a property for
00:46:50.889 --> 00:46:53.670
a weekend bender party. Now it's kids with airsoft
00:46:53.670 --> 00:46:56.690
guns or whatever it is. I don't think it's airsoft.
00:46:56.750 --> 00:46:58.070
That's something else. But at any rate, the...
00:46:59.280 --> 00:47:01.119
My other thought was, like, of course they don't
00:47:01.119 --> 00:47:02.820
know who the kids' parents are. I've never been
00:47:02.820 --> 00:47:05.420
asked who my parents are when I rent a car. Yeah,
00:47:05.659 --> 00:47:07.460
I was going to say they wouldn't know. How do
00:47:07.460 --> 00:47:09.500
they rent a robot? Well, first of all, you can't
00:47:09.500 --> 00:47:12.699
rent a car when you're under 25. Fair. Second
00:47:12.699 --> 00:47:15.619
of all, how are they supposed to rent a car?
00:47:15.820 --> 00:47:19.400
Do they have credit cards? They're 15. Yeah,
00:47:19.500 --> 00:47:21.739
15 -year -olds have a ton of credit cards. In
00:47:21.739 --> 00:47:23.380
Britain, they're not allowed to have credit cards
00:47:23.380 --> 00:47:25.900
until you're 18. Well, you can't have a credit
00:47:25.900 --> 00:47:28.489
card. You can have a card with money on it. Not
00:47:28.489 --> 00:47:29.969
the same thing. I don't think that would work
00:47:29.969 --> 00:47:33.130
for rental, by the way, anyway. They won't work
00:47:33.130 --> 00:47:34.909
for rental, but this isn't a rental car. This
00:47:34.909 --> 00:47:41.510
is a taxi. Yeah, so it's fine. Precisely. I mean,
00:47:41.530 --> 00:47:45.989
the same thing. I mean, these taxis being accessible
00:47:45.989 --> 00:47:51.489
to kids is part of the selling point. What level
00:47:51.489 --> 00:47:55.610
of illegal activity should the car company call
00:47:55.610 --> 00:47:59.619
the police? I think it has to affect the driving
00:47:59.619 --> 00:48:02.699
or safety of the car, surely. They could have
00:48:02.699 --> 00:48:04.780
just diverted the car into a parking lot and
00:48:04.780 --> 00:48:07.119
opened the door and refused to move the car further
00:48:07.119 --> 00:48:10.000
until the teens got out. That could be dangerous
00:48:10.000 --> 00:48:12.840
in itself. We are never going to do that because
00:48:12.840 --> 00:48:15.340
the whole philosophy of these cars is fuck human
00:48:15.340 --> 00:48:18.760
interaction at all levels. Yes, there's something
00:48:18.760 --> 00:48:21.320
wrong with that whole story. Somebody at Reno
00:48:21.320 --> 00:48:23.960
called the police. Calling the police was the
00:48:23.960 --> 00:48:27.900
most direct and easy way to offload the responsibility
00:48:27.900 --> 00:48:30.179
of any form of human interaction. And that's
00:48:30.179 --> 00:48:32.320
like that's that's just them doing the same thing
00:48:32.320 --> 00:48:34.599
that the American police are being asked to do
00:48:34.599 --> 00:48:36.380
all the time because they've removed other services.
00:48:37.380 --> 00:48:42.869
Sorry, go ahead. I agree with Ryan. You know,
00:48:42.889 --> 00:48:44.829
I think it's a really good way of offloading
00:48:44.829 --> 00:48:47.150
responsibility to someone else, you know, not
00:48:47.150 --> 00:48:50.969
deal with it. The moment, the moment it's, and
00:48:50.969 --> 00:48:54.389
also this is, this is absolutely break, like
00:48:54.389 --> 00:48:56.730
move fast and break things mindset where it's
00:48:56.730 --> 00:48:58.789
like the moment anything steps off the happy
00:48:58.789 --> 00:49:01.929
pass. Fuck you. Call the call the cops. So we
00:49:01.929 --> 00:49:05.289
can't be bothered to engineer for all these weird
00:49:05.289 --> 00:49:07.670
out there scenarios like irresponsible teenagers.
00:49:07.849 --> 00:49:09.909
How many how many really irresponsible teenagers
00:49:09.909 --> 00:49:11.730
could there possibly be out there? We can't be
00:49:11.730 --> 00:49:14.070
expected to anticipate that completely insane
00:49:14.070 --> 00:49:17.849
scenario. Just throw it to the cops. OK, so granted,
00:49:17.909 --> 00:49:22.670
granted all that. What if one of the teens had
00:49:22.670 --> 00:49:26.630
been killed? That would be extremely problematic.
00:49:28.469 --> 00:49:29.889
Ironically, I think it would have been a story
00:49:29.889 --> 00:49:32.070
that we rolled in the AI bucket because then
00:49:32.070 --> 00:49:35.690
it would have been a story about when AI does
00:49:35.690 --> 00:49:39.429
bad stuff. I would say also that they probably
00:49:39.429 --> 00:49:42.030
have somewhere a news license agreement saying
00:49:42.030 --> 00:49:45.530
we are not responsible if you die in response
00:49:45.530 --> 00:49:51.050
to transit. I was thinking more as a result of
00:49:51.050 --> 00:49:54.010
calling the cops. Oh, well, that's not their
00:49:54.010 --> 00:49:56.269
responsibility then, right? That's the cops'
00:49:56.349 --> 00:50:00.070
responsibility. Well, there was a person they
00:50:00.070 --> 00:50:02.849
asked, there was a quote that went viral a couple
00:50:02.849 --> 00:50:05.389
of weeks ago, where they walked up to a woman
00:50:05.389 --> 00:50:07.730
in the street and they said, if you had just
00:50:07.730 --> 00:50:09.409
killed a black person, where would you hide?
00:50:09.530 --> 00:50:12.829
And she paused for a second and she said, behind
00:50:12.829 --> 00:50:22.269
a badge. Yeah, yeah, yeah. Just be honest. And
00:50:22.269 --> 00:50:27.730
this is just another variant on that. Now, entertainingly,
00:50:27.789 --> 00:50:34.309
any police officer with a smidgen of self -preservation
00:50:34.309 --> 00:50:37.570
would have a lot of problems with that. And yet
00:50:37.570 --> 00:50:42.349
it's happening a lot. Yeah. I think, so I just,
00:50:42.389 --> 00:50:46.230
I want to go back to what I said. I think Ryan's
00:50:46.230 --> 00:50:49.610
correct. Waymo just passed it to the police,
00:50:49.730 --> 00:50:52.849
as Ryan said, because lots of people, lots of
00:50:52.849 --> 00:50:55.889
companies are just doing that. And then if anything
00:50:55.889 --> 00:50:59.309
happens beyond that, they just go, well, we gave
00:50:59.309 --> 00:51:02.590
it to the police and we're not responsible for
00:51:02.590 --> 00:51:05.329
it. And it's just this buck passing. And I think
00:51:05.329 --> 00:51:09.090
it is what happens a lot when it comes to these
00:51:09.090 --> 00:51:12.010
companies. They don't have any procedure for
00:51:12.010 --> 00:51:16.210
if this happens. pass it on and wipe their hands
00:51:16.210 --> 00:51:18.489
off air. It's horrible. It's disgusting. Actually,
00:51:18.489 --> 00:51:23.369
it's more specific. When we say Waymo, we mean
00:51:23.369 --> 00:51:26.710
a mechanical Turk in Cambodia. Saw it. Anon is
00:51:26.710 --> 00:51:29.250
sort of hinting at what circumstance it goes,
00:51:29.349 --> 00:51:31.510
blah, blah. Number 17. Yeah, if you've got here,
00:51:31.570 --> 00:51:32.989
ring the police because you don't have anything
00:51:32.989 --> 00:51:36.769
relevant to. That's all it is. I assumed it was
00:51:36.769 --> 00:51:40.369
a human at Waymo, but not. It'll be a human.
00:51:41.309 --> 00:51:43.329
Yeah, of course it was a human. Well, it'll be
00:51:43.329 --> 00:51:46.610
for some value of at Waymo, I think. Calling
00:51:46.610 --> 00:51:50.170
the car a snitch is a bit ridiculous because
00:51:50.170 --> 00:51:52.829
it wasn't the car that made the call. Calling
00:51:52.829 --> 00:51:55.110
the car the snitch is a great way to get them
00:51:55.110 --> 00:51:57.710
to be able to dodge the responsibility too. It
00:51:57.710 --> 00:52:02.710
wasn't us, Gov, it was the AI. If I was a snitch,
00:52:02.710 --> 00:52:06.650
what can we do? You're going to have to try again.
00:52:06.789 --> 00:52:11.090
You've gone and trained up a snitch. Yeah. What
00:52:11.090 --> 00:52:13.889
car do you drive? I drive a snitch. You know,
00:52:13.929 --> 00:52:15.690
I think that's another it's another example,
00:52:16.030 --> 00:52:20.050
though, of that. They always talk as though the
00:52:20.050 --> 00:52:23.150
only job the driver in a taxi does is to drive
00:52:23.150 --> 00:52:27.110
the taxi. But actually, the drivers have lots
00:52:27.110 --> 00:52:29.090
more functions than that. And one of them is
00:52:29.090 --> 00:52:31.969
15 year old boys don't get into taxis with an
00:52:31.969 --> 00:52:35.239
adult driver and start. They might drink alcohol,
00:52:35.360 --> 00:52:37.980
but they're not going to be shooting beads out
00:52:37.980 --> 00:52:41.400
of the windows. Well, also, adult drivers of
00:52:41.400 --> 00:52:43.840
taxis spot drunk 15 -year -olds and say, fuck
00:52:43.840 --> 00:52:45.420
no, get another taxi. Yeah, because they get
00:52:45.420 --> 00:52:51.260
a puke on their seats. Indeed. Or at the very
00:52:51.260 --> 00:52:53.079
least, if they take the fare, they give them
00:52:53.079 --> 00:52:54.980
the look of death and like, you start any trouble,
00:52:55.099 --> 00:52:57.639
I've got a baseball bat right under here. You
00:52:57.639 --> 00:53:01.690
know, I know I know you're laughing about that,
00:53:01.769 --> 00:53:03.530
but actually, I remember a long time ago when
00:53:03.530 --> 00:53:08.349
I used to... He's back from work in Bristol.
00:53:08.909 --> 00:53:11.989
One day, actually, a taxi driver kind of showed
00:53:11.989 --> 00:53:15.690
me... It wasn't a machete, but it was like a
00:53:15.690 --> 00:53:20.989
small machete. And he said, look, you know, lots
00:53:20.989 --> 00:53:24.250
of people run off. I have this. And if you try
00:53:24.250 --> 00:53:27.329
to think, I have this. He wasn't threatening.
00:53:28.500 --> 00:53:32.900
But he was clearly... I don't know. That comes
00:53:32.900 --> 00:53:34.619
pretty close to threatening there, Ian. Okay.
00:53:34.699 --> 00:53:37.800
But he was just... It was very clear that you
00:53:37.800 --> 00:53:41.699
don't do... He's ready. And it was a bit like...
00:53:41.699 --> 00:53:44.780
At best, it was a very proactive form of threatening,
00:53:44.900 --> 00:53:50.539
shall we say. I'm not saying it would be a problem.
00:53:50.760 --> 00:53:53.139
I'm just saying you're not going to be a problem.
00:53:54.099 --> 00:54:01.889
Yes. And last but not least on our lovely world
00:54:01.889 --> 00:54:05.449
of tech grumps, we have any other business with
00:54:05.449 --> 00:54:08.030
a couple of items. Ian, what do you say about
00:54:08.030 --> 00:54:12.690
EMF? Yeah, so EMF, by the time you hear this,
00:54:12.750 --> 00:54:14.750
you might have, it's probably too late to get
00:54:14.750 --> 00:54:19.670
a ticket, but EMF is happening and it's every
00:54:19.670 --> 00:54:23.489
two years, I think, yeah. And I will be talking
00:54:23.489 --> 00:54:27.190
about two, well, I'll be talking. one thing is
00:54:27.190 --> 00:54:31.190
around digital what is emf what is oh electronic
00:54:31.190 --> 00:54:36.909
magnetic field pump is a camp for geeks and it's
00:54:36.909 --> 00:54:39.750
it's great fun and it's the only camping i will
00:54:39.750 --> 00:54:42.690
do because i hate camping but when it's got power
00:54:42.690 --> 00:54:48.329
and reliable wi -fi i'm in okay it's got high
00:54:48.329 --> 00:54:50.170
speed internet high speed power capabilities
00:54:50.170 --> 00:54:54.449
but it only happens every two years Yes. It's
00:54:54.449 --> 00:54:57.050
definitely worth going for if you get a chance.
