Telepathy is Finally HERE! & This is EXACTLY how they did it | An Intuitive Introduction to Brain-Brain Interfaces | BrainNet

Michael Ye
Oct 9, 2019 · 11 min read

No, this isn’t one of your classic mainstream media clickbait articles about telepathic communication without any clear explanation of what the researchers actually did.

Instead, we’re going to take a deep dive into the realm of Brain-Brain Interfaces (BBIs) and fully explain the technical details of the BrainNet research paper published 6 months ago, build the BrainNet from scratch, and analyze the realistic implications of the paper. (also, there’s a surprise at the end…)

A High-Level Overview — Collaboratively Playing Tetris Telepathically

In a nutshell, the researchers created a bidirectional Brain-to-Brain Interface (can read and write to the brain at the same time) between 3 people (referred to as a triad) and had them collaboratively play a Tetris-like game.

Two people in the triad had the role of a Sender, and one person had the role of a Receiver. Only the Receiver can interact with the game by deciding whether to rotate the block by 180° or not rotate at all.

The twist is, the Receiver cannot see the bottom row; consequently, he/she is unable to make an informed decision.

Therefore, in order to win the game, the Receiver must rely on the Senders to telepathically communicate to them the correct answer.

The Receiver gets to make a decision to rotate twice per game. A second turn is in the game design because if the Receiver made the wrong decision previously, the Senders can essentially tell the Receiver to correct it, this utilizes a feedback loop to achieve higher win-rates.

And indeed, the setup worked as the avg. win-rate was 81.25%, which is significantly better than a random chance of 50% win-rate. (p-value=0.002)

Now, here’s how they did it, with the science…

The Basics

Before we can get into how the BrainNet works, we must start with the basics and understand its building blocks. Feel free to skip through the sections you already know!


Your brain is made up of ~100 billion nerve cells called neurons and they can have connections with other neurons.

For simplicity’s sake, you can think of neurons as dams in rivers, and the connections between neurons as the water in the river.

Each neuron has an input and an output, meaning the neurons control the flow of water a.k.a signal. And if the neuron has two or more connections, those signals are essentially added together. By default, the dam/neuron does not let any signal pass through, therefore the total signal from Neuron #1 and Neuron #2 does not get to Neuron #4 as it was blocked by Neuron #3.

However, sometimes the signal is really strong…

This time, the signal is passed through to Neuron #4 as the signals from Neuron #1 and #2 reached a certain threshold (can be visualized as water flowing over the dam). This passing on of information is called an action potential.

Now every time an action potential is achieved, it emits an electrical pulse and passes on the information onto the next neuron. Keep in mind of this emission of electrical pulse.

Reading from the Brain: BCIs

BCI stands for Brain-Computer Interfaces, and it’s exactly what the name suggests: extracting information from the brain and inputting it to a computer. The method we’ll focus on today is electroencephalogram or EEG.

EEG headsets use sensors called Electrodes to measure electrical activity in the brain. However, it’s not as simple as reading the action potential of individual neurons because 1) there’s 100 billion of them, and 2) the technology isn’t precise enough.

Therefore, instead of measuring individual neurons, you measure a group of neurons firing together at a certain region of the brain. (E.g. one electrode measuring the collective electrical activity of all neurons near the forehead)

An analogy is dropping TNTs and measuring the seismic activity or the magnitude of earthquakes. If you have a few TNTs and they explode (action potentials) here and there and try to measure the seismic activity over a whole 10-square-mile region, chances are, you’ll measure 0.

However, if you have a million TNTs detonating all at once in one place, you’ll measure a huge number for the magnitude.

So, by measuring the voltage (magnitude of the earthquakes) you can get a single value for the electrical activity of your brain in a certain region. Now you can take this measurement 300 times a second, and you get a wave of EEG signals coming from your electrode.

This surprisingly resembles audio waves, in fact, many audio analysis tools such as FFT (fast Fourier transform) or spectrograms can be used for EEG signal processing. In a way, Electrodes are essentially microphones for your brain’s electrical activity. And an EEG headset is just a whole bunch of microphones for each specific region of the brain.

However, the signal itself doesn’t tell you anything, you must decode it to get the result you want, e.g. the decision of the sender. It’s sort of decoding an audio wave file for speech recognition, but we’ll talk about decoding methods later.

Writing to the Brain: CBIs

CBI or Computer-Brain Interfaces essentially do the opposite of BCIs: directly inputting information into the brain. The method we’ll focus on today is transcranial magnetic stimulation or TMS.

If an EEG headset is a ‘microphone’, TMS is a ‘speaker’ — Instead of recording action potentials, it invokes them by stimulating the brain with a ‘speaker’ blasting strong magnetic fields at it.

But again, just like for EEGs, the technology isn’t precise enough to stimulate individual neurons, therefore you can only stimulate a specific region of the Brain. (Think back to that TNT & earthquake analogy)

Woohoo! Now that we’re done with the basics, let’s go on to build BrainNet from scratch!

Building The BrainNet

Step 1. Build a one-way telepathic communication channel a.k.a Brain-Brain Interface:

Using a BCI (more specifically EEG) you can get a real-time signal of the electrical activity of the brain on the left.

This EEG signal can then be decoded by a computer. In this Tetris game context, the signal will be decoded into a binary decision — ‘rotate’ or ‘not rotate’, and sent via the internet to the computer near the Receiver.

Then, the decision is encoded and sent to the CBI (more specifically TMS), and transmitted into the receiver’s brain.

We’ve just successfully built a brain-brain interface — a one-way channel to communicate telepathically!

Step 2. Make the Receiver’s Brain-Brain Interface Bidirectional

Bidirectional just means you can read and write to the brain at the same time — Allowing the senders to telepathically communicate the correct decisions to the Reciever (writing), and also allowing the Receiver to communicate his final decision (reading) back to the senders.

That can be achieved as simple as adding a BCI on the receiver’s end, and the process is essentially the same as how the sender’s decision was transmitted as described above:

“This EEG signal can then be decoded by a computer. In this Tetris game context, the signal will be decoded into a binary decision — ‘rotate’ or ‘not rotate’, and sent via the internet to the (Sender’s) computer”

Step 3. Add the Second Sender

Finally, we can now add a second sender. In fact, due to the simplicity of this design, you can easily scale it up to 3, 4, or even 100 senders! And… that’s it! We’ve now built BrainNet from scratch.

But wait… How was the signal decoded & encoded?

Oh yes, you caught me. I never explained how the EEG signals from the sender were decoded into the decision of ‘rotate’ or ‘not rotate’, and how this decision was once again encoded for the TMS to tell the Receiver the decision.

The reason I left this at the end is, it completely takes the ‘sexy telepathic magicalness’ out of the story. But fine… here’s how EXACTLY they did it.

Decoding: Steady-State Visual Evoked Potential (SSVEP)

When you turn a flashlight on and off really fast at someone’s eyes, their brain’s electrical activity or EEG signal follows a pattern — the frequency of the electrical activity is always a multiple of the frequency of the flashing light. Given this knowledge, you can easily differentiate whether or not someone is looking at light that is flashing at a frequency of e.g. 15hz vs 17hz.

If you want to sound smart, you can call the above technology Steady-State Visual Evoked Potential or SSVEP. Here’s an example of someone typing using an SSVEP-based keyboard.

Wait… So all the BCI is doing is having 2 lights flashing, one at 15hz and another at 17hz, then depending on which light the player looks at, it will correspond to a different decision?

Yep, this is the screen the Sender sees to “telepathically communicate” with the Receiver:

Okay, that’s pretty disappointing. Now the encoding and writing information to the brain is got to be a little more sci-fi than this…

Encoding: Phosphene Induction

Remember how the TMS can stimulate different sections of the brain? It turns out if you stimulate the occipital cortex (part of the brain responsible for vision), you can induce a ‘phosphene’, which is just a fancy way of saying making a flash of light appear.

