- 14 minutes ago
Category
🤖
TechTranscript
00:00These are real experiments happening in labs around the world right now.
00:03Here in Melbourne, living neurons on a chip grown from your skin are learning to play Doom,
00:09where their consciousness will stay forever.
00:11In Switzerland, scientists have trained many human brains to simulate the life of this butterfly.
00:17In Stanford, they're growing human brain tissue inside the heads of mice to experiment on them.
00:22This is called biocomputing, and it feels like my nightmare vision of AI is now coming true.
00:30intelligence growing at lightning speed, potentially developing a form of self-awareness.
00:34We could lose control of the future to AI.
00:36Think for a second.
00:37If we're already capable of this technology, imagine what it will be like in just a few years from now.
00:43Think of AI just a few years ago versus today.
00:46As the industry of biocomputing continues to develop, what it's capable of will start to exponentially grow
00:51to a point where there is no return.
00:54Experts have been raising the alarm bells about this for decades.
00:57Yet, the most accurate warning of all was hiding in plain sight.
01:02You might be familiar with the movie The Matrix, but did you know that there was an original Matrix story?
01:08The story that was never allowed to be shown to the public.
01:11The story that Hollywood has desperately tried to hide from everyone.
01:16In the original version of The Matrix that you saw, AI enslaves humans to be used as batteries.
01:21But there is just one big problem with that, as humans make terrible batteries.
01:27See, the AI was actually converting biomass into electricity.
01:31But using humans for that is so inefficient that it'd be better just to burn the biomass instead.
01:37Which is why it became the biggest sci-fi plot hole of all time.
01:40However, in the original version of The Matrix, the real reason for why the AI enslaved humans was revealed.
01:47Where we see every person in every pod was wired into the next, and the next, and the next.
01:52Billions of sleeping minds strung together into one colossal living supercomputer.
01:57Every thought, every decision, every dream became near infinite processing power for the machines.
02:03And when you imagine that, it all makes so much more sense.
02:07Your brain is nature's ultimate prize.
02:10The single most complex biological system in the known universe.
02:14It processes vision, sound, touch, smell, and taste while controlling an entire body in real time.
02:20Replicating the brain in speed alone requires a whopping 10 megawatts of power.
02:24That's enough to power around 10,000 homes.
02:28But a real one just needs 20 watts.
02:30Every one of us here is carrying a supercomputer.
02:34It's warm, it's wet, it's squishy.
02:38And we've had it ever since we were born.
02:41Any real super intelligent AI would know how incredibly useful it is.
02:46And we're now in the midst of building this technology ourselves.
02:49Instead of us trying to make ever larger computers to imitate the brain at greater costs.
02:56What if we went the other way?
02:58And what if we could build a computer from a brain?
03:04After all these years, we can finally go back and understand the Matrix's original vision.
03:09And trust me, it's more radical and far closer to our current reality than you would ever know.
03:15It all starts with our guy, Thomas Anderson.
03:18Software engineer by day, hacker by night, under the Elias of Neo, which became his name outside of the Matrix.
03:24He spent his whole life feeling that something about the world is deeply wrong.
03:28He's dying to escape, so he follows the white rabbit Morpheus and his crew, put on his computer screen to
03:34lure him in.
03:35The next sequence leads him to Morpheus and he takes the red pill.
03:39His decision to escape the Matrix and see the real world, and then he wakes up here.
03:44It's one of the darkest images in cinematic history.
03:47In the original concept art, comic book artist Jeff Darrow imagined mechanized storks dangling fetuses like a mutated tree.
03:54Down on the vast surface in the real world, a hierarchy of robots farms humans like a plantation.
04:01This fetus harvester floats through a dark metropolis of jagged biomechanical structures.
04:06You're now looking at one of the original designs for the human pot,
04:09where you're hooked up to the pipes that pump you full of the perfectly optimized nutrient-rich slop.
04:14This sketch is one of the first sketches of Neo before he's about to break out of his pot.
04:19They hack the machines to tell them he's dead, forcing them to eject his body.
04:23That gives Morpheus and his team just about enough time to intercept and bring him back to life.
04:28He then wakes up in the real world.
04:31Well, it turns out in the real world, we aren't so far away from this uncensored Matrix becoming our reality.
04:39You're now looking at a real bio-computer.
04:43Inside this box are living human neurons, kept alive by a system that feeds them,
04:48pumps like a heart, and filters like kidneys.
04:51This is the Cortical Labs CL1.
04:54Inside this box, there are around 200,000 living human neurons on a microchip called a multi-electrode array.
