The Singularity

[Chris Coles]

Re: The Singularity

While I do share your enthusiasm for the mechanical multi spindle, the German Index CNC multi spindle is a pretty amazing machine. (see: http://www.index-werke.de/de/englisc...x_ENG_HTML.htm). I worked for a company that had a mix of Davenport, Euroturn and Acme Gridley mechanical machines but the Index could do all that the mechanical machines could do and more.

Where I agree is that the machine (CNC or not) will just sit stupidly and do nothing without human intervention to set it up, program it and attend to its needs. Billions of years of development have gone into the animal brain and machines have a long way to go before they can approach a fraction of what humans can do.

Gosh, does anyone still use Acme Gridley's?? .... argh! Back in the late 1960's I was dumped with a full row of them, seven machines, turning everything from hex Gunmetal to EN32B where the tools wore out every few hours and the cross slide cam tips were breaking up due to wrong pressures set on the dead stops......... Blimey, that brings it all back.

Your final point is by far the most correct; until you have been there and seen it, you would not believe the enormous variety of failure modes for a mechanical system that has to be set by a human being that in turn has failure modes.

For those of you not familiar with the challenges of setting a multi-spindle lathe; I make one point. If you need to find employees that can do and will NEVER give up trying .... employ someone that has achieved in Multi-spindle setting. Because they have to operate in a mass production environment where a machine stopped must be back running again ASAP, (they run 24hr x 365days and never stop except for either setting, tooling or replenishing the bar feed); the setter can NEVER walk away from a problem. You have to stand there, come what may and SOLVE the problem; regardless of why it occurred; you have to achieve a working machine...... 12 hr shifts, half hour lunch, 33 mile drive to and from work. NOW, THAT is working for a living!

[unlucky]

Re: The Singularity

Your final point is by far the most correct; until you have been there and seen it, you would not believe the enormous variety of failure modes for a mechanical system that has to be set by a human being that in turn has failure modes.

Personally I work in the IT sector and let me tell you, it's the exact same story. There are thousands of IT staff working 24 hours a day in data centers all over the world, dealing with the multiple continuous and on-going failures of all the machines and software that constitutes the "Internet" (plus all the private networked services that firms use internally). Sometimes a failure is big enough to start affecting users (e.g. Blackberry) but the overwhelming majority are fixed before users ever notice.

And that doesn't include the software/hardware developers who create these systems (and their bugs) in the first place.

I recall EJ's comment on Y2K, that on 1 Jan 2000 a lot of buggy systems that break down all the time will keep on being buggy and breaking down all the time. Spot on. And "the singularity" will be the exactly the same!

[reggie]

Re: The Singularity

Agreed, and this is Lanier's point too. The idea of the singularity, and its great promise, encourages certain behaviors such as posting ideas, friends, likes, dislikes, etc in online public spaces (including these forums) for free. All of that information will be (is already) "controlled" in terms of access--perhaps not overtly but by the very technologies associated with the singularity--google search algorithm, adwords, and the like.

Imagine itulip if EJ never posted, but just arranged visitor post priority.

It's not just a tool for analysis, it's a tool designed for input back into the public as well. This is where John Boyd's OODA loop work is so important, because if one can possess the fastest processing power, than one can analyze inputs from the public and design outputs back into the public faster than the public can respond. This is what transformed the US Military, and this is what is being applied in the public sphere now.

Robots, are merely OODA Loop processors deployed in the field, creating greater and faster processing power for their controllers, while delivering some benefit to the users.

[Ghent12]

Re: The Singularity

I'm not a Luddite. I love technology. I just don't believe that once many blue collar manufacturing and data sifting jobs become partially or fully automated away that there'll be some sort of new field of work that'll automagically pop up to give all those out of work people something to do that will pay as much as their old job. After all why couldn't those jobs be automated away too by ever better machines? What happens to society when a large minority or even most of it becomes obsoleted and made useless by cheaper, faster, better machines that don't require healthcare and need only a little management?

I think either: laws will be passed to defend labor against automation OR those newly useless people will be effectively abandoned by society OR society rearranges itself to benefit most from the automation while attempting to distribute wealth in a equitable manner (ie. shorter work days and weeks, free or nearly free career retraining after college, higher wages or cheaper goods and services, etc.).

That last possibility would be the best IMO but I fear the 2nd one to be the most likely.

The shift need not invoke many laws at all, nor any specific response from any group. It is unfathomable that more automation will not be dealt with in the same way that automation was has always been dealt with: the price of raw, semi-finished, and finished goods are reduced at a given level of demand. Hence, some society with a theoretical super-automated manufacturing sector will probably just have cheap goods, but need not necessarily have some extreme labor surplus. Take a trip down to Lowe's and look at a few of the trucks that pull in to get an idea of how an economy can shift from making stuff to using stuff; people currently get by by providing services to each other.

Also, take a mental trip down the flip side of super-automation: who will provide the capital for that? Probably the banks for almost all of it. In order for the banks to provide capital for super-automated processes, there should be an expectation of customers for the products of such a process. That alone insures some employment.

But ask yourself this: what can you do with cheap raw and semi-finished goods? Can you build your own custom furniture since style innovation cannot be automated? How likely is it that plumbing services will be completely automated? Or any type of construction? As the inputs for all of these types of activities decline in price, it is probably a safe bet to assume that the intensity or magnitude of those industries will improve.

Not even the stupid have to fear some "super-automation" taking root in society. However, the smart stand to benefit quite significantly, just as they have from current and past trends in automation. Machines need builders, designers, and technicians after all. The bottom line is, however, that an economy ultimately is a system to supply the demands of the people within it generally according to the economic worth that the system places on the various individuals. Hence so long as people are able to provide some input into the system, they should be able to extract some output from it as well and no "super-automation" revolution will change that.

[mesyn191]

Re: The Singularity

The shift need not invoke many laws at all, nor any specific response from any group.

Capital and industry cannot control themselves, they have what can be most generously described as tunnel vision that is profit focused. Everything is secondary to the profit motive, certainly ethics and even people's lives too. When people stop becoming profitable to keep around they will be canned, its just that simple. This is the root of all outsourcing today. Americans are too expensive so they get some SE Asian or Chinese guy to do it for $1 or 2 a day. Eventually it'll be even those SE Asian's and Chinese will be too expensive vs. machines and they'll be canned too.

It is unfathomable that more automation will not be dealt with in the same way that automation was has always been dealt with:

You're not understanding the level of automation that is coming nor are you considering what happens to wages when everyone is trying to provide whatever service you can think of. The future is not going to be just like the past, nor will it rhyme in tune. Besides the obvious issue of the wages for most services dropping like a rock since there'll be a ton of people trying to do those jobs you'll also have the issue of competency to deal with if you want to consider retraining. Most people won't be able to retrain as an engineer or scientist or artist or software coder. Only a relative few are any good or even just decent at those jobs. Eventually a significant minority or even most people are going to end up being effectively useless to society at large.

the price of raw, semi-finished, and finished goods are reduced at a given level of demand.

You know when the whole trade imbalance started getting going with China you could get some really decent stuff for cheap, and people were happy. Now a days though the quality of the products is crap even if it is cheap, and people in general can hardly afford anything else since they don't have the wages for it. More automation will do more of the same. Goods will get even cheaper...but that will be due to the elimination of human workers mostly or altogether and then later on the sacrifices in product quality will come too in order to cut costs even further and maintain sales.

If people aren't making the money they won't be able to buy much or even anything no matter the price. In that scenario the government or corps start to have to guarantee a certain level of employment and/or wages just to keep things going.

Also, take a mental trip down the flip side of super-automation: who will provide the capital for that? Probably the banks for almost all of it. In order for the banks to provide capital for super-automated processes, there should be an expectation of customers for the products of such a process. That alone insures some employment.

Expectation of customers does not actually produce real world customers, otherwise the housing and .com bubbles never would've popped. And the banks were pushing heavily behind both of those bubbles.

Can you build your own custom furniture since style innovation cannot be automated?

You know most people don't build their own furniture now right? They just go to Sears or Ikea.

How likely is it that plumbing services will be completely automated? Or any type of construction?

This is already happening. Pre fab homes are becoming more and more common. Pre fab warehouses and multi story buildings made of ICF's and trucked out to location have been common for a decade now.

[c1ue]

Re: The Singularity

It is unclear to me that even were said fabrication machines to be cheap to build/operate and easy to manipulate into creating goods, that there will be any long or even medium term effect on overall demand.

As we've seen again and again in the various waves of the Industrial Revolution, the Electricity/Lighting/Household appliances Revolution, the Internet Revolution, and so forth - the net effect of greater productivity is demand for ever more complex products.

The housewife in 1900 used a natural material broom and maybe a Franklin stove.

The working mother today uses a vacuum cleaner and a microwave.

Clearly neither automation or manufacturing are the end-all, be-all of labor.

[mesyn191]

Re: The Singularity

It is unclear to me that even were said fabrication machines to be cheap to build/operate and easy to manipulate into creating goods, that there will be any long or even medium term effect on overall demand.

As we've seen again and again in the various waves of the Industrial Revolution, the Electricity/Lighting/Household appliances Revolution, the Internet Revolution, and so forth - the net effect of greater productivity is demand for ever more complex products.
....
Clearly neither automation or manufacturing are the end-all, be-all of labor.

Well you have to also consider the 8hr work day/40hr work week wasn't the norm back in and prior to 1900 and that almost everyone, even the kids except perhaps the wife in a family*, worked a fairly labor intensive job for 10-12 hours 6 days a week.

It wasn't until the 1930's or so that laws were passed to make the 8hr day/40hr week the norm. Which was something that labor had to fight very long and very hard for, since at least the 1880s in the US IIRC, and people take it for granted now. They also started to keep more kids in school instead of working, which drastically shrunk the work force and started training up more skilled labor. So the waves of Industrial Revolution here and world wide did indeed have a huge huge effect on people's lives and society outside of "you can get x widget for y cheaper and also maybe make some other more complex goods". There is more than a little historical precedent for what I'm saying. I'm simply also pointing out that there is a limit to what you can train most people up for and that capital will become more and more aggressively anti labor at some point, for some work fields possibly soon, due to their profit motive.

What we need is a 20hr work week over the next few decades, personally I think a 32hr work week would make lots of sense for many people right now. Capital will fight that tooth and nail though, just like they did the 40hr work week and 8hr days, much less niceties like overtime.

*she wasn't paid to do her job of raising the kids, cooking, cleaning, etc. which back then was just quite labor intensive and time consuming too

[c1ue]

Re: The Singularity

Well you have to also consider the 8hr work day/40hr work week wasn't the norm back in and prior to 1900 and that almost everyone, even the kids except perhaps the wife in a family*, worked a fairly labor intensive job for 10-12 hours 6 days a week.

*she wasn't paid to do her job of raising the kids, cooking, cleaning, etc. which back then was just quite labor intensive and time consuming too

My point was that even 40 years ago, wives didn't expend tremendous effort cooking or cleaning. There might not have been Dyson vacuum cleaners, but there definitely were vacuum cleaners. There might not have been microwaves, but housewives were already buying pre-butchered meat and vegetables in grocery stores, not killing/processing their own meat and harvesting food, and cooking in ovens and over electric or natural gas stoves as opposed to wood fired.

Yet today - with arguably more labor saving devices - there is more labor and less leisure than 40 years ago.

Part of this is simple over-consumption: we have bigger houses, keep them cleaner, and stock them with all manner of crap unnecessary to actual survival.

The other part is the ever increasing burden of FIRE.

I don't see either of these effects going away any time soon - house size may cease to increase but the cost and complexity of its contents is only going up.