00:54:57.550 --> 00:55:02.989
Okay, so I'm about digital privacy. I'm literally
00:55:02.989 --> 00:55:04.889
looking at the presentation right now. Machine
00:55:04.889 --> 00:55:09.750
wishes. And I'm actually in stage A on Sunday.
00:55:09.989 --> 00:55:13.329
The time is still being bounced around. I see
00:55:13.329 --> 00:55:17.150
another email in my client saying, we've changed
00:55:17.150 --> 00:55:20.389
your time. So it will be early, unfortunately.
00:55:21.239 --> 00:55:24.739
which also messes because the next one is doing,
00:55:24.820 --> 00:55:30.739
I'm DJing on Saturday night or Sunday morning.
00:55:32.119 --> 00:55:35.780
And so I'm going to have to DJ, I think the last
00:55:35.780 --> 00:55:38.739
set, Trance. So you should definitely be there
00:55:38.739 --> 00:55:40.760
if you're there. But then I also have to get
00:55:40.760 --> 00:55:45.019
up early to do my talk in 10A. So that's going
00:55:45.019 --> 00:55:48.840
to be fun. I'm going to be definitely very wide
00:55:48.840 --> 00:55:51.619
awake at that time. Where is it this year? I
00:55:51.619 --> 00:55:54.320
didn't recognize the name. It said somewhere.
00:55:54.559 --> 00:55:57.199
Is it not the same place as you? I think so.
00:55:58.679 --> 00:56:03.659
Maybe it is. I would like to make the point that
00:56:03.659 --> 00:56:06.820
now that you will be performing at EMF, Ian,
00:56:06.980 --> 00:56:11.659
you understand given this year's Eurovision entry
00:56:11.659 --> 00:56:15.179
from the UK. Uh -oh. I didn't watch Eurovision.
00:56:15.179 --> 00:56:17.360
I had no idea. Well, this year's Eurovision entry
00:56:17.360 --> 00:56:22.980
has also performed at EMF. wow i'm just saying
00:56:22.980 --> 00:56:26.579
yes this year's this year's like i was i was
00:56:26.579 --> 00:56:28.820
i do i do eurovision every year and i get ready
00:56:28.820 --> 00:56:30.300
to tear them all a new one and then i was reading
00:56:30.300 --> 00:56:33.159
up on the uk entry which was usually a prime
00:56:33.159 --> 00:56:36.440
choice for ridicule a year and year and year
00:56:36.440 --> 00:56:40.099
out but this guy despite having a an entry that
00:56:40.099 --> 00:56:42.139
was actually horrible and i did not like it but
00:56:42.139 --> 00:56:44.400
i i actually saw that he had performed at emf
00:56:44.400 --> 00:56:46.039
and a lot of people that i actually knew personally
00:56:46.039 --> 00:56:47.500
were like oh yeah he's nice he's really cool
00:56:47.500 --> 00:56:49.340
he's lovely and i was like okay i can't tear
00:56:49.340 --> 00:56:53.500
him a new one So I'm just saying, Ian, Eurovision
00:56:53.500 --> 00:56:56.519
next year, right? It could be there. It could
00:56:56.519 --> 00:57:03.039
be you. Yeah, it could be me. All right. You've
00:57:03.039 --> 00:57:05.440
got something else on Death Glitch? Yeah. So
00:57:05.440 --> 00:57:10.019
I follow on from my thoughts about Digital Legacy.
00:57:11.260 --> 00:57:14.599
There's a book I read recently called Death Glitch.
00:57:15.159 --> 00:57:23.039
It's by Tamara. I think that's her name. And
00:57:23.039 --> 00:57:27.480
it's about digital legacy and it talks about
00:57:27.480 --> 00:57:33.900
the whole death sector. And it's really interesting.