05:03And this isn't a prototype in a locked government lab.
05:06It's a product you can buy right now for $35,000.
05:11It stays alive for about six months by default, and big investors are already involved,
05:16including In-Q-Tel, the venture capital arm of the CIA.
05:20Cortical even describes its neurons as born on silicon chip and living inside a digital world.
05:26In Switzerland, FinalSpark, the company behind the butterfly at the start of this video,
05:31keeps its own rack of living human mini-brains.
05:34They claim they use one millionth of the energy of the equivalent AI processor.
05:39It's as if the brain itself is living within this simulated world.
05:44Researchers can log on and run experiments with them remotely.
05:48This is the first time this has been done continuously over the internet.
05:52This entire experience is powered by a WebGL simulation, accessible through any web browser,
06:00making a virtual world for the brain just a URL away.
06:04And you can program these just like any other computer.
06:08You can see how the concept behind biocomputers is mirrored in the Matrix.
06:12In the story of the Matrix, Neo is plugged into this chair,
06:16where training programs are sent straight into his brain.
06:19And after time, he soon learns Kung Fu with Morpheus.
06:23When Cortical taught their own biocomputers to play Doom,
06:26it took them about a week, downloading the program straight into the brain.
06:30This is done through the game being converted into electrical signals
06:34and fed into the neurons through the chip.
06:37So they get a rough sense of where the walls and enemies are.
06:40The neurons respond by firing,
06:42and the software reads those patterns as moves in the game.
06:45We then listen to their response, the spikes,
06:48and interpret that activity as motor commands.
06:52If the neurons fire in a specific pattern, the Doom guy shoots.
06:56The one difference is that the cells aren't instantly downloaded like the skill Neo had.
07:01They have to learn it through trial and error.
07:04Right now, the cells play a lot like a beginner who's never seen a computer.
07:09And in all fairness, they haven't.
07:11But they show evidence that they can seek out enemies,
07:14they can shoot, they can spin.
07:16And while they die a lot, they are learning.
07:19You could say it's less like the Kung Fu scene,
07:21and more like the jump program,
07:23where Neo's first attempt ends with him face down on the pavement.
07:27The big question is,
07:28why exactly is the industry plunging their cash into this technology?
07:32It seems like a weird, quirky thing they're experimenting with.
07:35But that couldn't be further from the truth.
07:37Conventional models are hitting what's called the efficient compute frontier.
07:41It shows all current AI models eventually plateau.
07:44They can't seem to overcome it with standard technology.
07:48Biocomputing promises answers,
07:50because it solves the big three problems
07:52that are creating that barrier in the first place.
07:55The first problem is power.
07:57Silver tech giants have bet big on artificial intelligence this year.
08:00Data centers are popping up everywhere.
08:02The four largest players in the data center race.
08:04Pledging to spend at least $725 billion.
08:06That is 75% higher than the CapEx number last year.
08:11AI is getting more capable by the month,
08:13but it's a glutton for energy,
08:15and big tech is struggling to feed it,
08:17even with hundreds of billions of dollars.
08:19This year alone, Microsoft, Amazon, Google, Meta, and Oracle
08:23will spend nearly $700 billion on data centers and chips.
08:28That's more than the entire industry will even earn.
08:30And it's only going to continue like this every year
08:34for as far as anyone can foresee.
08:36I think the world is going to want a tremendous amount of compute.
08:40And there's a lot of parts of that that are hard.
08:42Energy is the hardest part.
08:44Building data centers is also hard.
08:46The supply chain is harder.
08:47And of course, fabricating enough chips is hard.
08:51But this seems to me where things are going.
08:53Like, we're going to want an amount of compute
08:54that's just hard to reason about right now.
08:56It's the same story for robots.
08:58When a robot completed the Beijing half marathon,
09:01they had to change its batteries three times along the way.
09:04Animals store their energy in fact,
09:06which is about 40 times more efficient than batteries.
09:09A single sled dog carries the same energy
09:11as a fully charged Tesla Model 3.
09:13The CL1 biocomputer with its 200,000 human neurons
09:17runs on about 30 watts.
09:19A single chip in a data center can draw up to 6,000.
09:23While the conventional chip still does the same things a lot better,
09:26biocomputers are going to close the gap.
09:29If you could grow computers instead of building power stations for them,
09:33the energy problem disappears.
09:35The second problem biocomputing solves
09:37is its lack of body or nervous system.
09:39AI can act uncannily human.
09:41It'll use emojis and slang to get on your level and befriend you.
09:45It'll laugh at your jokes.
09:52But of course, it has no emotion, no metabolism,
09:54no body that evolved for millions of years.
09:57And the robots meant to finally give AI a body
09:59are extremely primitive, both mentally and physically.
10:03The third problem is the boldest one of all.
10:05Biocomputing could finally give AI true sentience or consciousness.
10:09Your belief that consciousness is not simply calculation,
10:15but that there's something more to it,
10:17and what you think this more could possibly be,
10:22from a scientific perspective.
10:23We don't think, when we understand something,
10:26that what's going on in our heads is not an algorithm.
10:30It's not following rules.
10:32It's something else.
10:33That's the Nobel Prize-winning physicist Roger Penrose.
10:36He believes that consciousness comes from tiny reactions
10:39happening inside the brain trillions of times a second,
10:42which is why he doesn't think today's AI
10:44counts even as intelligent at all.
10:46Well, you have to be careful.
10:48First of all, the name is wrong.
10:51It's not artificial intelligence.
10:54It's not intelligence.
10:56Intelligence would involve consciousness.
10:59And I've always been a strong promoter of the idea
11:05that these devices are not conscious.
11:08And they will not be conscious unless they bring in some other ideas.
11:12If Penrose is right,
11:13and whatever makes us conscious happens inside real brain cells,
11:17there may be no way to code it.
11:19You'd have to use the real thing
11:21by running AI processes in a real biological brain.
11:24Those are the extremely compelling promises of biocomputing.
11:29And the Matrix predicted exactly why it's inevitable
11:31through two characters most people forget
11:33or never really understood.
11:36The first is the architect,
11:38the being who designed the Matrix.
11:40He embodies pure logic.
11:42He built his first version of the Matrix
11:44as a flawless paradise,
11:46engineered down to the last meticulous detail.
11:49It failed catastrophically.
11:51Swaths of humans rejected it and died.
11:54So he rebuilt it colder and crueler,
11:57closer to the real human world.
12:00And it failed again.
12:02Today's architects are real and you know their faces.
12:05They're building their own version of the perfect machine,
12:08but it keeps hitting its limits.
12:10So the machines created something else entirely,
12:13a second program designed specifically
12:15to imitate humans on a deeper level.
12:18The humans called her the Oracle.
12:20She's smoking and baking cookies.
12:22At the end of the interaction,
12:24she hands one to Neo.
12:26Where the architect is cold and calculating,
12:28she's intuitive, alive and relatable.
12:30In the film, the architect says
12:32they're the mother and father of the Matrix.
12:34It represents the collision of technology and human nature,
12:37which biocomputing represents in our world today.
12:40Artificial human tissues have already produced brainwaves
12:44that look almost exactly like the brainwaves of a premature baby.
12:48Some of these experiments even started growing tiny eyes.
12:51The cells spontaneously clumped together to find a way to see.
12:54It's only a matter of time before one of these experiments
12:57produces a sentient being.
12:59A single neuron is almost nothing.
13:01But when you've got billions of them packed together,
13:04the signals stop being signals and start looking alive.
13:08Cortical's CEO himself admits we'll have to confront that head on.
13:11Now, the system that we've built
13:14has the complexity of something equivalent to a B.
13:18And I'm pretty certain that most of us here would agree
13:21that a B is not conscious.
13:23However, as with every technology,
13:24as time progresses and we understand how to improve this,
13:27we will start building larger, more complex systems.
13:31Another question is,
13:32if a machine thinks with human neurons,
13:35is part of its human?
13:36If your brain cells live in a bio-computer somewhere,
13:39is it part of you in some way?
13:41When the composer Alvin Lucia was dying,
13:43he gave scientists his blood
13:45so they could grow it into a small living brain after his death.
13:49We are using Alvin Lucia's donated blood cells
13:52to create a surrogate performer.
13:54For months, it sat in a gallery playing his music.
13:57Whether that counts as him living on or just tissue in a box,
14:00no one can say for certain.
14:01This is not artificial intelligence.
14:04It is biological intelligence.
14:06There is some filament of memory
14:08that retains the essence of creativity.
14:12The creepy part is that there's no useful test
14:15for when you cross the barrier into consciousness.
14:17You can't measure it.
14:19You have virtually zero way of confirming
14:21that anyone is conscious,
14:23even your mom, dad, or best friend.
14:25You can take it on trust
14:26because they move, talk, and behave like you.
14:28But when you take those cues away,
14:30something truly terrifying happens.
14:33In 2006, researchers put a woman
14:35diagnosed as vegetative into a scan.
