Slashdot Mirror
It's Time To Build the Analytical Engine
macslocum writes "John Graham-Cumming is launching a project to finish Charles Babbage's dream and build an Analytical Engine for public display. The goal: inspire future generations of scientists to work on their own 100-year leaps."
Question: If we had such a computer, or artificial intelligence. Would we be aware of it’s existence? I am leading to the idea that our military intelligence would likely have AI and suppress any knowledge of it until it is leaked or we are ready. I like to think that we have the tech now for cars to drive themselves but our society isn’t ready for a leap, so we are getting slow introductions to it – i.e. Microsoft Sync that is only available to those who want to buy a new Ford
Excuse me, I was thinking in relation sense, if we had a sudden leap in computer technology. I can't edit my previous comment :(
the ZX-81. I'm doing "scientific calculations" for Kim Jong Un.
Yours In Novosibirsk,
Kilgore Trout
Is it the Internet?
What today stands out as something that is so immediately useful and complex and ahead of its time that we as humans are lucky to have been around at the very start of?
Several others: Transistors, Fire, Radio, Electricity, Walkman
Unfortunately, one of the patent warehouse companies now holds the patent to the machine, and is asking $37B for the rights....
Check your premises.
What sort of framerate can it run Crysis at?
Reading TFA sent a very real chill down my spine. Who knows what we are overlooking everyday with all the science and engineering going on in the world? The shocking thing about this whole story is that in retrospect, his idea seems obvious and is scientifically sound, but was ignored. The real point I'm trying to make is how much CAD software and man hours will it take to simulate this - but he did it all without even a pocket calculator.
MAD is necessarily a livelock. Therefore it would choose to build up both offensive and defensive systems to react and overcome the matched threat.
A system that decided to stop developing in a MAD situation would be buggy and irrational.
This is much more than just building it for public display. The idea is to demonstrate that it was, indeed, a fully functional device, and to give credit where credit is due.
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
Doron Swade who wrote "The Difference Engine" (the non-fiction book, not the steampunk fiction by Gibson and Sterling) can tell you this:
It's not possible to create The Analytical Engine. Why? Because Babbage never stopped creating the designs. There is no one clean, complete set of designs for the Analytical Engine.
If someone were to build it, they would first have to pick and choose from among Babbage's numerous sketches, then fill in any of the missing bits. It's not a true, 100% authentic, Babbage design, unlike the simpler Difference Engine, which had a clean set of engineering drawings for its creation.
Then its like Difference Engine Forever then?
Neither is at all practical at the moment, but there's no reason they shouldn't work in theory, just like the analytical engine was in Babbage's day. And, like the engine, in 150 years, we'll be talking about how the idea changed everything.
Your assertion is false. Given that you cannot be sure your opponent will not make a first strike except for fear of retaliation, your force must be able to overcome any first strike and respond with enough force to destroy his. Therefore your strategy must be to protect against incoming attacks as well as overcoming enemy defenses. Since you cannot know exactly how your enemy is also progressing, it is absolutely imperative to 1) continue improving first strike capabilities to force the enemy to hold back on its first strike and 2) continue improving second strike capability in case the enemy decides to strike first.
What is false in your assertion is that there is such a thing as true assuredness. The program that decides such a thing exists and stops development and production of new weapons is buggy.
You mean he anticipated Versioning!
"There is no Definitive Firefox! They keep changing it!"
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
There's a mistake here.
He was only "100 years ahead of his time" because, er, well, 100 years passed. But he need not have been. Scientists say that sometimes "the mood of an age" is right for certain things to appear. So if some soft factors had gone the other way, he'd have only been 30 years ahead of his time.
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
not only is your continued assertion false, and mine obviously correct, but YOU ARE AN IDIOT.
that part really adds a lots to your argument
Jehovah be praised, Oracle was not selected
Yes, that's why the first step is "Figure out what the Analytical Engine is". The idea is that they would look over old drawings, use them where they make sense, and fill in the missing bits with whatever would have been available at the time. It would be *an* Analytical Engine rather than *the* Analytical Engine.
dom
Even if the AI would be fully demonstrated to the public, it will be a long time before we will recognize that the computers we have are actually artificially intelligent. It's not hard to see why: think back to the 18th century with its thoughts on black people and today's discussion around what constitutes artificial intelligence. My AI prof summed it up nicely (in the last century, yikes): if it works, it's an engineering problem. If it doesn't work, it's an AI problem.
Those who can, do. Those who can't, sue.
isn’t ready for a leap.... i.e. Microsoft Sync
Don't you mean, stumble?
Haven't you heard the Microsoft Car jokes?
"Lame" - Galaxar
Me too!
I also like to think that they are the ones who actually know the secret recipe of 11 herbs and spices involved in the KFC chicken, but a majority of the people out there couldn't handle it if they knew it.
I was on the fence, but once he pointed that out, how could I not agree? Well, if I was an IDIOT, too, I suppose I'd argue the point.
There's an Analytical Engine emulator available. It's a Java applet.
There's no fundamental obstacle to making a working replica, other than money.
Chicken poop.
Hmm... modern examples of overlooked technologies? Well, an emerging example might be the Memristor. Proposed in 1971, it wasn't until very recently that a practical example was constructed; it remains to be seen if they will remain niche curiosities, or become a common part of common electronic designs.
http://acarol.woz.org/difference_engine.html
YOU ARE AN IDIOT is the new YOU ARE NOTHING, apparently.
What a depressingly stupid machine.
In Soviet Russia, destruction assures you!
You mean sort of like how the facebook wouldn't exist if we had proper privacy regulations in place when it was being created? Now that's a scary thought.
Yeah, we have the tech now for cars to drive themselves. I used to work for a robotics company and we made a number of vehicles that could drive themselves safely (for the most part, anyway).
Anyway, the main thing keeping autonomous cars off the roads today is not some secret government conspiracy, but cost. We built a car for the Darpa Urban Challenge which was capable of driving safely in normal traffic conditions at speeds up to 40mph (and several of our engineers felt confident that it could have handled itself safely all the way up to around 100mph!) The final cost of the car was a little over 1.5Million dollars. A good portion of that cost went into the varied and *extremely* expensive sensors the car required. Our main sensor sensor alone cost about $600,000!
We don't need society to "be ready" for autonomous cars - we need mass production of parts which are currently very, very specialized and costly.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
isn’t ready for a leap.... i.e. Microsoft Sync
Don't you mean, stumble?
Haven't you heard the Microsoft Car jokes?
Jokes aside, the SYNC system is really nice
antipaucity
Imagine trying to explain to someone from the 18th century that we have a bodyless Intelligence lurking in the cloud, they would think we where talking about God.
Depends, are they going to add nipples to the suit?
Imagine a Beowulf cluster of them
Yes just the requirement for simulating it and debugging it says to me that Babbage didn't finish his machine. It smacks of when Bell and Curtis "debugged" Langley's aerodrome to show that he really "invented" the airplane first.
As it is Babbage is known as the father of Computers which he does deserve. Just the fact that he dreamed up this massive machine when he did shows what a great mind he had.
Now building one is a great idea. Shouldn't be too hard to simulate with modern cad and then use rapid prototyping to make the parts.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
What I sometimes wonder in hindsight, is could Babbage's machines have been built with the technologies of the time using different techniques that would have been more easy to achieve. He was always pushing the envelope of machining technology with axles and gears, partly in an effort to gain speed. (For example, in the difference engine the gear system had a very complex look-ahead carry feature to make it much faster.) The machines required very tight tolerances and a good deal of force to operate.
Gears work mostly on compressive forces. If instead he had built a machine based mostly on tension, like pulling strings wrapped around wheels and cogs, would the machines have been more practical to build? The machines might have been one or two orders of magnitude slower. However, the problems he was after, like computing logarithm tables, are highly parallelizable. Instead of trying to create one super machine (and never succeeding), would he have been better off with making a bunch of much slower, easier to build machines?
There is an entire scientific discipline (cognitive science) devoted to the creation of an AI. It is nowhere near succeeding. Unless the US military has managed to perform its own research (and I mean including basics like underlying philosophy which isn't even settled) then it is not possible for the US military to be harboring an AI. I know this seems possible from the outside because they get so much money... but money can't really make a few closed door researchers produce something more significant than an army of thousands of researchers sharing their data (academia) unless the money is giving those closed door researchers access to requisite hardware for the science. Hardware isn't currently the problem with AI. Currently, the problem is just figuring out what the "I" in AI even means.
This may well be the most retarded slashdot comment I've read. Ever.
Yeah, we have the tech now for cars to drive themselves. I used to work for a robotics company and we made a number of vehicles that could drive themselves safely (for the most part, anyway).
How well did your system handle pedestrian detection? Because, I recently (less than 2 years) attended a talk by someone from the Stanford autonomous cars lab. I was just about to get my license, so I asked if the system could pass the driving test. The answer was that it could not, because pedestrians would not be detected by the system. And that "for the most part, anyway" is trouble too.
Responsibility is an addiction
Virtue is a temptation
Community is a cartel
" I am leading to the idea that our military intelligence would likely have AI "
Every one in the military know that everything they get is brand new and way better then what is available in the civilian market. Not only that but we all know about the extrodonory budgets the military have to randomly develope AI. The most impressive part is the militaries ability to not only keep it secret but to completely build it in house without contracting it out like it does with 100% of non AI IT projects.
" I like to think that we have the tech now for cars to drive themselves "
Since humans have perfected driving and perfected the art of teaching driving they now have the knowledge base to teach robots to drive just as well. Computers are getting good at driving cars, but I dont think you are interested in paying for one. Last I heard the robots for darpa driving challenges usually had 10s if not 100s of thousands of dollars worth of just optics.
" our society isn’t ready for a leap"
Yeah because if modern business has tought us anything is that coporations will completely hold back on making money to allow society to adjust to changes.
That is what I really like about this brave new world. Humans have perfected their own skills, coporations are thinking about the needs of society, and the government is good at keeping secrets.
I'm sure that if someone was going to order 10k of those sensors, they could get them made for $5k.
A successful API design takes a mixture of software design and pedagogy.
... haven't the Science museum in London already done this?
America, Home of the Brave.
The DARPA Grand challenge has been won (easily and quickly, I admit) only a few years ago. Give people some time to adapt it into new cars. It take around 5 years from a design concept to a production car.
The Wise adapts himself to the world. The Fool adapts the world to himself. Therefore, all progress depends on the Fool.
37647 storage locations ought to be enough for everyone!
Now cue the beowulf cluster jokes...
There is a great difference between somebody who had a great idea, but was overlooked and somebody who blew it.
Babbage was the latter.
When he showed people a small prototype of his difference engine, they knew exactly what kind of potential it had. The TFA even said that the government backed him. I'll stop the press and let that sink in. The British government knew at the time just what a game changer this could have been. What TFA article doesn't say is the extent to which they backed him. In the prices of the day, they invested the equivalent of a fully kitted out and manned battleship in the project. A battleship. What happened?
Babbage squandered the money, fell out with every metal-smith in the country capable of building the difference engine and committed the ultimate crime of changing his mind and plans time and time and time again. Sure, he had a lot of plans for the Analytical engine, but he couldn't stay focused/act civilly enough to build the machine everybody wanted to begin with. After such an investment and nothing to show for it, nobody would give him the time of day, let alone commission him to build an even more complex machine with an unfinished design.
It could be said, rather than a man who had a great idea that wasn't realised. Babbage had a great idea that he killed so badly via his own incompetence, nobody touched it for another 100 years.
#13 has come to pass.
General Relativity: Space-time tells matter where to go; Matter tells space-time what shape to be.
The Ni Polar Option
military intelligence would likely have AI and suppress any knowledge of it until it is leaked or we are ready.
Assuming by "AI" you mean "HAL", I find it highly unlikely that the military or any other secret government agency would have it, let alone be able to keep a lid on it.
First, people aren't very good at keeping massive secrets.
Second, and most importantly, whlile it is true that governments employ some very smart people, and have access to some great resources. They don't have every smart person, or even a sizable minority, especially in computer science. No. If something like this would ever happen (and I kind of have my doubts, not necessarily for any metaphysical reasons, but rather for sheer practicality (i.e. no one would bother to even try), I would think it would come out of academia.
Third, since HAL AI is so difficult, and there are lots of smart people, if science got to the point where it was achievable and desirable, multiple people would notice that it was possible, and then multiple AIs would be created. Why? Because that's how it is with every scientific achievment.
John Mauchly, who knew a thing or two on the subject, made the same observation in a conversation that touched on the issue. It wasn't so much that Babbage was pushing the day's technology too far, he just never froze his design.
Credo sim. - I think I am.
Same with Aliens from outer space/other dimensions!
It isn't the analytical engine, but it works ... today ... and can be seen at the Computer History Museum in the SF Bay Area. http://www.computerhistory.org/babbage/
The first step of analytical engines is you don't talk about analytical engines!
Oh, wait...
well, yes, in that sense, he was most certainly the first computer programmer, as he was apparently always unsatisfied, and RECOGNIZED that he could never be satisfied with his design, so much that he could not produce a single stable design. He would have been welcomed heartily at Microsoft.
I watch a lot of cable news, and have trouble recognizing any natural intelligence.
Jesus was all right but his disciples were thick and ordinary. -John Lennon
"If someone were to build it, they would first have to pick and choose from among Babbage's numerous sketches, then fill in any of the missing bits. It's not a true, 100% authentic, Babbage design, unlike the simpler Difference Engine, which had a clean set of engineering drawings for its creation."
I believe there was also the issue of deliberate flaws in the drawings. In order to protect his works, Babbage (or his engineer) created different sets of drawings and they would have different flaws included. You could build a machine from a particular set of drawings, but it would instantly seize up if you tried to run it.
'The tyrant will always find pretext for his tyranny.' - Aesop's Fables
And the monkey, of course. Polite. Tips his little red hat. Capuchin. Ah, yes... Nostalgia for those bygone ne'er-do-wells of the Brddish Upper Crusk, bleedin' oddities that they were, and all.
``Tension, apprehension & dissension have begun!'' - Duffy Wyg&, in Alfred Bester's _The Demolished Man_
More than 150 years in coming. Beat that, 3D Reams.
Yes, that's why the first step is "Figure out what the Analytical Engine is". The idea is that they would look over old drawings, use them where they make sense, and fill in the missing bits with whatever would have been available at the time. It would be *an* Analytical Engine rather than *the* Analytical Engine.
dom
And then, bam! Dino DNA. :)
Ada magic rondevu.
I would think that the main thing keeping them off the road is liability issues, especially if there was a loss of human life, where an accident was caused either by software malfunction, or simple inability for the computer to anticipate certain drivers who do something that might not necessarily be very kosher or even legal, where an experienced driver learns how to "read" other drivers around them based on observations and naturally gives the most clearance to the ones that could be suspicious.
File under 'M' for 'Manic ranting'
8^) For obvious reasons, we couldn't do very extensive testing with pedestrian detection. Suffice it to say that HR felt compelled to convene a meeting with the engineers regarding the proper usage of interns...
:)
From what *ahem* testing we were able to perform, out system detected pedestrians just fine. We were using a laser-based detector from whose data we used to create a 3d mesh of the world in real time. The mesh was then compared with camera images to determine the location of obstacles and roads. If I recall correctly, I believe Stanford's solution only used lasers to determine "flat surface" vs "non-flat surface" and mainly relies on their cameras to determine the location of the road. So it may have trouble detecting a narrow moving obstacle ahead of (or just on a collision course with) the vehicle.
As for "the most part, anyway" we did have some situations in which our detectors failed. Chain-link fences were tricky (we almost always detected them, but not 100% of the time) and - though other types of vehicles presented no problems - our VP's new Mustang was consistently invisible.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
Maybe, but I doubt it. Sensors are precision instruments and each one has to be calibrated and tested to within an inch of it's life. That process only becomes "cheap" when you can afford to massively parallelize it. If you sell ten of these sensors a year and each one takes a week to be tested/calibrated properly by a trained engineer, they certainly cost a pretty penny. But you can't just pump out 10K and expect the engineer's time to suddenly get cheaper.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
That's certainly a concern, but the GP was positing a conspiracy to protect the people from the dangers of tech we aren't "ready" for. I agree that liability is a problem, but it's not an insurmountable one. Autonomous vehicles *are* coming. They won't be approved for public roads until the government is convinced of their (relative) safety, but they are coming.
No one is avoiding this research because of liability concerns.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
In mass production, all of the testing is going to be 100% automated. All the cost will be in developing a test/calibration jig. The testing/calibration will probably take a couple minutes, and be done wholly by an automated system. The engineering will be NRE cost to develop the jig, to run it you shouldn't need anything more than an industrial technician.
A successful API design takes a mixture of software design and pedagogy.
already available on display at Computer History museum in Palo Alto ?
Over 2,500 people have so far pledged to support the project to build an analytical engine. Minimum pledge is $10. John Graham-Cumming is looking for 50,000 people to pledge to start the project. You can pledge at http://www.pledgebank.com/babbage and get further info at http://plan28.org/.