00:57:34.679 --> 00:57:38.300
I recommend it to lots of people because even
00:57:38.300 --> 00:57:41.159
if you're not interested in this space, you're
00:57:41.159 --> 00:57:43.400
going to die at some point. And it's really,
00:57:43.519 --> 00:57:47.079
it sums up a lot of the... ideas and thoughts
00:57:47.079 --> 00:57:48.920
that people have when they actually think about
00:57:48.920 --> 00:57:52.880
this like what happens to your data so yeah what
00:57:52.880 --> 00:57:56.340
happens to your digital remains you know um what
00:57:56.340 --> 00:57:58.619
happens to your estate your digital estate like
00:57:58.619 --> 00:58:02.239
when you've got the likes of playstation and
00:58:02.239 --> 00:58:06.920
xbox or microsoft who are um are not providing
00:58:06.920 --> 00:58:09.320
digital discs for example you know those kind
00:58:09.320 --> 00:58:13.599
of all kind of factor into this um If you are
00:58:13.599 --> 00:58:15.980
interested in the Def Clitch, you'll see parts
00:58:15.980 --> 00:58:19.679
of it in my presentation. So come along at 10
00:58:19.679 --> 00:58:24.539
o 'clock in the morning to that. Well, interestingly
00:58:24.539 --> 00:58:28.559
related, my little NLB is that I finally, after
00:58:28.559 --> 00:58:31.380
having reacquired a source of income, treated
00:58:31.380 --> 00:58:37.199
myself to a new NAS box. For 10 years, I've been
00:58:37.199 --> 00:58:40.179
running off of machines that I bought cheap that
00:58:40.179 --> 00:58:42.099
were end of life from work and other places.
00:58:42.300 --> 00:58:43.599
And now I finally just went, you know what? I'm
00:58:43.599 --> 00:58:45.500
just going to actually buy a new one. And it's
00:58:45.500 --> 00:58:48.280
lovely. And they work really well. And this one
00:58:48.280 --> 00:58:50.000
is not even one of the really hardcore ones.
00:58:50.179 --> 00:58:53.000
I think it's a Ugreen 4800. It has four bays,
00:58:53.059 --> 00:58:55.980
which is all I need. But I've been putting everything
00:58:55.980 --> 00:58:59.239
up on it. I've now moved a lot of my media. And
00:58:59.239 --> 00:59:02.460
for me, the biggest thing was, and I... I was
00:59:02.460 --> 00:59:04.619
hoping this would exist in the world. It does.
00:59:04.760 --> 00:59:07.400
I now have set it up so that I have access to
00:59:07.400 --> 00:59:10.059
all of my audiobooks. And it was surprisingly
00:59:10.059 --> 00:59:12.219
easy to establish the kind of things that I did
00:59:12.219 --> 00:59:14.380
so that I have my audiobooks all on a server.
00:59:14.820 --> 00:59:17.699
It can connect to my phone directly. So anytime
00:59:17.699 --> 00:59:20.860
I want to be listening to one of my audiobooks,
00:59:20.860 --> 00:59:23.760
I have the ability to just grab it and download
00:59:23.760 --> 00:59:25.420
it anywhere in the world. And I'm very happy
00:59:25.420 --> 00:59:29.719
about that. And it's giving me joy. And that's
00:59:29.719 --> 00:59:32.380
it for any other business. That means it's goodbye
00:59:32.380 --> 00:59:34.739
from all of us here at TechGrums. And we'll do
00:59:34.739 --> 00:59:37.139
the final round of saying goodbye. My name is
00:59:37.139 --> 00:59:39.059
Ryan. I have been your host. I hope you have
00:59:39.059 --> 00:59:41.679
a great time until we hear you next time. Goodbye,
00:59:41.719 --> 00:59:46.099
Ian. And goodbye from me. Hopefully you'll see
00:59:46.099 --> 00:59:51.139
me DJ live at EMF camp or you will hear the record
00:59:51.139 --> 00:59:55.530
if I remember to record it on my PeerTube. Goodbye.
00:59:55.849 --> 00:59:59.190
David. Goodbye from me. You're here. Well, you
00:59:59.190 --> 01:00:01.510
weren't here, but you can read lots of World
01:00:01.510 --> 01:00:04.570
Cup live summaries on Blue Sky, which is where
01:00:04.570 --> 01:00:07.329
I do them at the moment. Goodbye from me. And
01:00:07.329 --> 01:00:11.070
Wendy. Goodbye. Bye. Bye. Bye.