Why we still can't upload our brains to the cloud

  • As long as people have been alive, they've wanted to stay alive.
  • But unlike finding the fountain of youth or becoming a vampire, uploading your brain to a computer might actually be possible.
  • Theoretically, we already know how to do it. Technically, we have a long way to go.
  • We explain the main technological advancements that we'll need to make whole brain emulation a reality.

The following is a transcript of the video.

Narrator: In the "Black Mirror" episode "San Junipero," Yorkie and Kelly meet and fall in love in a virtual afterlife. And in this world, the possibilities are endless: They can wear what they want, be in whatever decade they want, and die when they want, including never. This technology might actually be more realistic than finding the fountain of youth. So, will uploading your brain to a computer ever be possible?

Randal Koene: I'd be surprised and a bit dismayed if we didn't have it within the century.
Advertisement
Narrator: The concept of uploading an entire human brain - thoughts, feelings, memories, the whole shebang - and running it on a computer is called whole brain emulation. And there are three main areas of technology that need to advance for it to be possible: scanning, processing power and memory, and environment. Let's start with scanning, since researchers will have to start there too. The first bottleneck we're gonna hit is the connectome. The connectome is a complete map of the brain, basically the pot of gold at the end of a rainbow for neuroscientists. Brains are wildly complex. Scientists have only been able to map the complete connectome of one creature, a nematode, and a nematode's brain has about 302 neurons. A human brain has 86 billion, not to mention the 10,000 or so connections each neuron makes with other neurons. Our current brain-scanning tech, like MRIs and MEGs, just aren't good enough to map them all out. There are other methods that could show us a clearer picture, but they tend to be destructive.

Koene: So, the idea there, once you've preserved a brain, then you can slice it very thin and observe everything that's in there. You can reconstruct a lot of what's in the brain from its structure.

Narrator: For living brains, the most promising scanning technology we're looking at is electrodes with thousands of recording sites on them.
Advertisement

Koene: Now, it's extremely invasive, in the sense that there's a lot of surgery involved.

Narrator: These electrodes, when placed on precise areas on the brain, have the potential to show us thousands of neurons at a time. Koene: Because every single site can record from multiple neurons and identify which neuron it is based on the specific shape of action response.
Advertisement

Narrator: Whichever way, the tech has to be capable of scanning and recording massive amounts of data. By massive, we mean zettas of pieces of information. And we don't even know everything we're looking for yet. Many neurotransmitters have yet to be discovered - very necessary pieces of the connectome puzzle. So even we have all the info, if we don't know what we're looking at, then...

Koene: It's like a story that's written in a book, but if nobody is reading it and nobody's understanding the words, then it's just black stuff on white stuff.

Narrator: Right now, each scanned neuron has to be interpreted by humans in real time, zettas of information slowly sorted by hand. If we even want to begin to read the connectome, advanced machine learning and data mining are gonna be pivotal.
Advertisement

Koene: Because otherwise you end up with systems where the learning, the training of that model, just tuning parameters, takes forever. Literally forever. More time than there is in the universe, just because of how exponentially that grows.

Narrator: Computer processing power has also shown smooth exponential growth over the past century or so. Similar can be said about computer memory. If the pattern continues, we'll likely reach the necessary processing requirements fairly soon. But some experts worry we could be reaching the end of Moore's law.

Koene: I also read that, and it also concerns me.
Advertisement
Narrator: Moore's law at one point predicted that computer power doubles every 18 months. And in recent decades, these advances have slowed. If they come to a halt, our computers may never be fast enough to sort through the data.

Koene: It concerns me largely because of that big computational bottleneck that I mentioned earlier, the one where you're trying to fit models, and fitting models is really, really hard.

Narrator: Once we get past that, then we just might be able to get an emulated brain running. One of these WBEs is called a SIM, or substrate-independent mind. The SIM would be an exact digital replica of the mind it was emulated from. And we're actually pretty close to memory requirements, so we should have no problem storing SIMs once we're able to create them. But where they'll be stored is a little bit different than where they'll live, and we do mean live. Would a SIM be conscious?
Advertisement

Koene: Yes.

Narrator: Yes. Simple answer. Koene: That is the simple answer.
Advertisement

Narrator: And if a SIM is conscious, it'll need to exist in a place and interact with things. In "Black Mirror," the characters spend the rest of eternity in a luxurious beach town. Right now, our virtual reality capabilities aren't good enough to create a paradise like that. To experience virtual reality in the way that humans experience actual reality, SIMs would need high-fidelity sensors and systems to sense the world around them. Graphics, at least, are constantly improving, thanks to the speed and size of the video-game industry. It's the rest of the senses that will need major work, everything from tasting a rum and Coke to feeling the pain of a car accident. The bandwidth and signal complexity needed for this shouldn't be underestimated. And while virtual reality isn't the only option for a SIM, actual reality would still require a few upgrades. First, they'd need a robot avatar, and robots are generally really good at only one thing. One of the world's most advanced humanoid robots' claim to fame is its ability to climb stairs. Relatively impressive, but not even close to replicating the experience of having an actual human body. Virtual reality offers a human experience and more. In a digital world, SIMs could fly or teleport or even turn into a lion for a distinctly nonhuman experience. So, once we figure out the technical side of whole brain emulation, there's still the philosophical part of the equation. Would that emulation still be you?

Koene: You have to be very careful about these scenarios when you paint them.

Narrator: The simple answer is: We're not quite sure. Philosophers and scientists have debated, defined, and redefined the concept of "you" for centuries, and they're still doing it.
Advertisement

Marya Schechtman: If you think there's a difference between getting incinerated and having a replica made with a bunch of fake memories and really continuing to exist, tell me what that difference is.

Narrator: Either way, you or another you, a SIM is conscious, which adds a heaping helping of ethics to this dilemma. In the Amazon Prime show "Upload," SIMs can be owned by human beings, giving them the right to delete their SIMs on a whim. Is this murder? Apparently not in the "Upload" universe, but it is a question we'll need to answer in ours.

{{}}