And surprise… that’s all the CBI does, it makes the Receiver see a flash of light if a Sender says to rotate, and no flash of light if Sender says to not rotate.

The Truth Behind BrainNet

The first reaction I had when I got to the end of the paper was…

WHAT THE …? THAT’S IT? I THOUGHT THERE WAS ACTUAL TELEPATHY! Can’t the SSVEP just be replaced by eye-tracking software? And why bother to induce a phosphene when you can just show a flash of light on the monitor instead?

I was especially disappointed to see the truth after reading some of these hyped up articles with titles like “Brains of 3 People Have Been Successfully Connected, Enabling Them to Share Thoughts”

However, despite the harsh expectation vs reality paradigm shift, this paper is taking the right step towards true telepathic communication, and here’s why:

1. Demonstrated a scalable framework for an ‘internet of brains’

For the first time, scientists came up with a framework to collaboratively solve a task via only Brain-Brain Interfaces, which can also be scaled up to allow an entire internet of brains. While the way they did it was cheap, that was only because of the Brain-Brain Interface’s hardware limitations…

2. The Upgradability of Brain-Brain Interfaces

In the future, there’s no doubt BCIs and CBIs will get better, along with better approaches than SSVEP and Phosphene induction. Given this framework, you can just switch the new hardware in.

The Future of Telepathic Communication

While it’s easy to get clickbaited into thinking telepathy is already here, it really isn’t even close. There are some major problems to overcome before a full-fledged thought-transmitting system can come into reality. Here’s exactly what we need to work on:

The BCI Hardware:

Right now, the most reliable non-invasive approach we have is EEG headsets with SSVEP decoding. Eventually, we want a system that can directly record a signal that is detailed enough for a computer to decode direct thoughts. Some areas to look into are increasing the preciseness of electrodes — spacial resolution; and increasing the speed of measurement (sample rate) — temporal resolution.

Just to give you an idea, the default sample rate of an mp3 file is 44100 times per second. That amount of precision allowed us to create speech recognition software! Now compare that to EEG signals which at most samples at 500 times per second, the signal we’re getting is just way too uninformative to perform any meaningful inference on.

Computer-Brain Interfaces:

BCIs have been getting a lot of hype recently in the media, but CBIs aren’t getting the same attention it deserves. Our current method of TMS and phosphene induction cannot do anything more than just providing a binary decision or morse code at best. Again, having higher precision for TMS would help, but also we need to understand the brain better which is what the field of Connectomics is looking at.

So… What was the Point of this Article?

Was it to get your hopes up about telepathic communication being already here and then pull a sudden plot twist? Maybe…

But really, this article was to educate you about the field of Brain-Brain Interfaces and give you a direction of where the field is heading and what problems we need to solve in order to make telepathy a reality. We’ve talked about:

  • How your brain works (neurons)
  • How to read and write to the brain via BCIs and CBIs
  • How to build BrainNet from scratch
  • The best approaches we have right now: EEG -> SSVEP, and TMS -> phosphene induction

So although telepathy isn’t here yet, we do have a path forward and this was a step along that direction. Anyways, this article was more about education than telepathy, so go ahead and use what you learned to build something cool. Some of my friends at The Knowledge Society have even built brain-controlled toy cars!

👋 Hey! I’m a 16-year-old entrepreneur, speaker, & Machine Learning/Python/C++/Web developer located in Toronto.

If you have any suggestions, ideas for projects, or just want to connect, feel free to message me on LinkedIn( or email!

If you enjoyed reading this article, don’t forget to hit that Clap button below and share this with all your Medium friends!

The Paper: BrainNet: A Multi-Person Brain-to-Brain Interface for Direct Collaboration Between Brains
Linxing Jiang, Andrea Stocco, Darby M. Losey, Justin A. Abernethy, Chantel S. Prat, & Rajesh P. N. Rao

Michael Ye

Written by

machine learning developer, entrepreneur, life long learner

More From Medium

Top on Medium

Ed Yong
Mar 25 · 22 min read


Top on Medium

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade