WEBVTT

00:00:00.000 --> 00:00:02.640
Okay Chelsea, I have two different directions

00:00:02.640 --> 00:00:05.040
this will go and two different stories through

00:00:05.040 --> 00:00:08.199
each direction. So do you want an AI story or

00:00:08.199 --> 00:00:15.820
do you want a plastic story? I'm gonna go with

00:00:15.820 --> 00:00:22.539
AI. Okay, and do you want AI cope or AI humanism?

00:00:23.059 --> 00:00:25.960
Humanism. Okay, this story is... You just say

00:00:25.960 --> 00:00:28.629
the first thing that comes to mind. Okay, that's

00:00:28.629 --> 00:00:32.189
the best way to do it. Yep. Let's do it. Okay.

00:00:32.390 --> 00:00:35.070
How many choices? This article is on futurism

00:00:35.070 --> 00:00:39.890
.com written by Victor Tangerman. March 9th,

00:00:39.969 --> 00:00:43.689
2025 article title. Weird new computer runs AI

00:00:43.689 --> 00:00:46.770
on captive human brain cells and you can buy

00:00:46.770 --> 00:00:50.130
compute on the cloud. Okay. Australian startup

00:00:50.130 --> 00:00:52.390
Cortical Labs has launched what it's calling

00:00:52.390 --> 00:00:56.189
quote, world's first code -deployable biological

00:00:56.189 --> 00:00:59.810
computer." The shoebox -sized device, dubbed

00:00:59.810 --> 00:01:02.869
CL1, is a notable departure from a conventional

00:01:02.869 --> 00:01:05.969
computer and uses brain cells to run fluid neural

00:01:05.969 --> 00:01:10.519
networks. In 2022, Cortical Labs made a big splash

00:01:10.519 --> 00:01:12.840
after teaching human brain cells in a petri dish

00:01:12.840 --> 00:01:15.980
how to play the video game Pong. The CL1 however

00:01:15.980 --> 00:01:18.700
is a fundamentally different approach as New

00:01:18.700 --> 00:01:22.579
Atlas reports. It makes use of hundreds of thousands

00:01:22.579 --> 00:01:25.079
of tiny neurons, roughly the size of an ant's

00:01:25.079 --> 00:01:28.280
brain, each. which are cultivated inside a nutrient

00:01:28.280 --> 00:01:30.620
-rich solution and spread out across the silicon

00:01:30.620 --> 00:01:33.319
chip, according to the company's website. Through

00:01:33.319 --> 00:01:36.340
a combination of, quote, hard silicon and soft

00:01:36.340 --> 00:01:38.719
tissue, end quote, the company claims that owners

00:01:38.719 --> 00:01:41.560
can, quote, deploy code directly to the real

00:01:41.560 --> 00:01:45.060
neurons to solve today's most difficult challenges,

00:01:45.379 --> 00:01:48.219
end quote. Quote, a simple way to describe it

00:01:48.219 --> 00:01:51.959
would be like a body in a box. But it has filtration

00:01:51.959 --> 00:01:55.299
for waves. It has where the media is stored.

00:01:55.680 --> 00:01:57.599
It has pumps to keep everything circulating,

00:01:57.799 --> 00:02:00.500
gas mixing, and of course, temperature controls,"

00:02:00.500 --> 00:02:02.739
end quote. Cortical Labs Chief Science Officer

00:02:02.739 --> 00:02:05.760
Brett Kagan told the new Atlas late last year.

00:02:06.040 --> 00:02:08.360
Whether it will actually prove useful remains

00:02:08.360 --> 00:02:11.039
to be seen, but Kagan is excited for scientists

00:02:11.039 --> 00:02:13.539
to get their hands on the tech. Quote, there's

00:02:13.539 --> 00:02:16.080
so many different options, end quote, he told

00:02:16.080 --> 00:02:18.900
Australian broadcaster ABC News, suggesting it

00:02:18.900 --> 00:02:21.419
could be used for, quote, disease modeling or

00:02:21.419 --> 00:02:24.599
drug testing, end quote, quote. The large majority

00:02:24.599 --> 00:02:27.460
of drugs from neurological and psychiatric diseases

00:02:27.460 --> 00:02:30.300
that enter clinical trials testing fail because

00:02:30.300 --> 00:02:32.419
there's so much more nuance when it comes to

00:02:32.419 --> 00:02:35.379
the brain. But you can actually see that nuance.

