I don't feel like my ability to interface with the computer is a major bottleneck for me. If I'm writing code, for example, my typing speed is really not a factor. Even typing this very short comment, I paused several times to think of how to phrase things.
It definitely is for me. Not all the time - I can type at the speed of thought, and using powerful languages + some completions minimizes the time I spend frustrating over boilerplate - but frequently enough, and especially when conceptualizing and designing something.
When designing, even a whiteboard is a bottleneck for me, and so is pen and paper. I can't draw nearly as fast as I need, and then I can't really manipulate the drawings much either. I dream of being able to conjure 2D and 3D diagrams at will, of manipulating them and connecting them together. When focused, I really do feel like I have limited RAM and a very slow I/O for swap.
(I seem to suffer from aphantasia. I'm jealous of people who can freely visualize and manipulate stuff in their heads and actually "see" the results.)
I'm interested for the same reasons - I can't hold much of an image in my head for very long, and being able to persist what I can imagine so I can examine it in detail with my actual eyes would be extremely useful.
FWIW Tibetan Buddhism has concentration exercises revolving around visualisation that might help develop this ability. Not sure if it would work for 'true' aphantasia.
For anyone who wants to look into that, it's called "kasina practice." There's a chapter about it in Daniel Ingram's book Mastering the Core Teachings of the Buddha, though he describes it as an advanced technique for people who already have a high degree of concentration.
There's also "image streaming" which is similar, and might be more accessible:
From what I know, Kasina practice involves visualising certain simple shapes or objects such as candle flames and coloured disks.
Tibetan Buddhist practices involve visualisation of fairly complex 'deities'; supernatural beings wearing certain clothes, certain ornaments, surrounded by certain other complex objects. As the visual complexity of the imagined scene is much more than in Kasina practice, the 'phantastic' benefits might be different as well. The deities are convenient as visual object because (I suppose) Tibetans are familiar with their visual appearance. Choosing any other similarly complex familiar object should work as well.
Ingram described Kasina as gradually increasing the complexity of the images, after success with simple ones. But I confess I don't really know much about this.
I've stumbled across the "Image Streaming" notion before, on those same few websites. Is there any third-party support for the idea? Such as a published study, slatestarcodex-style book review, or (at this point) a few redditors claiming "it worked for me"...
It unfortunately sets off my BS radar in full:
* Most references to the technique appears to originate from a book called "The Einstein Factor: A Proven New Method for Increasing Your Intelligence", by the same Win Wenger as you linked above.
* Win Wenger is always written as "Win Wenger, Ph.D", but I can find exactly one paper from him on Google scholar [0], written solo, written well after he was already publishing books [1].
* The "Reinhart Study" cited by the book is listed on this website [2] with the heading "Reviewers have found that this study was statistically inadequate. Further and better studies are requested, please contact wwenger101@aol.com".
* The author of that not-published paper is listed as "Charles P. Reinert, Ph.D., Dep't of Chemistry/Physics" (not psychology).
* Other works attributed to Win Wenger on that same site are very obviously nuts [3], containing such gems as "It is unclear at this point whether anyone in fractile theory or in interference pattern physics has noticed yet the relationship between these two fields, each of which by itself is and will be totally transforming our understanding of our world and of ourselves within the next decade or so."
In short, Win Wenger appears to be someone without any actual credibility who is nevertheless attempting vainly to forge it. And I can't find any actually credible references to this technique working.
Imagine a computer interface capable of actualising the conceptual structures that you are imagining.
Imagine that this interface had a feedback mechanism simulating a virtual eye allowing you to perceive a full three dimensional rendering of that concept in real time.
Now think of an arraylist, think of how it responds to changes, visualise the corner cases.
Can we please not adopt this as a potential use case for this topic though? Real applications look more like this:
Imagine a movement. Follow an intense training protocol many times for each possible movement so that the signal processing component actually gets meaningful data it can differentiate and is certain enough that you want to trigger something. After multiple hours of training it might actually get good enough. Input dimensionality for that method would likely be simple action triggers (at least less than a traditional keyboard) instead of a complex concept abstraction.
Alternative: Stare at a board of visual stimuli, repeat for a few hours in order to train the system to your brain. Stare at board of visual stimuli again in order to actually use the thing. Again, slow paced and limited input dimensionality.
While these invasive models produce better signals than non-invasive BCI, they do not include a feedback mechanism and have far less throughput than using a keyboard. Which is great if you have a medical need for the technology and thus cannot use traditional interfaces, everything else is sci fi. I'm not trying to say that this isn't cool but we should really refrain from painting it as some miracle tech, otherwise we cannot really complain if media adopts that image and misrepresents the whole field again.
According to my 1990s graphics card packaging I have a dragon in my bedroom. I wouldn't expect a Musk company to produce underwhelming marketing claims but marketing departments don't change the state of the art in research.
Making this sci-fi levels of usable and getting it to consumers are both pretty unrealistic claims. I'm pretty sure you still couldn't get invasive research BCI past an ethics board around here if the patients skull isn't opened out of medical necessity anyway, doing so for fun isn't analogous to getting a piercing or tattoo, it's analogous to chopping off a limb to get a robotic arm.
the first electric car was invented in 1828, and the electric motor isn't that complicated of a device. Reusable rockets are a little more complicated, but fantasies about plugging the human brain into a computer are to those inventions what a sandcorn is to the sahara desert.
Musk has the habit of taking some established technology like solarpanels or pneumatic tubes and then repackage them in ways that get futurology type people all exited, but that doesn't explain how he's going to hook us into the matrix.
And yet, until Tesla came about we didn't have that many electric cars running around, nor did we have many companies investing heavily to change the status quo of gas cars. The same goes to reusable rockets. As many other people said in this thread, Musk is a PR nightmare and is a very peculiar public person when it comes to his tweets and such, but I wouldn't put it past him to actually find the talent and bankroll this until they achieve their objective.
There's no doubt that Musk has a knack for high risk entrepreneurship and he definitely very well how to repackage and promote things, but Neuralink wants to invent devices that come straight out of the science fiction realm.
Not only do we not have the AI required for natural language processing of the sort Elon is after, we also don't really have the neuroscience that explains how information is stored in the brain or let alone how we "write" to it. That's a different beast.
Sure, I didn't want to imply that they won't make money on this. Just trying to give a bit of perspective versus their marketing :) No matter what I think of the feasibility, a Musk-backed commercial player in the field will likely stir something up. With the pace of progress in the field over the last decades, that might just be a good thing.
> Sure, it's not like a Musk company ever broke the mold before.
Well to be fair they really haven't much, as far as technological advancements are concerned.
Reusable rockets weren't used before not because we didn't know how to do it, but because it doesn't seem to be economical. The judge is still out on whether SpaceX will actually be able to reuse the first stages in a way that's even remotely useful.
And there's even less innovation involved in Tesla cars. The only new thing they bring to the table is Autopilot, which arguably is a huge accomplishment, but not one that's "breaking the mold" in the hypercompetitive autonomous car space right now, and hasn't really delivered on its promises either.
What Neuralink proposes is so far ahead in terms of what we even glimpse to be technologically possible I have trouble thinking it's not basically his pet project that is guaranteed to receive press and funding just because the topic is sexy and it has Musk's name of it. If we are ever to get as far as NL implies, it will take more than a single consumer electronics company.
People who are not impressed by the innovations of Musk’s companies generally aren’t involved in those fields.
I had a NASA fellow (guidance, navigation, etc, so a relevant field) who literally told me landing on a barge is impossible, just months before they accomplished it.
I personally think this will take decades. And the implications of it for humanity if it works are probably greater than anything else Elon is working on.
Because Neuralink wasn't started in order to develop a medical device, it was started with an explicit goal to develop a BCI to ultimately improve the cognitive capabilities of healthy humans.
This has little to do with how code is written. It's so much more than "visualize an array". The gap from brain reading thread to what you described is like the gap between DeepMind and AGI
We're already close to replacing the eyes as a source of visual data.
We've already used brain waves to directly control input to a computer.
What I was talking about was essentially computer aided design with an integrated computer.
And I know how code is written, on many occasions I've known almost completely the structures that I'm going to spend the next two and a half hours translating and typing out.
The minutiae of the code that emerges as I'm typing is often an exception, but I generally have a fair idea of where and what the corner cases are going to be.
The biggest bottleneck by far comes in translating the internally visualised structure into an English derived programming language.
Obviously this is a very personal thing, but that is not the case for me. I often can “see” entire chunks of code in my head, and typing out those 20 lines ends up feeling slow and frustrating.
Of course that’s not systematically the case - in many instances, I am thinking as I am writing the code. But for many straightforward control functions, tests, constructors/destructors, etc. the keyboard absolutely feels like a bottleneck.
(There is the orthogonal debate that perhaps a better language would allow me to express more concisely these things that feel straightforward)
I know exactly what kind of frustration you mean. I think this frustration often comes about, because writing code up is a first quality control and it might seem unnecessary for straight forward stuff to be quality controlled.
While spending less work on trivial stuff seems reasonable, if I do introduce a bug in trivial code it will be harder to track it down, because I will always first check whether or not all the tricky parts work. On top of that you indirectly incentivize yourself for thinking something is trivial, because it would mean less work and somehow my brain is very good at tricking me into doing less work in situations where obviously more work is required.
So, even if it was possible to paste code from the brain I would recommend reading it again carefully and then the speed advantages are gone. The only situation where you wouldn’t need it are really repetitive tasks, but in these situations you can have macros/snippets.
For me too. Something like auto-completion for the brain. Not in the "let me complete this command/file name for you" But in the "let me write immediately and correctly the 20 lines of code for the exact procedure you think you need and you just thought in a second". I know it sounds like it is impossible now, but who knows? Maybe it will start with some macros for the brain.
That might not always be possible. Sometimes I don’t know the full definition of the function upfront; I have to discover/invent it one line at a time and can’t do that without pausing and thinking about the line I just wrote.
Even if the speed is identical to how fast you are thinking of words to type, there are plenty of applications (prevent RSI, accessibility for amputees).
Not at a desk and keyboard, but on the phone? Im excited for brain computer interface to liberate software development from the desk. I like to walk and think but today I have to sit at a computer to implement and verify my ideas. With a bci I could do more than theorize.
I'm also excited offload mathematics to a computer.
I don't feel like my ability to interface with the computer is a major bottleneck for me. If I'm writing code, for example, my typing speed is really not a factor. Even typing this very short comment, I paused several times to think of how to phrase things.
It feels like a huge rate limiting factor for me. Even typing 2x as fast would be a phenomenal improvement.
Speaking is even worse though. Communicating verbally with other humans is frustratingly slow and inaccurate. A new non-verbal means to communicate would enable us to understand others faster and better. This would solve a lot of problems immediately that arise from our inability to understand others or the amount of time it takes to arrive at understanding.
I respectfully disagree. Your first mistake is equating programming to input. Ask yourself how much time you lose to the externalities of coding (ci, context switches, the shell, etc). Consider the xkcd on automation [1] and programmers' obsession with screen real estate. Why do we work so hard to eliminate distractions while coding?
I think this idea that computation/programming is a thing that happens through a window into the world of computing is one of the major concepts of our time that will be laughed at in the future. Solving HCI is (imho) one of the most important problems of our time.
This is not just a neural keyboard. They are also working on write ability to the brain. I am a laymen, and this is far off, but imagine being able to access all of the docs for a given language...
You might think: “what are the params for that function?”
That question would be detected, the answer found online, and would be presented to you as a “memory.”
Don’t take the thought = input too literally, your computer input companion will be an AI system, not a keyboard.
Just think about how Siri’s AI can mostly understand what you’re saying even with background noise, stuttering, pausing, and coughing when talking. Now imagine talking to Siri at the speed of your thoughts.
I've always thought riding a motorcycle is sometimes hard to do because holding onto the bike is at odds with working the controls because both use your hands and feet.
Tasks along these lines would be a great match for better computer interfaces.
You should be gripping your bike mostly with your knees. Also, I think that sense of mechanical coordination is something most motorcyclists actually enjoy.
To refute your assertion more precisely, there is ample evidence that bikes have hand and foot controls that are frequently modified from the current controls. Some electric bikes move both brakes to the handlebars. Many roadracing motorcycles have added thumb brakes. And I've had plenty of situations on dirtbikes where I can't use to the shifter or rear brake because of silliness like water or mud. And then there's broken levers.
afaik it's more about brain to brain interface. I can vividly imagine an horse eating grass in a field next to a river with a sunset in the background, but unless I paint it I have no way to make you see the same thing, and even if I do paint it it won't be a 1:1 representation.
Another example would be feelings, describing love, fear, anger is almost impossible by words. I can also think about sharing dreams, memories, with sounds/scents/&c.
I'm not a huge fan of the thing, I'd probably never get one myself, but I see the potential real use cases.
It'll be interesting to see how things like that play out, once we have finer-grained read/write than we do now. Your eyes aren't cameras, and your memories aren't recordings of the outside world. So if you try to broadcast your mental image of a horse, imagine some extremely lossy compression where the recipient has different horse emojis. Your thinking of a horse in a scene may render to their image of a horse in a scene.