An example is television: a traditional CRT television is effectively a 70 year old manufacturing process. While not so easy to create at home, it is possible given a modest workshop and specifications.

An LCD television, on the other hand, is physically impossible for any person or even a group of 100 people to create.

[mesyn191]

Re: The Singularity

My point was that even 40 years ago, wives didn't expend tremendous effort cooking or cleaning.

Sure and people do buy random essentially useless stuff because its hip at the time or they get a kick out of it, if they have the income for it. And goods have gotten more complex, but who says you need a person to put them together or will in the future? Given what they can already do with pick and place machines today I don't see where your skepticism is coming from.

Yet today - with arguably more labor saving devices - there is more labor and less leisure than 40 years ago.

That isn't a sign of more jobs being automagically created to fill in the gaps, you sure you aren't essentially arguing Say's Law? People are trying to work more to make up for the lack of income, the dual income earner family and consumption society are not goods things for all sorts of social and economic reasons and are probably unsustainable given the likely hood of increasing materials and energy cost in the future.

[thriftyandboringinohio]

Re: The Singularity

...cost and complexity ... is only going up.

An example is television: a traditional CRT television is effectively a 70 year old manufacturing process. While not so easy to create at home, it is possible given a modest workshop and specifications.

An LCD television, on the other hand, is physically impossible for any person or even a group of 100 people to create.

That's my view of Kurzweil's singularity. It's usually discussed as a technical utopia as knowledge goes exponentially vertical, with medical science making us immortal and industry giving us endless material comforts.

I see complexity and interconnections going exponentially vertical, creating a world full of surprising unintended consequences. A flood in Thailand shuts down auto assembly lines in Alabama. Crooked bankers in New York ruin pensions in Europe. While advances in medicine might make it possible for people to live 200 years, it only takes a single criminal plant manager in China making toxic gluten to poison the daily bread of thousands of people around the world.

[Techdread]

Re: The Singularity

Clue what do you think of the Memristor?

http://www.theregister.co.uk/2011/05..._breakthrough/

[mesyn191]

Re: The Singularity

Memristors are really nice but probably aren't a game changer. Might allow for cheap-ish SSD's as main memory or for mass storage if everything works right. What they need for that singularity stuff to really have any hope of taking off is something that amounts to a super cheap 8nm etch-a-sketch for a ultra low power process that allows stacking of some ridiculous number of layers (certainly hundreds, probably more like thousands) to form 3D chips at very high clock frequencies.

Right now even doing something like memory stacking along with a GPU/CPU on a single package connected with a SI is absolute cutting edge but a joke compared to something like that. I don't know what the heck they're gonna do to get around the heat problems too.

Personally I suspect they're going to have to switch the fundamental way they produce CPU's/GPU's to keep scaling up performance as much as you need.

[dcarrigg]

Re: The Singularity

That's my view of Kurzweil's singularity. It's usually discussed as a technical utopia as knowledge goes exponentially vertical, with medical science making us immortal and industry giving us endless material comforts.

I see complexity and interconnections going exponentially vertical, creating a world full of surprising unintended consequences. A flood in Thailand shuts down auto assembly lines in Alabama. Crooked bankers in New York ruin pensions in Europe. While advances in medicine might make it possible for people to live 200 years, it only takes a single criminal plant manager in China making toxic gluten to poison the daily bread of thousands of people around the world.

I think you're generally right - except for the 200 year part (and you did say "might"). This "people are living longer" stuff always gets to me, so it's nothing against you. If Ramesses II could live to 87, then there is no magic now. They may prevent a heart attack or infection from being terminal where it otherwise would have been, but the upper bound of aging seems to be fairly consistent through time.

[c1ue]

Re: The Singularity

Sure and people do buy random essentially useless stuff because its hip at the time or they get a kick out of it, if they have the income for it. And goods have gotten more complex, but who says you need a person to put them together or will in the future? Given what they can already do with pick and place machines today I don't see where your skepticism is coming from.