00:02:35.340 --> 00:02:38.300
when you test these tools. Our hope is that we're

00:02:38.300 --> 00:02:40.860
able to replace significant areas of animal testing

00:02:40.860 --> 00:02:44.460
with this." Kagan told New Atlas. For now, the

00:02:44.460 --> 00:02:47.740
company is selling the device as a way to train

00:02:47.740 --> 00:02:51.800
biological AI, meaning neural networks that rely

00:02:51.800 --> 00:02:54.539
on actual neurons. In other words, the neurons

00:02:54.539 --> 00:02:57.590
can be taught via the silicon chip. Quote, the

00:02:57.590 --> 00:03:00.189
only thing that has generalized intelligence

00:03:00.189 --> 00:03:03.289
are biological brains. End quote. Kagan told

00:03:03.289 --> 00:03:06.650
ABC, quote, what humans, mice, cats, and birds

00:03:06.650 --> 00:03:09.849
can do. that AI can't is infer from very small

00:03:09.849 --> 00:03:12.430
amounts of data and then make complex decisions,"

00:03:12.430 --> 00:03:15.650
end quote. But the CL1 isn't about to disrupt

00:03:15.650 --> 00:03:18.509
the entire AI field overnight. Quote, we're not

00:03:18.509 --> 00:03:20.469
here to try and replace the things that the current

00:03:20.469 --> 00:03:23.930
AI methods do well, end quote, Kagan added. Nonetheless,

00:03:24.270 --> 00:03:26.229
the approach could have some key advantages.

00:03:26.530 --> 00:03:29.430
For instance, their neurons only use a few watts

00:03:29.430 --> 00:03:32.110
of power compared to the infamously power hungry

00:03:32.110 --> 00:03:34.590
AI chips that require orders of magnitude more

00:03:34.590 --> 00:03:37.479
than that. Apart from selling the CL1, Cortical

00:03:37.479 --> 00:03:40.759
Labs is also looking to sell compute via the

00:03:40.759 --> 00:03:44.039
cloud, using its own assembled racks of the unusual

00:03:44.039 --> 00:03:46.740
computers. In short, while it sounds like an

00:03:46.740 --> 00:03:48.780
exciting new take on conventional computers,

00:03:48.860 --> 00:03:51.360
Cortical Labs still has a lot to prove, especially

00:03:51.360 --> 00:03:53.560
when it comes to teaching neurons not unlike

00:03:53.560 --> 00:03:56.580
an AI. Quote, I know where it's coming from because

00:03:56.580 --> 00:03:59.460
it is clear that these human neuronal networks

00:03:59.460 --> 00:04:02.860
learn remarkably fast. End quote. A University

00:04:02.860 --> 00:04:05.939
of Queensland biologist and cell research specialist

00:04:05.939 --> 00:04:09.879
Ernst Wolvetang told ABC, quote, At this stage,

00:04:10.039 --> 00:04:12.719
I would like to reserve my judgment because learning

00:04:12.719 --> 00:04:15.159
Pong is one thing, but making complex decisions

00:04:15.159 --> 00:04:18.079
is another. End quote, he added, end of article.

00:04:18.980 --> 00:04:21.540
So, how do you feel about them growing brains

00:04:21.540 --> 00:04:25.360
in computer chips to BAI? Well, I could definitely

00:04:25.360 --> 00:04:28.300
use that technology to make decisions about what

00:04:28.300 --> 00:04:30.839
fringy mini we should be doing. it would be great

00:04:30.839 --> 00:04:34.339
for that actually seems to be a complex decision

00:04:34.339 --> 00:04:37.000
to make just based on how this episode started

00:04:37.000 --> 00:04:39.660
out and this is the first time i've used futurism

00:04:39.660 --> 00:04:42.939
for its articles so i'm not quite sure what their

00:04:42.939 --> 00:04:46.699
perspective is on ai but it was very rose -tinted

00:04:46.699 --> 00:04:49.040
glasses they didn't even talk about like the

00:04:49.040 --> 00:04:52.240
ethical weird parts of this yeah which is kind

00:04:52.240 --> 00:04:55.879
of my next thought like Where would that come

00:04:55.879 --> 00:04:58.379
in handy? Like, where would we be using that?

00:04:58.420 --> 00:05:00.920
Would that be in a situation like I believe it

00:05:00.920 --> 00:05:03.439
was a fringy mini we did before where it was

00:05:03.439 --> 00:05:07.000
the AI was going to kill the human life over?

00:05:07.819 --> 00:05:10.879
Was it? It was a pilot. Yeah, that one was really

00:05:10.879 --> 00:05:13.399
dumb because it turns out it wasn't really AI.

