18 ms·

So I've actually been paid to work on two projects of this kind: Apple's Newton, and Peter Neumann's Clean Slate project (https://www.nytimes.com/2012/10/30/sci

So I've actually been paid to work on two projects of this kind: Apple's Newton, and Peter Neumann's Clean Slate project (https://www.nytimes.com/2012/10/30/science/rethinking-the-computer-at-80.html https://www.nytimes.com/2012/10/30/science/rethinking-the-co...). In both cases, for different reasons, and in different contexts, the projects set out to reinvent computing from the ground up. The Newton failed in ways that make it a punchline, but the punchline obscures useful things about how it succeeded and how it failed. Clean Slate seems to have morphed into another research program called CHERI (http://www.csl.sri.com/users/neumann/20191213-ctsrd-ftr-final.pdf http://www.csl.sri.com/users/neumann/20191213-ctsrd-ftr-fina...), but it hasn't taken the world by storm, either. Both projects embodied some fairly radical rethinking of how computing should work. Both were fun and inspiring to work on--among the best things I've done in my career. Neither has conquered the world, though some of Newton's novel features have become commonplace, at least in some form. On the one hand, the fate of these projects is most likely the fate of any similar attempt to reinvent computing. The scale of the existing computing ecosystem is planetary. No small group of visionaries can exert the force needed to move the planet, and no group big enough to move the planet can be cohesive enough to embody a coherent vision. On the other hand, I think it's worthwhile to invest in pipe dreams like these. They sometimes invent compelling new things that go on to make positive contributions. They're also instructive. They teach you what is possible when you ignore inconvenient realities. Smart people given a green field and an inspiring vision can invent some truly wonderful things. They also teach you what is not possible. You aren't going to simply wipe away the existing infrastructure and replace it wholesale. Its gravitational pull is inexorable. At some stage you have to come to terms with what has gone before, unless you're content for your effort to become a footnote. Both things are useful things to learn: yes, you can do something novel and inspiring, and that's well worth doing for its own sake. No, you can't, simply through the charisma of a beautiful idea, undo all that has gone before. Nor, if you could, would that necessarily be a good thing. Finding workable compromises with your surroundings is, after all, a constructive endeavor.

buran77 6y ago
The problem with starting from scratch and doing it completely differently is that inevitably you'll end up researching only to get to the same point (for all intents and purposes). Some things just converge towards the same solution given the set of conditions. So between spending all that (partially duplicated) effort to rethink everything and the fact that the end result would have little buy in due to the high threshold of effort involved with adopting and using it, this is rarely a tempting proposition for many. Either to develop, or to use.
Communitivity 6y ago
Great innovations don't seem to come from just iterated evolution or some big revolutionary leap. Instead it seems to me to come from a series of iterative evolutions followed by a revolutionary idea and a good execution in part of the problem. Often it comes from a sequence of these evolution->revolution chains. When that happens the initial execution of the revolutionary idea is usually buggy and incomplete, but works well enough to demonstrate the idea. Other people then evolve that idea further.
cutemonster 6y ago
> some of Newton's novel features have become commonplace Was that because of the Newton project, or it happened independently?
- mikelevins 6y ago
  A mix of both, I think, and it's hard to be sure which is what.
  - mikelevins 6y ago
    I can be a little more forthcoming than this with the benefit of a night's sleep. :-) The problem with this question is that the answers are fuzzy. Newton had a lot of features that are everywhere now, but that were not when it was being developed. Of the features that have become widespread, it's not perfectly clear which ones have Newton ancestry, which share common ancestry with Newton, and which are mostly independent but possibly influenced to some degree by Newton. There were a lot of ideas swirling around in the late 80s and the early 90s, and a lot of smart people trying them out in different ways. Some of Apple's best people worked on Newton, and afterward went on to work in other places, taking with them their intelligence and creativity and things they learned working on Newton. It's hard to be sure what they did later, or who they influenced with those ideas. Take Javascript, for example. It's pretty similar to Newtonscript in several ways. Is that because Brendan Eich knew about Newtonscript and was influenced? Or is it because Brendan Eich was influenced by some of the same things that influenced Newtonscript's creator, Walter Smith? Things like Scheme and Self, for example, and like the idea of using a syntax chosen to seem friendly to C (or Java) programmers. Someone who knows Brendan Eich could ask him, or Walter might know something about it; he often comments here. Newton had a bunch of features that were wild in 1992, and aren't as wild now. Features like transparent wireless network roaming, seamless migration of data between ephemeral and persistent storage, a purely gestural interface built on machine learning, and a simple knowledge-based help system. It was almost small enough to fit in a pocket--but not quite. I think that's one of the major things that worked against it in the market: it wasn't small enough for a pocket, and it wasn't big enough to have a nice screen. It was exactly in the awkward middle. Its handwriting instantly became a joke, but it was actually pretty good. Once it was trained, it was actually very good, but one of the things that we learned from Newton was that very good is not good enough in a consumer product. Every small step away from perfection loses you customers. (Newton's handwriting recognition was unusually good for me, and for the other Newton engineers, because our handwriting training data was built into the firmware. Any of us could make any Newton look nearly flawless because it already knew us.) Its recognition system was actually more sophisticated than I've allowed: it was in fact several different kinds of recognizers working in parallel, coordinated by a blackboard system. Its ability to recognize drawn shapes and convert them into vector graphics that you could immediately edit in-place was very cool. The fact that you could have all these different recognizers all working together on what appeared to be a single sheet of "paper", doing the right thing, whether it was converting handwriting, recognizing drawings, or responding to command gestures, all interleaved, was even cooler, and I'm not aware of another system that achieves that seamless kind of feeling even now. The first time I ever saw a freely-roaming networked device, it was in Newton's Bubb Road building in Cupertino in about 1992. It was a Newton prototype PC board ribbon-cabled to a Macintosh SE that was being wheeled around the building on a cart so that the engineers could watch as it negotiated connections with different networks. We all imagined a world where you could have a computer in your pocket and, wherever you went, it would just always have network connectivity. That was crazy talk in 1992. Newton probably isn't why we have that now; rather, more likely, Newton was one of the earliest manifestations of a dream that lots and lots of people were dreaming, and that someone was eventually bound to get working. Still; 1992. It was science fiction then. Okay, our present mobile devices are not as seamless as the Newton team imagined, but we do have network connectivity pretty much everywhere. It's so ubiquitous now that people probably don't think it's a big deal, unless they're old enough to remember when we didn't have it.
ellis0n 6y ago
I built Animation CPU from scratch for the same reason https://animationcpu.herokuapp.com/ https://animationcpu.herokuapp.com/

buran77 6y ago

The problem with starting from scratch and doing it completely differently is that inevitably you'll end up researching only to get to the same point (for all intents and purposes). Some things just converge towards the same solution given the set of conditions. So between spending all that (partially duplicated) effort to rethink everything and the fact that the end result would have little buy in due to the high threshold of effort involved with adopting and using it, this is rarely a tempting proposition for many. Either to develop, or to use.

Communitivity 6y ago

Great innovations don't seem to come from just iterated evolution or some big revolutionary leap. Instead it seems to me to come from a series of iterative evolutions followed by a revolutionary idea and a good execution in part of the problem. Often it comes from a sequence of these evolution->revolution chains. When that happens the initial execution of the revolutionary idea is usually buggy and incomplete, but works well enough to demonstrate the idea. Other people then evolve that idea further.

cutemonster 6y ago

> some of Newton's novel features have become commonplace Was that because of the Newton project, or it happened independently?

mikelevins 6y ago
A mix of both, I think, and it's hard to be sure which is what.
- mikelevins 6y ago
  I can be a little more forthcoming than this with the benefit of a night's sleep. :-) The problem with this question is that the answers are fuzzy. Newton had a lot of features that are everywhere now, but that were not when it was being developed. Of the features that have become widespread, it's not perfectly clear which ones have Newton ancestry, which share common ancestry with Newton, and which are mostly independent but possibly influenced to some degree by Newton. There were a lot of ideas swirling around in the late 80s and the early 90s, and a lot of smart people trying them out in different ways. Some of Apple's best people worked on Newton, and afterward went on to work in other places, taking with them their intelligence and creativity and things they learned working on Newton. It's hard to be sure what they did later, or who they influenced with those ideas. Take Javascript, for example. It's pretty similar to Newtonscript in several ways. Is that because Brendan Eich knew about Newtonscript and was influenced? Or is it because Brendan Eich was influenced by some of the same things that influenced Newtonscript's creator, Walter Smith? Things like Scheme and Self, for example, and like the idea of using a syntax chosen to seem friendly to C (or Java) programmers. Someone who knows Brendan Eich could ask him, or Walter might know something about it; he often comments here. Newton had a bunch of features that were wild in 1992, and aren't as wild now. Features like transparent wireless network roaming, seamless migration of data between ephemeral and persistent storage, a purely gestural interface built on machine learning, and a simple knowledge-based help system. It was almost small enough to fit in a pocket--but not quite. I think that's one of the major things that worked against it in the market: it wasn't small enough for a pocket, and it wasn't big enough to have a nice screen. It was exactly in the awkward middle. Its handwriting instantly became a joke, but it was actually pretty good. Once it was trained, it was actually very good, but one of the things that we learned from Newton was that very good is not good enough in a consumer product. Every small step away from perfection loses you customers. (Newton's handwriting recognition was unusually good for me, and for the other Newton engineers, because our handwriting training data was built into the firmware. Any of us could make any Newton look nearly flawless because it already knew us.) Its recognition system was actually more sophisticated than I've allowed: it was in fact several different kinds of recognizers working in parallel, coordinated by a blackboard system. Its ability to recognize drawn shapes and convert them into vector graphics that you could immediately edit in-place was very cool. The fact that you could have all these different recognizers all working together on what appeared to be a single sheet of "paper", doing the right thing, whether it was converting handwriting, recognizing drawings, or responding to command gestures, all interleaved, was even cooler, and I'm not aware of another system that achieves that seamless kind of feeling even now. The first time I ever saw a freely-roaming networked device, it was in Newton's Bubb Road building in Cupertino in about 1992. It was a Newton prototype PC board ribbon-cabled to a Macintosh SE that was being wheeled around the building on a cart so that the engineers could watch as it negotiated connections with different networks. We all imagined a world where you could have a computer in your pocket and, wherever you went, it would just always have network connectivity. That was crazy talk in 1992. Newton probably isn't why we have that now; rather, more likely, Newton was one of the earliest manifestations of a dream that lots and lots of people were dreaming, and that someone was eventually bound to get working. Still; 1992. It was science fiction then. Okay, our present mobile devices are not as seamless as the Newton team imagined, but we do have network connectivity pretty much everywhere. It's so ubiquitous now that people probably don't think it's a big deal, unless they're old enough to remember when we didn't have it.

ellis0n 6y ago

I built Animation CPU from scratch for the same reason https://animationcpu.herokuapp.com/ https://animationcpu.herokuapp.com/