Hacker News Reader

Comment by dang | Hacker News Reader

A 1966 demo of his pioneering Ph.D. work on interactive visual programming: https://www.youtube.com/watch?v=NLyIYmPfCps.

His Ph.D. dissertation is here: https://dspace.mit.edu/handle/1721.1/13474. It was some of the earliest work on dataflow and graphical programming. I know this because Alan Kay told me to read it, so I did. You should too.

Edit: Bert, of course, was Ivan Sutherland's (of Sketchpad) older brother. There's a delightful dialogue with the two of them from 2004: https://www.youtube.com/watch?v=sM1bNR4DmhU ("Mom Loved Him Best: Bert & Ivan Sutherland").

View on HN

Love the PhD. "Thesis advisor: Claude Shannon".

For those that aren't familiar, the TX-2 is one of the systems that led to modern interactive computers. It was far more powerful than you would expect for a computer in 1958, the company DEC was basically founded on this computer's architecture, remarkably advanced computer graphics and image perception was being done (https://dspace.mit.edu/bitstream/handle/1721.1/11589/3395912...)

lincoln labs, where it was hosted, was a fertile area of research at a time when far more people were in favor of academic military research (MIT played a huge role in WWII).

I built a visual programming system for grid computing a while ago and it's interesting how often this paradigm keeps coming up in random places, like music making, blender effects pipelines, etc, etc. You'd think that researchers would design neural network architectures in visual programming space, not in a programming language.

> You'd think that researchers would design neural network architectures in visual programming space, not in a programming language.

Give it a few years, this will definitely happen.

Er, may I ask why you'd think that?

Other than the notion of "layers of neurons", there isn't much about NNs that I would consider graphical - and that term is merely a remnant of the original motivation for NNs back in the day. Today, this metaphor has mostly been abandoned, and we think more about NNs in what they really are: combinations of linear and non-linear functions.

You may be thinking of sequence processing where a basic network is repeated in various different ways. This is already today often displayed as a graph of sorts.

But I don't see how this would obsolete using a programming language. The strength of visually laying out a network architecture is in communicating it to other humans, not to implement it for a machine.

Whether you call the substance of NNs "layers of neurons" or "combinations of (non-)linear functions", I think very often a graphical programming process will be the most ergonomic way to manipulate those essential building blocks.

This is because the building blocks are often "stable" in the software-maintenance sense, and when designing a special-purpose NN most of the hard work seems to be in their configuration and composition. In many projects you'd only ever configure/compose pre-existing functions like "Conv2d" and "Softmax", rather than write them yourself. And the graphical programming model excels at configuration and composition, so it's a natural fit. However, when you have to customize the building blocks, or break away from the "data flowing through functions" pattern, that advantage quickly vanishes.

I'm not claiming that every implementation of a NN would go better with a graphical programming tool. But just like with Audio/DSP tools I can easily see lots of people being able to justify using that kind of thing.

Wow, I think they still have some pieces of that machine on display down in the library at Lincoln.

It was always cool to see it, but I wasn’t aware of the significance until now.

Edit: Looks like I was wrong. The computer in the LL library is more likely the predecessor TX-0.

Isn't that his younger brother Ivan?

No, that is Bert.

Looks like they were doing pretty similar work! Here's Ivan's Sketchpad https://www.youtube.com/watch?v=6orsmFndx_o

Please take the trouble to look more closely at what they did, and when.

At 2:22 in the video, Ivan's manager says (paraphrased): "We are making the computer the human's (engineer's) assistant".

In 2020, he would say: "All engineers are the computer's assistant".

Where did we lose the way or we haven't?

I'm not sure who appears in the video, but the work being presented is Bert's work on graphical dataflow programming.

Reply on news.ycombinator.com