00:05:13.579 --> 00:05:16.220
Remember, it was the military one. where to get

00:05:16.220 --> 00:05:18.399
points it was going to attack the target, but

00:05:18.399 --> 00:05:20.439
they were only theorizing it. Yeah, they didn't

00:05:20.439 --> 00:05:22.759
even run it through an AI. And is that like what

00:05:22.759 --> 00:05:25.500
they're theorizing that like, like, where would

00:05:25.500 --> 00:05:27.720
you use something like that? Where would have

00:05:27.720 --> 00:05:30.800
to be capable of complex reasoning? Well, I know

00:05:30.800 --> 00:05:33.199
where you would use it. Obviously, you would

00:05:33.199 --> 00:05:36.399
regrow the brain cell tissue of somebody who's

00:05:36.399 --> 00:05:39.660
died for the seance AI. Like, that's clearly

00:05:39.660 --> 00:05:42.680
where you're supposed to be using this. Right.

00:05:43.129 --> 00:05:47.170
We've done so many AI episodes. 20 minis, rather.

00:05:47.329 --> 00:05:48.949
They're not... Yeah, we've never actually done

00:05:48.949 --> 00:05:52.930
an episode. Okay, I could see where it would

00:05:52.930 --> 00:05:55.910
be, but we already have enough to do that. Yeah,

00:05:56.029 --> 00:05:59.750
but I don't know. This really creeps me out.

00:05:59.930 --> 00:06:02.910
And I don't know how to feel about them growing

00:06:02.910 --> 00:06:07.850
a brain, parts of a brain to run AI and particularly

00:06:07.850 --> 00:06:11.050
where they think the line is where they can stop.

00:06:11.240 --> 00:06:14.860
Growing more person for AI because that's where

00:06:14.860 --> 00:06:17.800
their ethical line is like is it a whole brain?

00:06:18.139 --> 00:06:21.160
Is it a head the whole body? That's okay What

00:06:21.160 --> 00:06:25.100
like would we use them as war robots like the

00:06:25.100 --> 00:06:27.199
implications of that are pretty like what would

00:06:27.199 --> 00:06:30.079
it be used for? Yeah, and then I don't know if

00:06:30.079 --> 00:06:32.980
they're it's pretty sinister and I would worry

00:06:32.980 --> 00:06:35.860
on the other end of it too Where could we turn

00:06:35.860 --> 00:06:39.079
actual humans into AI with their brain tissues?

00:06:39.420 --> 00:06:41.980
Like I was just kind of talking about jokingly

00:06:41.980 --> 00:06:44.839
at the seance, but like, I don't know, could

00:06:44.839 --> 00:06:46.740
there now be a product out there where you use

00:06:46.740 --> 00:06:49.839
your own brain cell tissue that has been replicated

00:06:49.839 --> 00:06:54.000
to make your own AI? You could just become a

00:06:54.000 --> 00:06:56.180
vampire and then that takes care of that. That

00:06:56.180 --> 00:06:59.100
takes so much less ethical obscurity. It really

00:06:59.100 --> 00:07:02.279
does. It really does. I don't think we have to

00:07:02.279 --> 00:07:06.100
say anymore. No, I'm just worried that Zuckerberg

00:07:06.100 --> 00:07:08.420
is going to learn about this and his whole like

00:07:08.420 --> 00:07:11.279
metaverse. What's his AI like world that he wanted

00:07:11.279 --> 00:07:15.019
to do? Meta. Yeah, but like, is that what it's

00:07:15.019 --> 00:07:16.839
called? This is like the virtual world that he

00:07:16.839 --> 00:07:20.079
wanted. Oh, it is called meta now. Okay. Yeah,

00:07:20.079 --> 00:07:23.019
the whole company is meta. But I'm just worried

00:07:23.019 --> 00:07:25.019
he's going to get control of this and say that

00:07:25.019 --> 00:07:27.339
take all of our brain cells and put them in that

00:07:27.339 --> 00:07:29.720
and we're going to have to like live in a virtual

00:07:29.720 --> 00:07:32.750
office for an immortal amount of time. Yeah,

00:07:33.009 --> 00:07:35.410
that's what it is. Meta. Because we're already

00:07:35.410 --> 00:07:38.149
living it. So yeah, you think about it that way.

00:07:38.250 --> 00:07:41.089
It's pretty... I'd rather be a vampire. Oh, yeah.

00:07:41.310 --> 00:07:44.470
This is so much the lamer version of immortality.

00:07:44.850 --> 00:07:48.850
It's totally lame. Yeah. Way lamer. And the real

00:07:48.850 --> 00:07:50.730
weird implications are is how do we know this

00:07:50.730 --> 00:07:53.189
hasn't already happened to all of us? You've

00:07:53.189 --> 00:07:56.009
got 48 hours to figure that out if we're just

00:07:56.009 --> 00:07:59.329
brain cells attached to a microchip. But in any

00:07:59.329 --> 00:08:02.050
event, we'll see you back here as the microchip

00:08:02.050 --> 00:08:03.569
ordains. So bye.