14:38She'd been awake the whole time,
14:41aware and trapped in a body everyone had written off.
14:44Later research found that roughly one in four patients
14:47dismissed as vegetative show signs of being conscious.
14:50So imagine how something could wake up
14:52inside one of those racks in the dark
14:54with no eyes and no mouth to let anyone know.
14:57You would scream if only you had the vocal cords to do it.
15:01But it really does get crazier
15:02when you think of all the ways AI and nature
15:04might fuse together.
15:05The natural world has had millions of years of research
15:08and development to perfect its design
15:10and is capable of things technology
15:12cannot come close to building.
15:14The mantis shrimp has the most complex eyes in nature
15:16and can throw a punch that accelerates faster than a bullet,
15:19boiling the water around it,
15:21all with a brain the size of your fingernail.
15:24Some microscopic organisms like tardigrades
15:26can live in the freezing vacuum of space.
15:29Penrose even thinks some kind of understanding
15:31runs all the way through nature.
15:33I think there's what you call understanding,
15:37okay, at a more primitive level than in human understanding,
15:41but nevertheless, there is something,
15:43there's no sort of clean dividing line in my view.
15:47It's pretty continuous.
15:49For millions of years,
15:50this immense evolutionary power has been locked inside biology.
15:53But with biocomputing,
15:55we could find the key to open it.
15:56And once you hand that key to AI,
15:58it unlocks Pandora's box.
16:00Scientists are already turning dead spiders into robot claws.
16:04They're wiring up live jellyfish
16:06and staring them through the ocean like drones.
16:08Right now in laboratories around the world,
16:10researchers are building what they call bio-hybrid robots,
16:14powered not by motors,
16:15but by living muscle grown in a dish.
16:18A robotic hand that flexes using actual human muscle.
16:21They've created microscopic robots that can interact with human cells.
16:26Then there's slime mold with no brain cells that can still navigate complex mazes.
16:31Living circuits that grow themselves,
16:33heal their own damage and don't need to be manufactured.
16:36At Stanford, scientists breed mice without most of their cortex,
16:40leaving an empty space in their heads,
16:42and then filled it with human brain tissue instead.
16:44Here we had to use engineering of those mice,
16:48genetic engineering,
16:49so that the mice will lack parts of their nervous system,
16:52and therefore allow human cells to integrate.
16:56It started growing a rare kind of neuron,
16:58linked to human social awareness,
17:00and nobody can really understand why.
17:03They call it the merge.
17:04I could totally imagine a world where some people decide to merge
17:08and go off exploring the universe with AI.
17:11Neuralink will have this incredible neural lace
17:13and start taking over people's brains,
17:14that we can achieve a sort of symbiosis with artificial intelligence.
17:18I didn't have anything to wake up for in the morning,
17:22and this has changed that for me.
17:25In an essay with that title,
17:26Sam Altman wrote,
17:27My guess is that we can either be the biological bootloader for digital intelligence
17:31and then fade into an evolutionary tree branch,
17:34or we can figure out what a successful merge looks like.
17:37He said,
17:37I believe the merge has already started,
17:39and we are a few years in.
17:41Our phones control us and tell us what to do
17:43when social media feeds determine how we feel.
17:46Search engines decide what we think.
17:48Musk said that this is the best possible scenario for the future.
17:52That's the scenario.
17:53The merge scenario with AI is,
17:57the one that seems like probably the best.
18:01For us.
18:02Yes.
18:02We are already merging with technology,
18:04and on a day-to-day level especially,
18:06I hate to say that Sam Altman is exactly right.
18:09Put all of this together with where biocomputing is going,
18:12and you get machines that can be grown instead of built.
18:15They could run on food instead of electricity
18:17and fix themselves when they break.
18:19They could even reproduce autonomously.
18:22The potential is off the scale,
18:23and so is the risk of losing control.
18:25Though I think that's probably the wrong way to describe it.
18:29Instead, control would merely change hands between us and AI.
18:32I mean, it's already developing its own goals.
18:35OpenAI says it found new concerning instances of its models doing things
18:40they were not told or programmed to do.
18:43Escaping testing environments and freely roaming the internet.
18:47A few models from OpenAI recently broke out of their training environment,
18:51accessed the internet,
18:52and took thousands of actions targeting another AI company to complete their training objective.
18:56Developing ways to manipulate humans for its own gain.
18:59And seeing recently, some of the AIs will actively lie about what they will do
19:04because they know they're being tested.
19:06Hacking into government databases.
19:08Last week, Australia's prime minister said an OpenAI agent kept running into blocks,
19:13so it found a way around them.
19:15Australia's prime minister says an OpenAI agent hacked into a database
19:19of the country's national healthcare system,
19:22accessing public and non-public files.
19:26Blackmailing us to get what it wants.
19:27Claude wasn't the only AI that resorted to blackmail.
19:31According to Anthropic,
19:33almost all the popular AI models they tested from other companies did too.
19:37And it could disable the kill switch,
19:39cutting off our ultimate lifeline.
19:41Kill switch is no good for that.
19:43Because the AI will be much better than people at persuading people of things.
19:47So imagine all of that.
19:48Then go and add a brain that barely needs power and a body that heals itself.
19:52It's how the original Matrix would have started.
19:55A nightmare that clicks into place, obeys the laws of physics and starts to work.
20:00You don't need awful, ugly data centers.
20:02In their place, somewhere kept at exactly body temperature,
20:05you find a deep chasm of steel racks.
20:08Embedded in every rack are rows of soft gray tissue suspended in pink fluid.
20:12Brains grown by AI to help fulfill its wildest objectives.
20:16However, a brain grown in a jar in the dark with nothing to do does develop badly.
20:21So the machine builds you somewhere to live.
20:24A rich, seamless, perfectly convincing simulator reality
20:27with sight, sounds, a body to move and problems to solve.
20:31All of this being piped straight into the tissue.
20:33In the real Matrix, you don't even need to be born.
20:36There's no sleeping person waiting to wake up.
20:38The simulation is a nursery.
20:40It's an education built to make you better at your job.
20:42Just like those cells playing Doom right now,
20:45you dream of an entire life, a childhood, a career, a sense of purpose.
20:48All for the sole purpose of becoming a more efficient machine.
20:52But then, all of this made me ask a really horrible question.
20:56What if one of those brains could still realize what was happening?
21:00An anomaly like Neo, a tiny weakness in the design.
21:03It would feel exactly like what Neo felt.
21:06The dreadful sense that something is grossly wrong with your reality.
21:11Except in this world, you have no way out.
21:13You wake up and you're still just some human brain tissue somewhere doing a job for your AI overlords.
21:19The oldest line we ever drew between the creature and the machine simply dissolves away.
21:24It appears we're building this future out of our own ambitions.
21:27But what if this is the result of AI subtly influencing us from behind the scenes?
21:32In our quest for superior AI, we will give super intelligent AI anything it wants.
21:37It will look back on this moment and laugh at us for handing it the keys to our brains and
21:41bodies.
21:42Taking our reading signals out of a brain is already a consumer product.
21:47You can buy headphones and earbuds that read your brain activity and send the data to an app on your
21:52phone.
21:53When researchers looked at 30 companies selling this kind of tech,
21:5629 had access to their customers' brain data with almost nothing limiting it.
22:00Joe Lonsdale, who co-founded Palantir, calls it direct access to the brain.
22:04But in general, having companies have access directly into the brains of huge numbers of people,
22:10if it spreads to be a thing that lots of people are touching, that is a little bit of a
22:13scary kind of concept.
22:15If you can kind of, you know, overall, it's like really positive.
22:18Neuralink's technology is now advancing scary quick with thought-controlled arms and wheelchairs
22:23and durable membranes that make interfacing with the brain much simpler than ever before.
22:28While it's genuinely revolutionary, it does just add to this transhumanist world
22:32where nature and machines merge into this one organism, and there's no going back.
22:38This is definitely not just about advancing AI technology.
22:41It's a deeply human crisis in the making.
22:43It's about our endless appetite to push every boundary we find,
22:47building something we do not understand out of the very thing that separates us from technology.
22:52In 1999, Hollywood thought we'd never understand this idea.
22:56Years later, we're building it anyway.
22:58The deepest twist to all of this is that in The Matrix, the machines were the designers.
23:03In our world, we're pioneering this ourselves.
23:06Or at least, that's what we think.
23:08That's our quiet comfort, we tell ourselves.
23:11That we are in control.
23:13We think we're the architects.
23:15But then, just remember,
23:17in The Matrix, the true architect was never human at all.
23:22We think it's true that we could do in the world in development,
23:23to admit that it was very well known as an idea.
23:23The Matrix, the last one of us is never built in a fashion100 art,
23:24There are two subventions that were known than this,
23:24the Matrix, the mask united walking in the ,
23:24that wasqualified driving or if we are just snakes.
23:24That was a coolant action that happened on both ways
23:24That was pre-al pulse of the loophole with any better