Bret Victor update
worrydream.com
worrydream.com
There are prescription sunglasses for shortsightness, astigmatism, etc.
So the sunglasses-form VR would just have to cater for that. Apple's headset-form VR already does (offers the option to have prescription lenses in it).
Day to day, I wear glasses, and have no plans to change that. If I'm going to a show or a movie, or if I'm planning on spending a lot of time outside, I wear the contacts. The contacts are set up for distance not reading. and that's becoming more important as I get older.
It can be disorienting for a few minutes, I forget how much glasses distort peripheral vision.
I get a small supply of daily disposables. I keep a pair in my laptop bag. With vision as poor as ours, you know how difficult it is if your glasses are lost or damaged. It's not life changing, but it is a genuine quality of life improvement for me.
I encourage you to get a contact exam, to verify you can wear contacts. They should teach you how to put them in and take them out. It can be stressful, but you never touch your eye. The doctor or assistant can answer any questions you have. They're not particularly expensive (in the US anyway). You don't have to wear them. Try them out, they might have some benefits you haven't realized.
That said, I don't think i'd wear contacts just for AR/VR. The headsets already seem like a hassle, and extra tooling to make my eyes use them effectively, even more of a hassle.
How many megapixels equivalent does the eye have?
The eye is not a single frame snapshot camera. It is more like a video stream. The eye moves rapidly in small angular amounts and continually updates the image in one's brain to "paint" the detail. We also have two eyes, and our brains combine the signals to increase the resolution further. We also typically move our eyes around the scene to gather more information. Because of these factors, the eye plus brain assembles a higher resolution image than possible with the number of photoreceptors in the retina. So the megapixel equivalent numbers below refer to the spatial detail in an image that would be required to show what the human eye could see when you view a scene.
Based on the above data for the resolution of the human eye, let's try a "small" example first. Consider a view in front of you that is 90 degrees by 90 degrees, like looking through an open window at a scene. The number of pixels would be 90 degrees * 60 arc-minutes/degree * 1/0.3 * 90 * 60 * 1/0.3 = 324,000,000 pixels (324 megapixels). At any one moment, you actually do not perceive that many pixels, but your eye moves around the scene to see all the detail you want. But the human eye really sees a larger field of view, close to 180 degrees. Let's be conservative and use 120 degrees for the field of view. Then we would see 120 * 120 * 60 * 60 / (0.3 * 0.3) = 576 megapixels. The full angle of human vision would require even more megapixels. This kind of image detail requires A large format camera to record.
Second, that _completely_ ignores touch, which is the largest of the senses. There's a reason car companies are going back to knobs!
That said, it makes more salient for me the dichotomy between demos and reality. This demo looks very much like molecular biology does in the movies.
But what is it to look like something "in the movies"? I don't know exactly... All tools have overhead, no matter how cleverly designed. Maybe "the movies" is what you get when you ignore this overhead.
There's a clear phenomenon in looking back on old demos and concluding we must have taken a wrong turn because our systems are still not as good. This is true of Victor's own demos from a decade ago, and many well-known demos from decades earlier.
But rewatching a Victor demo earlier this year, which had blown my mind when I saw it live, it seemed full of a kind of naivety (and a bit of sleight of hand). Of course, he made clear it was only a demo. It was meant to inspire. But now I feel that's an easy excuse for not investigating what really advances our use of tools – and what holds it back.
> …what really advances our use of tools - and what holds us back.
It’s money! It’s cost! The cheaper tool wins.
“Worse Is Better”
Billionaires still want ROI.
July 4, 2023
Hi
Dynamicland is still going, just quietly. We closed the Oakland space for covid, but Realtalk development and collaborations have continued basically as originally planned, if more slowly due to our small size.
I'm hoping to spend the summer working bionano, and get back to preparing new on the Dynamicland website in the fall. It might be ready by the end of 2023? It'll have everything.
