Slashdot Mirror
Gates: Hardware, Not Software, Will Be Free
orthogonal writes "That's small-'f', not capital-'F' free:
according to Bill Gates, "Ten years out, in terms of actual hardware costs you can almost think of hardware as being free -- I'm not saying it will be absolutely free --...." Gates expects this almost free hardware to support two of the longest awaited breakthroughs in computing: real speech and handwriting recognition. He further predicts -- ugh! -- that software will not be written but visually designed."
Ah,
but who will visually debug the visual designer?
"/Dread"
Is this kind of like in the 50s when some expert said that nuclear power was going to make electricity free?
Heck...if I made Bill's salary, I'd already think of hardware as free. In any case, if I was running a company and had global influence, what better model could there be than to dictate that the hardware required to run my product should be (virtually) free, but that my product is too valuable to be expected to be given away. DAve
Nice try, Bill.
He's saying the tangible parts of the system (the hardware) will be virtually free while the freely duplicated software will not be. Fabrication plants cost millions, each chip has a real cost, each resistor has a real cost. Software, once written, can be copied countless times..
You'd think Bill had a vested interest in all this..
Trolling is a art,
Hmmm... This sounds vaguely familiar:
"Remember, quality is our top priority."
.sig wanted. Inquire within.
How much stock can you put in his predictions?
About 640,000 shares - should be a big enough investment for anyone!
SpamNet - a spam blocker that really works
Even if software development becomes putting lego blocks together, it's not going to make specifying algorithms, keeping track of data structures and debugging any easier.
Billy should know better.
The perfect sig is a lot like silence, only louder
"Many of the holy grails of computing that have been worked on over the last 30 years will be solved within this 10-year period, with speech being in every device and having a device that's like a tablet that you just carry around,"
For the last time, Bill...I still don't want a tablet pc!!!!!
Dave
You'll get free TCPA enabled hardware but it'll only let you run software by a certain company, software you'll have to pay for.
And Bill Gates frequently talks out of his ass. I seem to recall that the Web wasn't important (and then we got IE a year later), that MS Bob was going to make computers usable by everyone, and that no one would need more than 640K of RAM.
blog |
> He further predicts -- ugh! -- that software will not be written but visually designed.
"Let's start with a blue background that fills the whole screen..."
Sheesh, evil *and* a jerk. -- Jade
Of course, if you'r stock is worth a few billion dollars, the cost of hardware is 'almost free' :)
--
If code was hard to write, it should be hard to read
Wow another great prediction from the anti-psychic Bill Gates.
Sorry Bill, but software is far more replicable than hardware. It's the SOFTWARE that is becoming more free as we go along.
As far as visual goes, I don't think that's correct. He's envisioning a workflow type application for controlling logic. Diagramming most code is far more difficult then simply writing it. 4GL is a pipe dream.
I DO believe that future programmers will be more like carpenters. High levels of modularity will make custom software construction as practical as cutting and nailing/gluing/screwing together the components down at Home Depot. Programs that ARE sold will be far more extensible (plugin enabled) with managed code.
The future of software is changing. As usual, Gates doesn't have a clue. He was right about ONE thing 30 years ago. He swindled the owners of Q-DOS and IBM. He's been riding that ever since.
-------- -------- Support Wesley Clark for president!!!
This is why you need to study computer science, rather than "programming". Programming is a skill that can be useful, but is, by its very nature, transient. Remember, at one point in time, auto mechanics were considered a very skilled white collar position.
Computer Science on the other hand, is a mathematical discipline which involves working out how to do things better, faster, and with less energy. It's about algorithm design, and ways in which to make a computer most efficiently process mathematical representations. It'll be useful far beyond the use of general "coding".
Coding itself is becoming more and more prevalent. I have many friends who aren't even scientists who know how to code, and were even required to for their humanities classes (from English, to History, to Foreign Language). This is a good thing, IMHO. Coding is a great general purpose skill.
Don't devote your life to the practice of programming, devote it to understanding why certain things work better, and how to further refine our techniques of computation. Work on understanding the hardwaresoftware interface, and you open up all kinds of new fields, from embedded engineering, to robotics.
Take the hint from the majority of good Universities who teach computer science, where you are simply expected to pick up a language in your spare time, because that aspect is secondary to the theory, and the easier of the two.
To be fair, it seems the (in)famous Gates 640K quote is an urban legend.
Step 1: Predict everything that can possibly happen. Nothing is too wild. Some examples:
* In the future, every home will have a robot that carries a machine gun.
* Cars will not only drive themselves, they'll demand equal rights.
* Computers will be made only of light and sound.
