Up next in 10
Discover the groundbreaking uses of AI with AI Facial Recognition Glasses that can identify anyone! See how AI music scams are exploiting streaming platforms and explore how AI is being used to debunk conspiracy theories. Join our insightful AI discussion! #AIFacialRecognition #AIGlasses #AIMusicScam #ConspiracyTheories #ArtificialIntelligence #TechNews #Innovation #FutureTech #DataPrivacy #VRTechnology
Show More Show Less View Video Transcript
0:00
Let's start with a story that's just
0:02
come out this week of two Harvard
0:04
students in the US who've hacked Meta's
0:07
Ray-B band glasses in order to turn them
0:10
into makeshift um uh facial recognition
0:14
systems. Right? So, all you have to do
0:16
is look at someone and it will pull up
0:18
information on that person about who
0:20
they are, where they live, what their
0:21
phone number is, and a lot more. Take a
0:23
look. We built glasses that let you
0:26
identify anybody on the street. The
0:27
information a tool collects from just a
0:29
photo of your face is staggering.
0:34
God, I'm scared. Oh my god. To use it,
0:37
you just put the glasses on. Then as you
0:39
walk by people, the glasses will detect
0:41
when somebody's face is in frame. This
0:43
photo is used to analyze them, and after
0:45
a few seconds, their personal
0:46
information pops up on your phone. Hi,
0:48
ma'am. Wait, are you a Betsy? Yes. Oh,
0:52
okay. I think I uh I think I met you
0:53
through like the Cambridge Community
0:55
Foundation, right? Yeah. Yeah. Hey, it's
0:57
great to meet you. I'm Kane. Now, you
0:59
might be wondering what that's got to do
1:00
with artificial intelligence. Like with
1:02
many applications, it's just making
1:04
things easier and faster. A lot of this
1:06
technology was already available, but
1:08
what they've done is basically um create
1:10
a system which essentially takes a photo
1:12
of someone's face. Then it uses the
1:14
internet to search for that face and
1:15
pulls up lots of URLs. And then those
1:18
URLs are fed into a large language model
1:20
which passes through all of those
1:22
different web pages and pulls out all
1:24
the personal information and
1:26
connections. it can possibly find to
1:28
bring up a profile of this person. Um,
1:31
now if you think you're safe from this
1:34
because you don't put your personal
1:35
information, like your phone number and
1:37
your address online, well,
1:38
unfortunately, there have been a lot of
1:40
big uh leaks and hacks over the years
1:42
that have caused lots of people's
1:43
personal information to uh be put into
1:47
databases on the internet and it does
1:49
use some of these databases as well. So,
1:51
it just shows how powerful the use of AI
1:53
can be when when applied to something
1:55
like this. Meta has told 404 media that
1:58
this is a breach of their terms of
1:59
service. Um, but it's just more evidence
2:02
for people who are calling for more
2:04
regulation on something like virtual
2:05
reality systems. The tech itself keeps
2:08
getting more and more sleek. We've seen
2:10
two two weeks ago Meta announce its
2:13
project prototype Orion which are the
2:16
most sort of discrete slimline um
2:19
virtual reality glasses they've they've
2:21
ever developed. So it it just showing
2:24
now that um a lot can be done uh with
2:27
just a pair of glasses like the ones I'm
2:29
I'm wearing already. It's slightly
2:30
terrifying and if some students can do
2:33
very smart students imagine what in the
2:36
wrong hands this could this could look
2:38
like. Uh, okay. So, this is not the only
2:40
scandal that you're bringing us today.
2:42
You've got another one involving uh AI
2:45
generated music. Yeah, that's right.
2:47
Let's take a quick listen to some of it
2:49
now.
2:52
[Music]
2:55
Okay, so obviously not the best thing
2:57
I've heard of. We know we know because
2:59
we've said it's AI generated music.
3:00
We're talking about AI. We know that
3:02
it's AI generated. But if this were just
3:04
to come up on your Spotify playlist that
3:06
you're listening to in the background
3:07
and you've got a similar music theme
3:09
you're listening to, I don't think many
3:13
people would notice straight away. And
3:14
that's exactly what a man called Michael
3:16
Smith um is was alleged to be exploiting
3:20
when he pulled off um a massive scam
3:23
that US prosecutors are alleging he's uh
3:26
pulled off. So he generated tens of
3:29
thousands of fake accounts to listen to
3:32
hundreds of thousands of fake songs that
3:35
he put across lots of different
3:36
streaming platforms all of which were
3:39
generated by AI. the music, the titles
3:42
of the songs, the names of the artists
3:44
and he used these to stream stream these
3:47
songs uh bill literally billions of
3:49
times and from 2018 to this year he
3:52
generated more than 10 million US in uh
3:56
royalty revenue from these streaming
3:59
platforms. So that's what's in the
4:01
indictment that's been unsealed by US
4:03
prosecutors. He's faced with facing
4:06
three charges now which each carry 20
4:08
years in prison including wire wire
4:10
fraud and money laundering. I think
4:12
what's crazy to me is that the scheme
4:14
started in 2018 as I said that's a long
4:16
time ago in terms of AI and things have
4:19
got so much better and more convincing
4:20
since then. Um I also just love to point
4:23
out a few of the names of the artists.
4:25
Some of them could genuinely be actual
4:27
band names. Callus Humane, Calm
4:29
Knuckles, Calvin Man. Some of them are a
4:31
bit weirder. Um, you've got things like
4:33
Caliop, Eratum, Camel Edible, and a
4:36
calorie event. I think that one's
4:38
definitely going to be what my lunch
4:39
will be like after this, Charlie. It'll
4:42
be a major cal calorie event, I think.
4:44
All right. Fake fake listeners, but a
4:46
lot of real money at stake there. Okay.
4:48
You've also brought us a study that
4:50
shows how AI can be used for good to
4:53
tackle conspiracy theories. Yeah, this
4:55
is quite positive story. Um it was a
4:58
study published in science and uh
5:00
they've basically the researchers took
5:02
more than 2,000 self-proclaimed
5:04
conspiracy theorists. So people in the
5:07
US who for instance believe don't
5:09
believe the moon landing actually
5:10
happened believe the covid-19 vaccine
5:13
gave us all microchips or believe in UFO
5:16
visitation. So um amazingly on average
5:20
each of these people after talking to a
5:22
chat GPT which had been told to convince
5:25
them that they were wrong about 20% of
5:28
them sorry uh most of them on average uh
5:32
reduced their belief in this conspiracy
5:34
theory by about 20%. Um so and that
5:37
actually persisted several months after
5:39
the study was finished. So it did seem
5:40
like according to this uh piece of
5:42
research that they were able to
5:45
permanently reduce these people's belief
5:47
in these conspiracy theories. So that is
5:50
is for instance I think something we can
5:52
say is quite a positive uh use of AI
5:54
particularly given the rise of
5:55
conspiracy theories in young