I'll try posting at @bret@dynamic.land (currently Mastodon until we have time to do our own thing in Realtalk). Id appreciate if you don't ask for my Opinions about things.
Thanks, Bret
<img width="1056" height="792" src="July2023.jpg" alt="Hi,
Dynamicland is still going, just quietly. We closed the Oakland space for covid, but Realtalk development and collaborations have continued -- basically as originally planned, if more slowly due to our small size.
I'm hoping to spend the summer working on bionano, and get back to the new Dynamicland website in the fall. It might be ready by the end of 2023? It'll have everything.
I'll try posting at @bret@dynamic.land (currently Mastodon until we have time to do our own thing in Realtalk). (I'd appreciate if you don't ask for my opinions about things.)
Thanks, -Bret">
Just put the plates on a liquid handling robot - stop trying to treat the person as a robot.
The other aspect - the human -> computer ( as oppose to the computer -> human ) interface is interesting - it sort of reminds me of the first multi-touch demos.
ie exploring how best for humans to communicate intent to computers - though again - a bit like the example above around automation - in my view the best UI is no or less UI!
For small experiments, when setup is 90% of doing the job, then you just don't win much by letting a robot do the remaining 10%.
Also, there are usually a lot of steps between the pipetting that the pipetting robot couldn't do like loading stuff in a centrifuge, checking DNA concentration, putting it in the PCR machine, etc.
Programming the robot, and putting the and register the reagents and source material.
Given that they can produce instructions for humans, I don't see why they could produce instructions of the robot ( with perhaps some human instructions for making the reagents and source material etc available ).
A lot of traditional liquid handling robots are large and have relatively large buffer tanks etc - so need flushing etc etc. But I don't see why you could use something like RoboCat to be much more flexible for simple container to container pipetting.
One interface I think is quite good is the spreadsheet - ( and I'm not talking horrible ribbons or poor charting - I'm talking about the core concept of a reactive grid of cells ).
You enter numbers and the computer immediately updates the results - the cells updating is the computer -> human interface, but it feels almost invisible as part of the task. The user has a mental model of how the spreadsheet works and treats it almost like a real world object.
You get a similar effect with the drawing/painting tools which use touch.
http://worrydream.com/LadderOfAbstraction/
http://worrydream.com/#!/ScientificCommunicationAsSequential...
Only dependency is on openCV through GoCV, otherwise pure Go. Supports a webcam/beamer setup. Uses Lisp for scripting instead of Lua (see https://github.com/deosjr/elephanttalk/blob/main/cmd/elephan...).
It's a lot slower/less mature than paperprograms, but it does attempt to implement the wish/claim/when model using a homebrew datalog implementation :)
I feel this.
Nice looking web page also.
Heck, even within one's field, Bret may have neat ideas about interfaces for programming and attained a sort of celebrity status for some of us, but that doesn't mean he can confidently answer random questions from strangers. He's a researcher, not a guru.
Use good science, but run weird experiments!
Jürgen Klopp when asked about this opinion on COVID?
https://www.youtube.com/watch?v=tZ4bKnpxbYc — great answer in a time where every-fucking-body had to share their opinion on it, and still do.
Also the message was great, the combination rarely happens!
I’m not familiar with him or his work beyond the cursory glance at his Wikipedia page that I made after recognizing the homepage from marginalia.nu binges, but I associated the statement with a certain level of exhaustion, or disinclination rather, from the desire to publicly comment on whatever bleeding edge or attention-grabbing topic a man of his profession and experience may encounter. I have no clue what this guy does with certainty, it sounds sort of Alan Kay-like? But based off this post it sounds like there’s a lot of ground he intends to cover with whatever he’s got going on that requires a lot of thought. Thought that he is reluctant to share at a toot’s notice.
Unfortunately, the demo he showed in that talk never really turned in to a product or movement that revolutionized programming as at the time many of us hoped it might.
And, having been one of those impressed with the demo, now I look back on it thinking "this business of software is cargo-culting the whole way down, isn't it."
A particularly simple and amazing tool for this is p5.js, with an editor that, to me, enables exactly the kind of workflow shown here: editor.p5js.org
Whatever you are going to do, you can draw alongside to help you make sense of it better. I do this on my own mind sometimes, but for example visualizing and small neural networks (which I have been messing around with) can't be done on my imagination alone.
I think the important is just that you have some library that makes drawing easy, and put it alongside your work. Of course, it's extra work to make diagrams of everything. But sometimes they really help. I think it's important to simply have this possibility in mind.
(maybe it should have a name? Visualization-enhanced programming? Print-programming? I think (e.g. jupyter) notebooks strive toward this, although tbh I prefer the p5js environment without a linear format)
It should be up to us to decide if and when this extra work is worthwhile. I think really designing (and discovering) things is what benefits the most from this style (i.e. game design, product design, scientific experimentation). So you need programmers working on a design problem and/or designers working on a programming problem; or you could have them side-by-side collaborating (and not just a designer imagining things, and a programmer following specifications to the letter).
It also requires a kind of artistic taste to make things look good, but of course that's something you can develop.
It's a really enjoyable process to be able to get a feel for what you're doing!
(See here: https://editor.p5js.org/gustavo.nramires/sketches/sLA07wpa_ for some personal experiments in visualizing graphs and sparse neural nets, https://editor.p5js.org/gustavo.nramires/sketches/1Bb5VIHi2 for designing an algorithm to efficiently make a roguelike map)
I think you're saying is that we don't get this for 'free'. There's probably no tool (although maaaybe AI could help?) that just instantly visualizes anything you could do. Instead it's up to us to do it :) (although good graphics libraries and a tool that encourages you to recompile often is almost essential imo!)
And here's the kind of art I could likely never just simply imagine to existence, experimentation is essential! https://editor.p5js.org/gustavo.nramires/sketches/pVFFT_8E5
That p5 editor looks nice for this sort of thing. It's important to be able to get the visualizations in quickly, otherwise you risk wasting time.
A tool I've dreamed up, that I never seem to have time to implement, is to send debug information to some form of database, and then being able to query and render that data as you like (from another client). To see bar/line charts of data, spatial visuals, more abstract graphs like in your examples, timeline scrubbers, etc... maybe some day I'll get around to making it.
Anyway, thanks for your comment. I'm interested in this kind of thing.
He was already pretty famous before that from Magic Ink, if not even earlier work.
You can argue that it shouldn’t be zero sum. But time and attention both are, and you can’t focus fully if you’re trying to “grow an audience.”
A lot of views are contrary and he presents his views strongly, so it's kind of off-putting.
I have often envied how (I romanticize) it was before social media and hyper-connectivity, where you might be able to just fade out a little bit to focus on a project. Its a bit hard to get that long term focus when people expect you to be available and able to reply within the hour. Just saying "no" drains enough as it is. I don't have the backbone to put out that message though.
Might not work for you, but I found this surprisingly liberating: allowing myself to just drop the ball, not respond, or just say "Sorry I can't" with no justification, so I can focus my energy where I need it
It's challenging to be a public figure, present yourself on contemporary social media, and not be pressured (implicitly or explicitly) to speak up on everything.
Though I don't read this as being about political or social issues. How else would you have someone politely say, "I'd like to be left alone to do my work, but I'd also like to share with you what I'm working on."
It's the sort of statement that might get someone called out as being somewhere negative on some political spectrum, but I read it here as academic.
Dude probably gets absolutely spammed with, "can you read my paper | review my project | comment on this article?" requests.
Other than that, once zooming a hair, I found it extremely quick and easy to read. For whatever reason, moreso than if it were in a standard font. Three full paragraphs with zero scrolling.
"Though it may not seem to visible in the screenshot, the designer certainly went with the preferred technique of the wooden table / digital camera."
It's been about 5 years since I last played with it.
It’s great fun and should be much more interactive with 5+ years of browser and hardware development.
Fyi for those interested: you need a color printer. You need a good projector. You need a good webcam. I already had the printer, got a projector as a hand-me-down, and shelled out the $ for a good webcam.
It can be a little finicky if you don’t have the perfect setup. For me, I got it to work in an hour or so, once I figured out a good mounting location for the projector. Then I had to periodically fiddle with the webcam settings as the light in the room changed. I’m considering investing in black fabric to improve the contrast.
Also, I suspect much of Dynamicland’s magic is in RealTalk, which is not open source. The Paper Programs repo uses Javascript with a realtalk-esque mini framework. This means that Paper Programs doesn’t quite achieve the goals set out by Dynamicland, ie, that it spills be approachable and imminently learnable to all ages, backgrounds, skill levels. My partner and I are both programmers, and both patient, so we’ve been able to replicate a little bit of Dynamicland’s social magic.
Even getting professional programmers to reason correctly about their program is hard. And getting them to not build "intuitive" tangles that devolve into complete spaghetti is even harder.
Everything I know tells me that the software we consider good and intuitive actually requires an enormous amount of unintuitive design and engineering to pull off. Not appreciating this, and not knowing what you don't know, is a plague in modern application development, the kind which makes people boot up Windows 3.1/NT and remark how utterly instant everything is.
Some kind of paper computing collaborative art project?
This won't tell you who he is, exactly, but it does give insight into how he thinks. And it's a pretty great talk, in my opinion.
Maybe it can be slightly edited to reflect that.
This reduces the scientist or designer to some small cog in the machinery, to the point where they are further alienated from their work.
Remove the scientist and replace them with an algorithm doing some poor facsimile of gradient descent for some protein energy metric. Or make it multiplayer where some crowd is doing the depth first search. No individual.
The only analogy I can think of is cooking. You can reduce all of the steps into a checklist, which can be optimized, but instinct is still the differentiating factor. Or rather, taste.
Maybe this helps designers develop taste better, but imo the interface is not the problem. It's training scientists to have better sense of taste or smell. There's a dimension above all of this that is still yet to be tapped.
Like if these were real processes you were doing in the lab and they were somehow being captured more accurately or efficiently, I could see that perhaps making sense.
But this just seems like extra steps and the novelty is you’re working on a table. It feels claustrophobic somehow.
That is essentially the AR value proposition in it's entirety in my opinion. Meeting the human experience half way, so we can use our immense spatial reasoning skills in an environment that supports them.
I wonder if someone needs to better introduce the SpaceMouse to this crowd.
Similar to teaching yourself to always put tools back in the toolbox immediately after use lest they go for a walk, really.
I mean, I do see your point about -you- not wanting to deal with a bin of parts, I'm not sure I would either - but for the people who've already encountered that problem and found a way to deal with it that works for them, it doesn't seem like it'd be an obstacle.
For years I tried to figure out why my short term memory seemed so good even though I'm neuroatypical in a way that's usually associated with reduced working memory. Turns out I've been building mind palaces basically for longer than I can remember, but not with images, which is why none of the descriptions ever resonated for me. Mind palaces with a blindfold on, if you will.
Being able to rotate items in my brain sure came in handy as a bike mechanic. I'm also handy in a move when you have one box that won't fit in your car/truck. But it also led to arguments with my father about whether I was turning a bolt the wrong way, when wrenching a bolt that is on the back side of something.
Someone once said that Civilization is defined by our ability to think in the abstract.
Scientists spend most of their time thinking in the abstract. Philosophers too. Designers and teachers are translators from the abstract to the real and back, or at least the good ones are.
Giving scientists 'tools' that reduce abstract thought is a crutch, that when overused will lead to injury.
I don't think that's who Bret is targeting, but if it is I hope he goes back to his first principles soon and doesn't end up where many of us are, working to create something and finding we contributed to the creation of something a lot darker.