1:24:40 Interesting to hear his view on the iPad: Apple simplified the interface to the point where it appeals to the lowest common denominator but in doing this reduced the flexibility of the device, restricting the potential for creativity and learning on it.
@mechadense
4 жыл бұрын
An especially interesting section here starts at 1:34:30 -- Alan Kays thoughts on modern FPGAs finally getting to a point that in principle can allow people to escape the proprietary hard-coded uninventive stagnant corsett and getting rid of the von Neumann bottleneck, so that the hardware architecture can actually be matched to the software system (programming language) running on it. He doesn't mention it but one thing one could do is e.g. implement something lik a SECD machine (see Wikipedia) for functional programming based on lambda calculus - or category theory - like Conal Elliotts work on compiling to categories suggests.
@danielsmith5626
3 жыл бұрын
I love listening to the people who have false starts when talking about something because you know they're really thinking about what they're saying and what comes out next is awesome.
@user-pk5kq2mu4x
9 жыл бұрын
This talk was incredible!
@miikapirtola2339
9 жыл бұрын
Eye-opening. Important stuff.
@yuriykochetkov
4 ай бұрын
36:04 Today, let's face it, we should be just programming in terms of specifications or requirements. It is hard job to come up with requirements.
@nanthilrodriguez
6 жыл бұрын
did anyone find that ometa in the browser link?
@da009999
6 жыл бұрын
If you need more info. on Ometa, this might help: Ohm is the successor to Ometa: github.com/harc/ohm Here is some more info. on Ometa: en.wikipedia.org/wiki/OMeta
@SoeaOu
8 жыл бұрын
what a great talk
@melancholiquemelancholique9623
6 жыл бұрын
51:39 i can't recognize what did he said, can anyone help?
@GurtTarctor
6 жыл бұрын
"...a much better idea called psuedo-time was invented about the same time as the semaphore, but it was not invented by a person that was held in as high favour as the people who came up with the idea of a semaphore, so that idea just never got used over the last fifty years in spite the fact that it is a superior idea. So that's an example of just picking up one of those idea from the past that got discarded for no particular good reason."
@stevenclark2188
9 жыл бұрын
Does anybody have some links to the stuff used in the demo? A lot more detail would be useful. I sorta get the thesis: Small, higher-level code may produce more efficient machine-code/operation than large hacked-together low-level code even though it's "Close to metal". The high level language allows for correctness rather than a million special cases. But we took a heck of a long time to hint at that here. Seeing what the demos look like would help bring it into focus.
@da009999
9 жыл бұрын
His institute's website is www.vpri.org/. You might have some luck emailing the e-mail address at the bottom of the homepage for a complete list. *Some* of the software he uses in the demos: www.vpri.org/vp_wiki/index.php/Main_Page More background info: www.vpri.org/html/writings.php
@ivanbulanov5754
9 жыл бұрын
Is it without an essence? Or is it just all over the place?
@gigiduru125
5 жыл бұрын
sort of all over the place but not really, the point is to show that the ideal would be to write in very high level languages instead of struggling like worms in C or golang or whatever
@ryanjackson0x
8 жыл бұрын
Anyone have a list of the books he mentions?
@da009999
8 жыл бұрын
+ryanjackson0x This is a complete list he prepared for his students: www.squeakland.org/resources/books/readingList.jsp I am not sure how up-to-date it is. There are different versions of it on google. Use keywords: alan kay reading list.
@ryanjackson0x
8 жыл бұрын
+da009999 Ah. Thank you so much!
@da009999
8 жыл бұрын
Glad I could help. Fun fact: His personal library/collection is 5,000 books. But, I am not sure if he ever posted that list. Source: c2.com/cgi/wiki?AlanKaysReadingList
@SoeaOu
8 жыл бұрын
+da009999 in that list there i think this one is missing: Molecular Biology, the Cell www.amazon.com/Molecular-Biology-Third-Bruce-Alberts/dp/0815316208/ref=sr_1_6_twi_cd__3?ie=UTF8&qid=1450693733&sr=8-6&keywords=the+cell+3rd+molecular
@da009999
8 жыл бұрын
It would make sense books are missing in the list. He has a library of thousands of books, so I would imagine the list is hard to keep updated. The book list is also copied on various sites so it's not surprising books are missing.
@IgorKim-Alyona
6 жыл бұрын
at kzitem.info/news/bejne/1piXvmWJpplno44 Alan talks about semaphores vs pseudo-time (?) anyone has a reference to this? quick duckduckgo does not come up with relevant things.
