I believe I saw an article about an AI that was able to decode human vision 'brain-waves' in real-time, which would create a blurry image of what the human was seeing.

This AI Decodes Your Brainwaves and Draws What You're Looking at

Is anyone aware where I can find this?

  • $\begingroup$ This is a fairly niche and very intense interest. I do think an expert analysis would be better. I'm not sure about the ethics of this, too. $\endgroup$ Nov 5, 2019 at 7:25
  • $\begingroup$ (en.wikipedia.org/wiki/…) Is this what you are looking for? I was interested in interfacing video games with CBIs but decided against it (still have those emails somewhere). Also, you can edit the article you saw into the question. [Not medical advice.] $\endgroup$ Nov 5, 2019 at 8:08
  • $\begingroup$ Yes, that's the one @TautologicalRevelations $\endgroup$
    – Albert
    Nov 5, 2019 at 8:50
  • $\begingroup$ I was able to understand you better because of these comments (improvements). Best wishes and good luck with your A.I. journey! $\endgroup$ Nov 5, 2019 at 10:04
  • $\begingroup$ Maybe this question would be better in one of the biology or medicine Stack Exchange websites? It does require both heavy understanding of technology and biology, so I'm not sure. $\endgroup$ Nov 26, 2019 at 8:48

2 Answers 2


Direct Answer to the OP's Question

"Have any AI's been able to decode human vision 'thoughts'" ~ Albert (Stack Exchange user, OP)

This is technology that can produce pictures of what the user is thinking about through scanning a brain.

"Is anyone aware where I can find this?" ~ Albert (Stack Exchange user, OP)

Emotiv is the most accessible commercial model (circa. late 2019).

The OP is probably interested in consumer brain–computer interfaces (also known as BCIs). These are varied technologies which range from:--

  1. Simple "yes-no" brain-interface (e.g. for people in a coma)
  2. Advanced programs that control video games through thought, such as a high fantasy wizard duel (Do a search about this on YouTube!)
  3. Technology that can produce pictures of what the user is thinking about through scanning a brain. (This was mentioned by the OP.)

This Wikipedia page, < https://en.wikipedia.org/wiki/Consumer_brain%E2%80%93computer_interfaces >, compares different models of BCIs.

There also some very serious ethical issues regarding being able to "read brains." I mentioned the medical use; and I hope it goes in this direction.

Deep philosophical discussions can be had on whether it is appropriate to read the brain of a supposed criminal. (I would personally say no.)


[I can't comment on the technical details of this. It is outside of my purview. For example, if you need information about the Python-BCI interface, you will need an expert.]

[This is not medical and/or legal advice. This is theoretical discussion.]


There have been studies in University of Oregon and Kyoto University to be able to visualise thoughts and dreams on a screen using voxel values of an FMRI scan as input and an estimation of an image of the thoughts as the output. Instead of linking you to these studies and papers - you could just watch this episode of mind field where both these studies are demonstrated and linked.

The idea behind this is easier to understand if you have a good understanding of generative networks such as generative adversarial networks or so. Essentially in GAN's you'd map a known latent distribution to images in pixel-space. You would be doing the same thing here, just that the latent distribution would now be the FMRI scan input and the mapping would be made in a supervised setting where they are initially showed images. A very rough understanding of the idea can be drawn on these lines.

  • 1
    $\begingroup$ this sounds pretty sweet, very good explanation of the training i was struggling to conceptualize how they could do this but if they control what the person is viewing when capturing the brain activity that makes total sense, upvoted! $\endgroup$
    – nickw
    Nov 6, 2019 at 20:14
  • 1
    $\begingroup$ Haha thanks! I think its very hard not to be mind-boggled while seeing or hearing about this for the first time. Just to mention - this is what I was able to take from the demonstrations in the video considering all the data they would end up collecting. The paper from Kyoto university is to be published sometime in 2020 from their estimated timeline and I guess we'd know for sure only then. $\endgroup$
    – ashenoy
    Nov 6, 2019 at 20:27

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .