CreateDebate is a social debate community built around ideas, discussion and democracy.
If this is your first time checking out a debate, here are some quick tips to help get you started:
Arguments with the highest score are displayed first.
Argument replies (both in favor and in opposition) are displayed below the original argument.
To follow along, you may find it helpful to show and hide the replies displayed below each argument.
To vote for an argument, use these icons:
You have the power to cast exactly one vote (either up or down) for each argument.
Once you vote, the icon will become grayed out and the argument's score will change.
Yes, you can change your vote.
Debate scores, side scores and tag scores are automatically calculated by an algorithm that primarily takes argument scores into account.
All scores are updated in real-time.
To learn more about the CreateDebate scoring system, check out the FAQ.
When you are ready to voice your opinion, use the Add Argument button to create an argument.
If you would like to address an existing argument, use the Support and Dispute link within that argument to create a new reply.
One could work in secret. The vast majority of people probably haven't seriously considered the possibility of such a chain of events. And will not, until it's too late.
There is a non-zero possibility that somebody somewhere in the world has been sitting in a basement for the last ten years, making significant progress on this dark project.
Also, most people lack the skills necessary to carry out such a task. Computers act like an ability multiplier. That is, the utility of a piece of software tends to be a multiple of the ability of the programmer. Thus we would expect to see a widening divergence in individual capabilities as time progresses. The possibility of one becoming unstoppable is forever increasing.
It doesn't matter how powerful your computer-powered multiplier of power becomes. The first time he eats a pizza with something good on it, has a beer or a little bit of grab-ass with a woman not wearing a beekeeper suit, Muslims will commit to blowing up and/or beheading him, and by sheer number alone, they will succeed.
I'm pretty sure you're joking... But still, if one has the power to acquire and manipulate knowledge sufficiently well, then one can learn to protect himself from bombs and such. My house was recently afflicted by ants. Massive amounts of ants. Yet their numbers did not amount to much after the strategic placement of a few of these.
While it is possible that an individual, given the unlikely environment might be able to achieve something like this, it is unlikely because there would always be a group of similar-minded people working together for the same ends, and a group is more capable at such a task as designing technology.
The group would have the advantage and arrive first at any technology which could achieve their aims.
The other important factor is the cost of the technology itself. Computers and machines require special supplies and tools to manufacture. A single person would either not have access to these things, be in a position where he could have access but it would draw suspicion towards him, or build the tools himself which could also blow his cover since these tools require special compounds which would become a spectacle if one person wished to obtain all of them.
When it comes to programming, it is generally accepted that large groups make for slower development (see Mythical Man Month). As you add each person, their marginal contribution to the overall productivity of the group decreases. The sweet spot seems to be around three to seven people. My point here is that the difference between the effectiveness of an individual and the effectiveness of a group is not nearly as large as it may seem. One sufficiently skilled, sufficiently obsessed individual could beat a group. See Richard Stallman vs. Symbolics, for example. Furthermore, as I have mentioned, one could work in secret until he had amassed a degree of power others couldn't match.
And even if you replace the individual with a small group, the story is basically the same. An oligarchy instead of a dictatorship.
The other important factor is the cost of the technology itself. Computers and machines require special supplies and tools to manufacture.
A large amount of computing power can be obtained relatively cheaply. I'm typing this on a $300 netbook with 1 GB of RAM. Way more than enough to do all but the most complex computations. And one can use services such as Amazon's cloud computing to do heavy duty work amazingly cheaply. The bottleneck to human capabilities is not hardware, it's software -- by a huge margin.
When it comes to programming, it is generally accepted that large groups make for slower development (see Mythical Man Month). As you add each person, their marginal contribution to the overall productivity of the group decreases. The sweet spot seems to be around three to seven people.
You just contradicted your initial premise. You are relying upon a single person. A group is more than one person and it is more productive than a single person in large programming tasks. This is why we have teams of programmers developing operating systems, drivers, and complicated software like games.
You simply cannot expect a single person to program a modern operating system in an acceptable time frame. MenuetOS is an example of this, an operating system programmed mostly by one person which has taken three years to have USB 2.0 support. It is still a bare-bones operating system not ready for desktop use in any productive sense. If you follow opensource software, you note that individuals are effective at starting projects but quickly lose ground after a year or two when the project grows in complexity, and if no one else has joined their team.
My point here is that the difference between the effectiveness of an individual and the effectiveness of a group is not nearly as large as it may seem. One sufficiently skilled, sufficiently obsessed individual could beat a group. See Richard Stallman vs. Symbolics, for example. Furthermore, as I have mentioned, one could work in secret until he had amassed a degree of power others couldn't match.
Operating systems are too complicated and downright arcane to be programmed by individuals in the modern world. Computer games are not as complicated or esoteric but a single-man game has tradeoffs in art, music and sound, or interface as a general rule. Drivers may be simple enough for an individual to program, but can quickly grow in complexity with the hardware. Try talking with device driver developers to hear the horror stories of growing spaghetti assembly code.
And even if you replace the individual with a small group, the story is basically the same. An oligarchy instead of a dictatorship.
You keep missing the simple point here though, and that is if it is a group there will be a much greater chance of exposure. Further, it is most probable that there would be many groups working together on similar technologies.
A large amount of computing power can be obtained relatively cheaply. I'm typing this on a $300 netbook with 1 GB of RAM. Way more than enough to do all but the most complex computations.
I'm going to guess that you are a normal computer user with a statement like that.
If you want to solve problems that are so well-beyond everyone else that you could force them into submission, then you will need to build your own semiconductors. You will need a factory. There is a reason that we don't have armies of genetically enhanced supermen fighting us, that is because protein folding is complicated and requires more than supercomputers at this point. The problems that must be solved are beyond the ability of wealthy corporations to solve in an acceptable timeframe, we are a decade behind the computing power necessary to solve quantum problems and protein folding problems efficiently.
And one can use services such as Amazon's cloud computing to do heavy duty work amazingly cheaply. The bottleneck to human capabilities is not hardware, it's software -- by a huge margin.
Remember one could work in secret until he had amassed a degree of power others couldn't match? Cloud computing would expose this.
Have you ever tried to write really efficient software? For simple problems it is difficult, complicated software multiplies this because you're optimising for hundreds, or even thousands of algorithms, in assembly code. Then a corporation builds a quantum computing core and makes all your work moot.
You simply cannot expect a single person to program a modern operating system in an acceptable time frame.
Not with the current means of developing software. My argument is that if one developed tools to make software development easier and used those tools to develop better tools and so forth, the productivity gained from such a cycle would eventually enable them to do something as difficult as writing an OS on their own.
If you follow opensource software, you note that individuals are effective at starting projects but quickly lose ground after a year or two when the project grows in complexity...
1) The money issue plays a big role here. People get distracted by needing to earn a living. One could solve this either by A) Being rich, having passive income, etc. or B) living extremely cheaply (if you can avoid paying rent, you can live on practically nothing.)
2) That's why I suggested developing tools to make software development easier. These tools would likely be largely oriented toward making this exploding complexity more manageable.
Operating systems are too complicated and downright arcane to be programmed by individuals in the modern world... Computer games...
Again, you seem to be arguing from the current state of software development. I'm arguing that one could change the way software is developed and thus fundamentally redefine notions about what one person is capable of.
You keep missing the simple point here though, and that is if it is a group there will be a much greater chance of exposure.
It is unlikely but possible that one individual would be willing to ruthlessly sacrifice short-term benefits (making money, living a normal life) in exchange for greater long-term benefits (power gains achieved by improvements in productivity) over a long period. It is even more unlikely that this person would be able to find others to work alongside him. Further, as you say, this individual could reason that working with others would increase his chance of exposure. Thus he could decide to work on his own until he had enough power to oppose any group.
If you want to solve problems that are so well-beyond everyone else that you could force them into submission, then you will need to build your own semiconductors...
I don't believe that's necessarily true. The limits of software to improve one's capabilities are unknown. The possibility space of what one can achieve even on relatively low powered computers is huge and largely unexplored.
Also one would not necessarily need to make a jump overnight. One could work in secret until he had enough strength to carve out a sort of beachhead, and then work from there.
There is a reason that we don't have armies of genetically enhanced supermen fighting us, that is because protein folding is complicated and requires more than supercomputers at this point.
I would argue another reason is that computer's have only been around for ~50 years and people haven't quite grasped the extent of their power. If I can write software that finds, parses, connects, visualizes, and displays information then I can learn faster than others. I can learn about the limits of human intelligence and how to push them or circumvent them altogether. I can learn all about protein folding and state-of-the-art algorithms which define the speed at which such computations can be done. Perhaps I can find a way to improve them. Perhaps I can focus on ways of obtaining power which don't require supercomputers until I reach a point where they become accessible to me. Use your imagination.
Remember one could work in secret until he had amassed a degree of power others couldn't match? Cloud computing would expose this.
No it wouldn't. Do you think Amazon monitors each of the millions of transactions that passes through its servers? Agendas can be hidden.
Have you ever tried to write really efficient software?
I don't think efficiency is that important. The bottleneck in software development is not the speed at which a program runs or the amount of memory a program uses. The bottleneck is the mind of the programmer. This is why people are moving toward more powerful, dynamic languages like Python even if they are slower than C. This is why people use C instead of assembly. When I talk about developing software tools to develop software tools, I'm talking about developing tools which augment the limited capabilities of the human mind. It is here that we find the potential for explosive gains.
Not with the current means of developing software. My argument is that if one developed tools to make software development easier and used those tools to develop better tools and so forth, the productivity gained from such a cycle would eventually enable them to do something as difficult as writing an OS on their own.
High level language is limited. The cost of inventing a high level language that speeds up software development is a limit to the number or pre-programmed libraries which you can use to build your solution, and libraries are intrinsically slower than low level programs.
Compilers may be advanced so that they are smarter at optimising code, but they lack a sophisticated AI which is capable of human judgments which are often necessary for specific hand-optimisations.
Both of these are extremely complicated tasks which would take a single person years and they wouldn't be able to keep up with everyone elses' technology.
1) The money issue plays a big role here. People get distracted by needing to earn a living. One could solve this either by A) Being rich, having passive income, etc. or B) living extremely cheaply (if you can avoid paying rent, you can live on practically nothing.)
2) That's why I suggested developing tools to make software development easier. These tools would likely be largely oriented toward making this exploding complexity more manageable.
The big issues for making software development easier is intelligence in the HLL, and the consequent smart algorithms chosen. This is a task that a single person would be unable to solve in a respectable time frame.
Again, you seem to be arguing from the current state of software development. I'm arguing that one could change the way software is developed and thus fundamentally redefine notions about what one person is capable of.
It is a hardware issue, not a software issue. If it were as simple as a software issue, we would program an OS in Python or C#. An operating system is complicated because you need to access specific hardware functions which are poorly documented, and case by case. You also have very little program memory to work with when writing your bootloader, which must function as a bootstrap into reading your filesystem and loading the programs from there.
It is unlikely but possible that one individual would be willing to ruthlessly sacrifice short-term benefits (making money, living a normal life) in exchange for greater long-term benefits (power gains achieved by improvements in productivity) over a long period. It is even more unlikely that this person would be able to find others to work alongside him. Further, as you say, this individual could reason that working with others would increase his chance of exposure. Thus he could decide to work on his own until he had enough power to oppose any group.
Simple rational problem:
Group risks exposure versus individual is lost behind modern technology.
I don't believe that's necessarily true. The limits of software to improve one's capabilities are unknown. The possibility space of what one can achieve even on relatively low powered computers is huge and largely unexplored.
You are not talking about simple programs like genetic algorithms and games. You are proposing state of the art software, which is clearly limited by current hardware, as you might experience if you ever try running benchmarking softwares on your platform of choice.
I would argue another reason is that computer's have only been around for ~50 years and people haven't quite grasped the extent of their power. If I can write software that finds, parses, connects, visualizes, and displays information then I can learn faster than others. I can learn about the limits of human intelligence and how to push them or circumvent them altogether. I can learn all about protein folding and state-of-the-art algorithms which define the speed at which such computations can be done. Perhaps I can find a way to improve them. Perhaps I can focus on ways of obtaining power which don't require supercomputers until I reach a point where they become accessible to me. Use your imagination.
Computing power is a limited resource like anything else. This is why we cannot solve complex problems on laptops no matter how optimised the software is.
No it wouldn't. Do you think Amazon monitors each of the millions of transactions that passes through its servers? Agendas can be hidden.
Of course it's monitored. When you enter someone else's domain you are monitored by them using auditing software, which is designed to sort through all that data and make it comprehensible.
When I talk about developing software tools to develop software tools, I'm talking about developing tools which augment the limited capabilities of the human mind. It is here that we find the potential for explosive gains.
Which in the real world is bottlenecked by the silicon you are running it upon. Explosive gains are impeded by slow hardware.
Yes, but they are only one of a wide variety of tools one can use to speed up development.
Compilers may be advanced so that they are smarter at optimising code
I'm not talking about building compilers. Programs generally run fast enough as is. I'm talking about building tools to augment human mental processes.
The big issues for making software development easier is intelligence in the HLL, and the consequent smart algorithms chosen.
Not sure what you're saying here. Intelligence in the language? What does that mean? Are you talking about AI turning English into machine code? Yes, I agree that will be next to impossible any time soon, both for individuals and huge groups. My argument is that there are better ways of speeding up development time that don't require solving hard AI.
It is a hardware issue, not a software issue. If it were as simple as a software issue, we would program an OS in Python or C#. An operating system is complicated because you need to access specific hardware functions which are poorly documented, and case by case.
I would say it's neither hardware or software that's the bottleneck in developing an OS. As you say, the problem comes from handling lots of obscure details. The issue is the human mind's slowness in diagnosing lots of niggling issues. I'm saying software can be written to facilitate the process of handling obscure details.
Group risks exposure versus individual is lost behind modern technology.
I don't understand this sentence. Are you agreeing with me? Either an individual or a small group could follow the path I outlined initially. Either would have different strengths and weaknesses.
You are not talking about simple programs like genetic algorithms and games.
I am talking about simple programs. Those would be a starting point. Would you not agree that simple programs could improve one's productivity significantly? Could one not use that increased productivity to develop slightly more complex programs which yield slightly greater increases to productivity?
You are proposing state of the art software, which is clearly limited by current hardware...
That is just wrong. There is a huge amount of useful software waiting to be written which does not push the limits of existing hardware. The reason those programs don't yet exist is not because hardware isn't good enough, but because humans haven't found time to build them.
Of course it's monitored. When you enter someone else's domain you are monitored by them using auditing software, which is designed to sort through all that data and make it comprehensible.
I think you're underestimating the difficulty of finding the needle in this particular haystack and overestimating Amazon's willingess to do so. Of course they collect aggregate data about usage -- but pinpointing individual user activities? I seriously doubt it. And as I said, agendas can be hidden.
Which in the real world is bottlenecked by the silicon you are running it upon. Explosive gains are impeded by slow hardware.
No, the human mind is a much greater bottleneck than hardware capabilities. Here are some things that could make me more productive: a better mindmapping program, a better find and replace program, more intuitive ways of collecting and organizing data, eye tracking software that doesn't suck, better content search of my own stuff, a less shitty version of this website. I could go on all day. Which of these things are limited by processor speed? None of them.
Yes, but they are only one of a wide variety of tools one can use to speed up development.
They are presently the primary tool.
I'm not talking about building compilers. Programs generally run fast enough as is. I'm talking about building tools to augment human mental processes.
You aren't going to design advanced software that can design iterative generations, when it's slow as sludge because of poor code.
Not sure what you're saying here. Intelligence in the language? What does that mean? Are you talking about AI turning English into machine code? Yes, I agree that will be next to impossible any time soon, both for individuals and huge groups. My argument is that there are better ways of speeding up development time that don't require solving hard AI.
Most of the problems you mention require or are critically benefited from a true AI. Mapping the mind, following the eye, recognising speech, and so on.
An intelligent HLL is one that can choose the best instruction code for tasks which are programmed using a language and syntax which doesn't impede the human thought process.
I don't understand this sentence. Are you agreeing with me? Either an individual or a small group could follow the path I outlined initially. Either would have different strengths and weaknesses.
I am saying that your proposition is a rational choice where the best decision is the one with the least risk versus highest productivity.
I am talking about simple programs. Those would be a starting point. Would you not agree that simple programs could improve one's productivity significantly? Could one not use that increased productivity to develop slightly more complex programs which yield slightly greater increases to productivity?
Up and until you can automate the problem-solving process used to program code in the first place, you're looking at the same problem repeated with ever more complicated software tools.
The bottleneck isn't the human mind, for it can always learn and develop solutions, but the computer's inability to solve its own problems. We must always do its thinking.
That is just wrong. There is a huge amount of useful software waiting to be written which does not push the limits of existing hardware. The reason those programs don't yet exist is not because hardware isn't good enough, but because humans haven't found time to build them.
If you're talking about ruling the world that requires state of the art software to solve scientific problems faster than a team of government researchers.
I think you're underestimating the difficulty of finding the needle in this particular haystack and overestimating Amazon's willingess to do so. Of course they collect aggregate data about usage -- but pinpointing individual user activities? I seriously doubt it. And as I said, agendas can be hidden.
You cannot encrypt the code which the processor runs.
No, the human mind is a much greater bottleneck than hardware capabilities. Here are some things that could make me more productive: a better mindmapping program, a better find and replace program, more intuitive ways of collecting and organizing data, eye tracking software that doesn't suck, better content search of my own stuff, a less shitty version of this website. I could go on all day. Which of these things are limited by processor speed? None of them.
You would still be impeded by the hardware because it cannot think for itself. Until it can, you're just building steps up to realising AI.
AI, researcher software, protein folding, quantum decryption, these are all advanced problems necessary for beating a government, and they all require superior hardware.
Think of IDEs and all of their constituent mini-tools. Think of text editors, mindmappers, lint like programs, project management software. These are all ways of making development faster which don't involve high level languages. I don't see how can say HLLs are "the primary tool".
You aren't going to design advanced software that can design iterative generations, when it's slow as sludge because of poor code.
Being slow as sludge tends to have more to do with high level architecture than anything else. Obviously if a program run too slowly to be useful it's not worth much, but the thing is most programs don't need to be all that efficient. It doesn't matter if an operation takes .1 seconds or .0000001 seconds if my mind can only operate in 1 second intervals. Surely you see my point here?
Most of the problems you mention require or are critically benefited from a true AI. Mapping the mind, following the eye, recognising speech, and so on.
I think you misunderstood what I meant by mind mapping software.
Eye tracking doesn't require AI, it's just a matter of hooking up a webcam, detecting pixels, dealing with noise and then turning that into OS level actions.
I don't know where I suggested "recognizing speech", but turning speech into text can be accomplished without AI.
An intelligent HLL is one that can choose the best instruction code for tasks which are programmed using a language and syntax which doesn't impede the human thought process.
Well that could include a wide range of software of varying levels of sophistication. Some versions of this idea are definitely within reach today.
I am saying that your proposition is a rational choice where the best decision is the one with the least risk versus highest productivity.
The optimal decision for a self-interested individual wouldn't necessarily be the one which results in highest productivity, but the one which results in the greatest relative power increase. He doesn't just want to be productive, he wants productivity gains which can then be used to forward his own ends. Government researchers would only get a tiny slice of the fruits of their labor. A single individual would get the whole pie. I think this would be worth a considerable amount of risk.
Up and until you can automate the problem-solving process used to program code in the first place, you're looking at the same problem repeated with ever more complicated software tools.
Yes, but the problem-solving process gets faster and easier as time goes by due to the increasing effectiveness of the tools. Easy tasks become trivial. Hard tasks become easy. Seemingly impossible tasks become feasible.
The bottleneck isn't the human mind, for it can always learn and develop solutions, but the computer's inability to solve its own problems. We must always do its thinking.
That's like saying the bottleneck to travel isn't the horse because a horse can always eventually get you from point A to point B. The central point is that airplanes get us there faster and easier.
Yes lack of strong AI is a greater bottleneck than the human mind, but augmenting the human decision making progress is a much easier problem to solve. Let's say we can develop tools that make programmers 100x more effective. Maybe then AI would be feasible.
If you're talking about ruling the world that requires state of the art software to solve scientific problems faster than a team of government researchers.
There are large regions along the cutting edge of software development which are not hardware bound. Solve these and you end up with an unknown set of capabilities which you could then potentially use to come up with any number of creative solutions to the various problems you've mentioned.
You cannot encrypt the code which the processor runs.
So now Amazon is monitoring every single instruction that passes through its CPUs?
You would still be impeded by the hardware because it cannot think for itself. Until it can, you're just building steps up to realising AI.
Ok, so those steps lead to AI. Eventually this guy reaches the point where he develops a strong AI for himself. AI would be just another step in this process.
AI, researcher software, protein folding, quantum decryption, these are all advanced problems necessary for beating a government, and they all require superior hardware.
You're making some big assumptions here.
1) Solving those problems is necessary to beat the government.
2) Those problems necessarily require superior hardware.
3) One person could not acquire the necessary hardware.
#1 Solving easier problems could potentially be sufficient.
#2a. AI doesn't necessarily require superior hardware. Nobody knows what the final answer here is going to look like. Same with researcher software.
#2b. Creative hacks are discovered every day which obviate the need for superior hardware. Why should a moderately enhanced human not be able to achieve such hacks?
#3. As I have said cloud-computing is one means of accessing this. Getting rich and purchasing it is another. Building it himself is another.
Also, there are probably many, many ways of circumventing all of these problems which I haven't thought of.
This guy would be able to solve these problems faster than the government researchers because while the government guys are chipping away at these problems with their hammer and chisel, Mr. God-Emperor is off inventing a jackhammer.
Think of IDEs and all of their constituent mini-tools. Think of text editors, mindmappers, lint like programs, project management software. These are all ways of making development faster which don't involve high level languages. I don't see how can say HLLs are "the primary tool".
High Level Language, as opposed to Low Level Language. Python, C#, BASIC, are all High Level Languages which facilitate rapid prototyping because they have built-in garbage collection, DMA, and APIs which facilitate most tasks effortlessly.
Assembly, C and arguably C++ are Low Level Languages, they lack garbage protection, DMA, and APIs which means you must write everything from scratch.
The IDE and RAD are platforms built upon these languages. They can never be as helpful as the language itself unless they write the language for you.
It doesn't matter if an operation takes .1 seconds or .0000001 seconds if my mind can only operate in 1 second intervals. Surely you see my point here?
We're talking about developing advanced algorithms, so the difference between .1 seconds and .0000001 seconds per iterative task in that algorithm is the difference between a solution taking one second and the same solution taking over a year to calculate.
I think you misunderstood what I meant by mind mapping software.
Apparently so.
Eye tracking doesn't require AI, it's just a matter of hooking up a webcam, detecting pixels, dealing with noise and then turning that into OS level actions.
Not a very simple task. Computers lack our heuristics.
I don't know where I suggested "recognizing speech", but turning speech into text can be accomplished without AI.
You didn't, but it is an example of a software solution that has eluded programmers for decades and which an AI could solve easily.
Well that could include a wide range of software of varying levels of sophistication. Some versions of this idea are definitely within reach today.
Yes, but they are still miles away from where it ought to be.
Government researchers would only get a tiny slice of the fruits of their labor. A single individual would get the whole pie. I think this would be worth a considerable amount of risk.
Which then puts him at the "lost behind group researchers and technological revolutions" end of the equation.
Yes, but the problem-solving process gets faster and easier as time goes by due to the increasing effectiveness of the tools. Easy tasks become trivial. Hard tasks become easy. Seemingly impossible tasks become feasible.
This is correct.
That's like saying the bottleneck to travel isn't the horse because a horse can always eventually get you from point A to point B. The central point is that airplanes get us their faster and easier.
Actually in your horse analogy it should be:
The rider is the common element in the horse and airplane, and all iterations between them. Therefore the rider is the common thread and the horse was the bottleneck, then the car, then the plane, and then the jet. The vehicle is always the bottleneck because it cannot operate without the rider's direction.
Yes lack of strong AI is a greater bottleneck than the human mind, but augmenting the human decision making progress is a much easier problem to solve. Let's say we can develop tools that make programmers 100x more effective. Maybe then AI would be feasible.
This is correct.
So now Amazon is monitoring every single instruction that passes through its CPUs?
It is capable of it with debugging software, and you can bet that logs would exist.
You're making some big assumptions here.
1) Solving those problems is necessary to beat the government.
2) Those problems necessarily require superior hardware.
3) One person could not acquire the necessary hardware.
The only big assumption is number one, since ultimately government has the weapons. Two and three are obvious.
#1 Solving easier problems could potentially be sufficient.
The easier the problem, the more likely that it has been solved or is on the verge of being solved by your competitors.
#2a. AI doesn't necessarily require superior hardware. Nobody knows what the final answer here is going to look like. Same with researcher software.
Modeling of a mouse's cortex required a supercomputer. A human brain has more neurons and synapses than our most advanced supercomputers, by a large margin.
#2b. Creative hacks are discovered every day which obviate the need for superior hardware. Why should a moderately enhanced human not be able to achieve such hacks?
Hacking a neural network to use fewer nodes only causes each node to bear more weight, and at worst impedes the emergent property of AI.
#3. As I have said cloud-computing is one means of accessing this. Getting rich and purchasing it is another. Building it himself is another.
If you had access to every computer in the world simultaneously, you might have the computing power necessary to emulate a human mind.
This guy would be able to solve these problems faster than the government researchers because while the government guys are chipping away at these problems with their hammer and chisel, Mr. God-Emperor is off inventing a jackhammer.
To extend a metaphor, while Emperor-sama is inventing a jackhammer, the government researchers have finished their task and are moving to version point two.
The IDE and RAD are platforms built upon these languages. They can never be as helpful as the language itself unless they write the language for you.
Who says C++ the language enhances productivity more than Visual Studio the tool? Just because a shirt can't exist without cotton does not mean raw cotton is more useful than a shirt.
We're talking about developing advanced algorithms, so the difference between .1 seconds and .0000001 seconds per iterative task in that algorithm is the difference between a solution taking one second and the same solution taking over a year to calculate.
Only if you need to run the task many times. Often .1 seconds is sufficient. My point was that many potentially useful programs don't require heavy hardware.
Not a very simple task. Computers lack our heuristics.
Whether it's simple or not, it's clearly doable, today. Software exists which can do it.
You didn't, but it is an example of a software solution that has eluded programmers for decades and which an AI could solve easily.
Eluded for decades? There's software that can do this too.
Which then puts him at the "lost behind group researchers and technological revolutions" end of the equation.
Again, group or no group the story is largely the same. Productivity improvements in a virtuous cycle until one small set of individuals hold a vastly disproportionate amount of power.
And if this guy has built tools which enable him to absorb and act on information more easily, would technological revolutions not merely serve to further differentiate his abilities?
Actually in your horse analogy it should be:
The rider is the common element in the horse and airplane, and all iterations between them. Therefore the rider is the common thread and the horse was the bottleneck, then the car, then the plane, and then the jet. The vehicle is always the bottleneck because it cannot operate without the rider's direction.
Yes, that's exactly the point I was making. Without removing the constraint of the rider we can still achieve improvements of several orders of magnitude.
Let's say software is currently at the point of early cars. We could expect to see improvements of several orders of magnitude in human productivity before we hit a limit requiring AI.
It is capable of it with debugging software, and you can bet that logs would exist.
I think you are pretty obviously wrong that there would be significant difficulty in conducting nefarious computations via the various cloud computing platforms. But I don't think you will budge on this point, so I'm going to call this an impassse.
The only big assumption is number one, since ultimately government has the weapons.
In a digital world the meaning of the word "weapon" is not so clear.
The easier the problem, the more likely that it has been solved or is on the verge of being solved by your competitors.
It is likely that the vast majority are going after the overwhelming array of low-hanging fruit rather than pursuing the grinding work necessary to vastly increase one's personal power.
And, again, programming tools are a multiplier so we should expect to see diverging power levels. The strongest of these competitors would become stronger over time until one or a few eventually dwarfed the capabilities of all others.
Modeling of a mouse's cortex required a supercomputer. A human brain has more neurons and synapses than our most advanced supercomputers, by a large margin.
AI wouldn't necessarily require modelling the human brain.
Hacking a neural network to use fewer nodes only causes each node to bear more weight, and at worst impedes the emergent property of AI.
I think you may have misunderstood what I meant by "hack". I meant "creative solution to a difficult problem." There was a time when searching all of one's hard drive could take hours, but with the hack of indexing everything beforehand it can be done in milliseconds. Who is to say some other such creative insight won't allow for easy development of AI?
Take a clone of yourself that is ten times as capable as you. That guy can do things that you fundamentally cannot. What you see as impassible roadblocks he might circumvent easily. The fact that you with your limited knowledge think that something can't be done does not necessarily make it so.
To extend a metaphor, while Emperor-sama is inventing a jackhammer, the government researchers have finished their task and are moving to version point two.
I think you put far too much stock in the capabilities of government researchers. By the time this guy finishes his jackhammer the government researchers will only be halfway through the funding requisition process for a project of dubious worth.