@monad_tcp
6 жыл бұрын
It's so sad I still have to use C, perhaps I should consider doing my own hardware to get rid of it. C is not fast, it's our current hardware that's useless for other languages.
@ispinozist7941
6 жыл бұрын
The mouse pointer in the middle of the screen
@AndyThomasStaff
7 жыл бұрын
I'm at 1:07:42. Can anyone link to when the talk actually begins? This is fairly incoherent so far. Not yet sure what he's trying to say. Something something humans built an arch, something something someone's throwaway object oriented thesis.
@josipmiskovic
7 жыл бұрын
He argues that even complex project don't have to be complicated if you make right tools to support your abstract ideas. For CS, he claims that we are on wrong path for last few decades, that we should step back and reconsider the tools (programming languages) with which we express our ideas.
@jasonsebring3983
7 жыл бұрын
Alan's talk is pretty complicated and confusing. Maybe that was his goal to demonstrate by example.
@jasonfrazzano
7 жыл бұрын
he is basically describing regional graph databases that directly render. its what webgl will be next year to code 3d holography emotiv device
@jlifts4893
2 жыл бұрын
This video is really long. I think there are faster ways to get the same point across. Going forward, I do not recommend this video to be used as an assignment for class work. Have some compassion for the students.
@fschutt247
5 жыл бұрын
His numbers are greatly exagerrated... Windows is about 50 MLOC, not 120, Office is about 40 MLOC, not 220. Second, the thing about Nile is that they don't care about hardware acceleration at all. It just uses a plain memory buffer, of course it's "simple". Damn caching, to hell with 60fps, let's just use the CPU for everything like we did in 1980. You can do that, but in a real-world application Nile would simply be slow as shit. I've seen these kinds of "simple solutions" fail over and over again, there's a reason such a "simplicity" isn't encountered in the real world. And of course a DSL makes the code easier, still doesn't solve the problem of complexity. But it's like saying that you don't have to write a lot of GLSL in order to shade a triangle - no shit sherlock, it's a domain specific language made to do one specific thing and the complexity is in the driver code. It's just someone elses problem now, but the problem still exists. And I mean wow, he points to a working bacterium with 50 million atoms and says "yep that happened randomly by pure chance, 50 million atoms positioning themselves perfectly to form a working bacterium - happened by pure chance, can't see anything wrong with that theory". Yeah sure, my 50 million lines of code also wrote themselves without a creator and without any design or planning. I just put a computer in a room, waited a billion years and during this time the computer started itself and wrote Google Chrome via the input of cosmic rays flipping bits on the harddrive. Yeah, that surely happened. And to the question in the title: yes, it's complicated because computers are dumb as rocks, and no, we didn't make it complicated. This is just incoherent rambling about how coding is oh-so-complex, and the worst thing, he doesn't even offer a solution to this problem. "We should take a step back" - and do what, sing kum ba yah? That won't get a program working. I mean, he's good at buzzwords, I'll let him have that. "We should switch everything to constraint solvers" - yeah because those don't have runtime complexity or anything, they work by magic and fairy dust and will solve everything. Who needs complexity analysis, let's just use X (in this case constraint solvers), it'll solve everything. Last time I heard something like this it was a guy selling people on how Visual FoxPro was the future. This just your typical academic talk - lots of hot air with no actual solutions.
@mpweiher
3 жыл бұрын
Actually, he understated. Office alone is over 400 MLOC. And yes, Nile is a proof-of-concept. Nevertheless, it is already pretty fast. And what's more, its structure is dataflow, and that's in large part what makes it so compact. Well, turns out that the graphics *pipeline* is also very much dataflow-oriented. So much that the graphics APIs had to be changed (Metal, Vulkan) to more closely reflect that structure.
@absurdengineering
3 жыл бұрын
Nobody says to run this stuff on the main CPU, where did you get that? I think it just shows your preconceived notions. A competent Smalltalk implementation will necessarily translate the serial composition code into parallelizable chunks that can run on the GPU. Contraint solvers on GPUs are extremely powerful due to the abundance of resources compared to a CPU and the only reason they are not used more often is due to the impedance mismatch between what one is familiar with (eg C++ and C#) and the minutiae of submitting that code to run on the GPU. And just as C++ compilers strive to make it easier, so can a Smalltalk platform, except that developing said platform is much easier (and enjoyable, due to interactivity) than working on an LLVM backend.
Пікірлер: 37