* Computers will learn to upgrade theselves - not because of initial programming, but as a survival mechanism to prevent obsoletion.
* IT outsourcing will be controlled in some sectors by organized crime and gangs. This will start in Las Vegas and move outward.
* Email will be beamed directly into your brain. You will be able to type an answer in your head.
Step 2: Wait for at least one prediction to come true ( even slightly true ) and be declared a prophet.
Visual programming is one of the canonical examples of "Gee, I have no clue how it works but wouldn't it be really cool if...". Nobody has a clue how to do significant programming in it; it's never even had a decent prototype, let alone any reason to think it will work in general.
.Net team, who probably are also cringing and shaking their heads privately as well.
Sure, there are isolated instances of it being useful, mostly in drawing flow diagrams for signal processing, but that's far from the general case.
Other then that, though, it's been a miserable failure. Software doesn't look like anything in real life, and real life metaphors are effectively useless for manipulating it. Every tried to use a multi-level UML diagram, where each box contains boxes that contain boxes? That's what visual programming looks like. A confusing, ultra-hyper-dimensional object, where every detail is critical (even the ones you can't see), where to understand a system requires hundreds of little abstract entities on the screen.
Software has more "moving parts", by factors of magnitude, then any other human endeavor; the largest software projects dwarf the complexity (in part count) of even the Space Shuttle. (We get away with it because we use effectively 100% reliable parts, whereas the Space Shuttle does not, the problems that causes and the solutions they require mean the Space Shuttle is still IMHO a superior engineering work to an office suite. Nevertheless, don't make the mistake of underestimating the complexity of software; even the smallest program can dwarf a small car in complexity.)
With a clearer understanding of what is being asked for, it is easy to see that visual programming has been a disaster for fundamental reasons, not ones that can be abstracted away. Imagine the Mozilla source code. It contains megabytes upon megabytes of code. Each and every line must be represented to understand the whole correctly (although no one person may need to understand the whole.) One way or another each line must be represented on the screen; if you're trying to do it "visually", then you're hosed. You can't abstract "(cutcrn*)DO_LOAD((void *)nm_mungl, andlefle->getLumpiness(MAX_LOAD_LUMP_COUNT, (int)uniQuad), USER_MACROS(LOAD));" visually, because you'll either lose critical information, or have an unusably cluttered screen.
There's just no way around it.
"But what if I design special modules that can be hooked together cleanly?" Then you'll have special modules that can be hooked together cleanly, as long as they do exactly what you need, which they won't. We also have tons of experience with such special modules, and they never work completely in general. You can build a DSP out of such things and that's about it... and even then, that's just compositing the existing DSPs together, I wouldn't want to build the insides in a visual language in the general case. (You could get some milage out of it, but you'd still be shelling out to text code.)
You think I'm wrong, you think you have some clever way to reduce the amount of necessary information on the screen without throwing away something the user needs, show me the code. To date, nobody else has managed that, despite a lot of trying by smart and dedicated people, and given that we clearly don't need faster computers to do "visual programming", I think you ought to consider that a damned big clue before you consider punching the "Reply" button and making vague, hand-wavy gestures to the effect that I'm wrong.
Consider the source: I think there's a reason you're hearing this from Bill Gates, who probably hasn't coded significantly in decades, and not the
Hardware will never be free as long as software continues to bloat with an ever-expanding list of core features and peopel continue to believe the MHz myth.
At one level, hardware can already be free. I saw a small PDA with about the same specs as the original Palm Pilot selling for $19.95. Yet such devices are NOT popular because everyone wants the latest wiz-bang features on their PDA.
Its the same reason why laptops get such aweful battery life. I'm sure that someone could create a very functional laptop with a 50 MHz processor that does a competent job running a basic office suite and have superb battery life. As a real-life example, my Psion 5Mx gets 30 hours on 2 AAs and does a great job of basic office work on a 37 MHz ARM processor. You don't need battery-sucking GHz to do the job.
Yet nobody wants to buy "under-powered" devices because they have been trained for 2 decades by Wintel that they must have the fastest machine to get decent performance.
Two wrongs don't make a right, but three lefts do.
Although he claims it will be falling prices, somehow I see the Gillette model creeping in (give away the razor, sell the blades at a premium) - mainly because the hardware will never be "free," as there is always manufacturing cost involved.
:)
Basically, what he's saying is that hardware prices will drop to the point where they can charge for software and give the hardware away for free. I find this quite ironic because it used to be the other way around - sell the hardware at a premium and toss the software in for free.
If I had my way, hardware prices would drop to nothing as Bill proposed, and I'd create free software for it, making it a free-for-all... nah, it'll never happen, but wishful thinking
There are some caveats on this statement however. First, software systems have discrete, digital states rather than analog behavior. That makes them quite succeptible to error behavior in boundary cases. And the state space for software is extremely large. Universal use of components developed in either an object-oriented or functional way could divide this state space up into manageable components. But one issue that is often overlooked by methodology enthusiasts is that this only increases the size of the building blocks and decreases the number of blocks used for a particular size of project. It does not eliminiate the problem that bigger programs are made from a larger number of component parts. The complexity of a program grows as a function of the complexity of the underlying problem. You can change the function with different tools, but the relationship will still exist.
But the visual aspects of pure "compatible" HTML (as in not CSS and Divs, which many design shops still stay away from) are hacks. So you have these editors trying to visually do something that HTML was never intended to do. Dreamweaver, the best of these editors, was oft called "the moody woman" at one shop I worked at, as you had to know just how to coddle it it wouldn't do what you wanted, or even what it was supposed to. Handwriting the code was still superior for these hacks...
Then CSS/Layers became totally (mostly) supported. Now WYSIWYG editors work QUITE well... (Even some non editors generate perfect code. Photoshop's image ready generates some very nice code)
Anyway, point being, when something is designed to be designed visually it can be visually designed much easier. *grin*
Look at it this way: the 48K Apple II was introduced in the US at $1795. Now, a typical bottom-end cell phone has much more computing power. You could put the entire Apple II on a $20 FPGA, or make it an ASIC and the price would be $1 or less in quantity.
I bought a 333MHZ Pentium II based PC in 1998. For software development and everything else I did, it was fine for the following five years. I finally upgraded to a 3GHz P4, just because it was cheaper than upgrading the OS and various parts individually. In my timings, this PC is roughly 15x faster than the old one, plus the video card is at least 10x faster than the one I bought in 2000. This is a lot of power, and it's the least I've ever spent on a PC.
Or consider game consoles. A $150 game system is more powerful, in all ways except memory, than a computer from 5-6 years ago. Video-wise, they're much more powerful. Next generation consoles are going to outrun current desktops...for $200.
The short version is that computers get more powerful, then they get cheap. At some point power ceases to matter, especially if you have a GPU or video compression chip to offload lots of work to. Imagine if a 2 or 3GHz chip could be made to run at 10 watts of power and cost $5. For a 65nm PowerPC, this is reasonable. What's needed is economy of scale. An alternate approach is that "low end" processors in cell phones and digital cameras get to where they're fast enough to usurp a desktop. Then put a video compression chip in there, or other custom hardware to the bulk of the work. At $20 for a complete system, that's a big deal.
Or even consider alternate, custom CPUs. An x86 desktop CPU is expensive because it includes all sorts of junk, like MMX support and 16-bit mode and legacy instructions and SSE2 and all this other marginalized stuff. And still they're too general purpose. C++ doesn't matter any more. Well, it matters because it's "fast," but not because people really like it. C++ doesn't make you happy the way Haskell or Python or Smalltalk do. Take a minimal instruction set designed to support one of those languages, then implement a simulator for it, then an FPGA, then an ASIC. Keep it simple, keep it fast. You could easily have a 20MHz part pacing high-end desktop processors for most tasks. Again, combine this with an ASIC for doing heavy lifting like graphics and compression.
Gates is wrong. Hardware prices don't depend as much on technology but on what people are willing to pay. A PC costs $1000 because that's what people are willing to pay for it, and they happen to get as much hardware and software for that as they can.
.NET-based office suite using COM, DCOM, SOAP, DHTML, and whatnot, on the other hand, won't. But Microsoft has to keep changing things in order to get people to buy and pay them more money.
I'm sure Gates would like the entire $1000 to go to Microsoft, but that's not going to happen. It's not going to happen because Microsoft isn't going to produce $900 worth of software that is capable of running on whatever $100 buys you in hardware. That's not a problem with hardware design, it's a problem with the kind of software that Microsoft develops: big and resource intensive.
On the other hand, you will probably be able to get a really cheap computer that runs Linux and runs it well. We are already beginning to see this with Mini-ITX and Nano-ITX systems: they run Linux so much better than Windows. For $200, you get a full desktop system capable of pretty much everything that a home user needs.
What really helps Linux is that it doesn't have to push an agenda or "innovate" constantly. If a 1995 word processor written in C runs fine on $1000 1995 hardware, it will run really well on a $100 2005 Mini-ITX system, with a few `bug fixes and feature enhancements. Microsoft's new