You're still missing the point.

Theoretically the machine noted above could create a CRT television, if it could somehow assemble the vacuum tube, but that machine could never, ever be able to create an LCD television.

The sheer scale of complexity embedded in an LCD television is such that even just the design is far beyond the capabilities of all but a small group of people equipped with million dollar software packages, years of experience, and educated in that specific field.

The sheer cost of the equipment necessary to manufacture LCD panels, plus the electronics around them, is more than the GDP of most nations: the cost of a modern semiconductor plant itself would put it as #145 or so in the list of world nation's GDPs.

Note the difference with the Industrial Revolution and Electricity/Home appliance revolutions: these 2 revolutions extended the reach of individuals in terms of ability to work at what they needed to do. In a real sense, these changes extended the capability of individuals to be self reliant - because the capital cost of this equipment was more than repaid by the productivity enhancements.

In contrast, in the modern era, I greatly suspect the productivity enhancements in terms of self sufficiency of the first 2 revolutions simply aren't there.

That isn't a sign of more jobs being automagically created to fill in the gaps, you sure you aren't essentially arguing Say's Law? People are trying to work more to make up for the lack of income, the dual income earner family and consumption society are not goods things for all sorts of social and economic reasons and are probably unsustainable given the likely hood of increasing materials and energy cost in the future.

I'm not sure what you're getting at.

While certainly there is a class of people in the United States for which Say's law applies, I'd argue that this is in fact the opposite for most people.

They don't have infinite resources to buy infinite goods; they have a bare minimum plus a few luxuries. Even then, these luxuries are by and large electronics - most expenditures go towards food, housing, and transportation as has been the case for centuries.

I strongly recommend you look at Dr. Elizabeth Warren's early work on household spending between 1974 and 1994.

Clue what do you think of the Memristor?

It is an interesting technology, but I don't know the details enough to be able to compare its potential vs. traditional transistors.

What I can tell you is this: transistor count is no longer an issue. 15 years ago there was a significant cost to silicon area due to transistor sizes; even 5 years ago this became moot.

At that time the shrinkage in physical dimensions was such that transistor counts no longer mattered, but secondary effects like heat and substrate leakage came to the fore.

One of the reasons AMD was able to 'surpass' Intel in the processor wars in the mid '90s was that Intel had concluded that these heat/substrate leakage/other issues would became the gating factor in processor design - and redirected their design and process development technology to focus on these other issues rather than sheer clock speed.

AMD showed that Intel was being overly conservative - that there were 1 (actually 1.5) generations still left even using the 'old' paradigm, but of course the Empire has since Struck Back.

The immediate issue I see with memristors even in its immediate application - competing vs. Flash - is durability.

While theoretically memristors can behave like flash - i.e. retain information even without power - the question is going to be how many times it can do this at 99.9999% success rate. Flash does this via quantum tunneling to put an electrical charge basically in a sealed container; memristors seek to do this by having a material which 'remembers' what last happened to it. But because the material is fundamentally changing and changeable, there are almost certainly effects over time which make the 'memory' less reliable. Think of a slinky that's been stretched too many times. While a given bit of memory might cycles at mere millions of cycles per second compared with processor billlions of cycles per second, that's still a whole lot of slinky-stretching.

The second issue with memristors and the Singularity is that they still require power. A memristor is just a resistor with some 'memory' capabilities; well, one major effect of any resistor is to convert some current to heat as current passes through. This is where my lack of familiarity prevents me from making any judgments; the question is whether the set of memristors which mimic a transistor will dissipate more heat than just a small transistor. If it does, then the revolution is not going to come because heat is already the single largest issue in processor design. And the memristor itself might not be the issue; arguably the wiring these days consumes more power in the form of heat dissipation than the transistors themselves. A memristor based design would not have any less wiring, and probably would have more, than an equivalent transistor design, though the difference is probably relatively small.

Memories don't have a heat issue because they cycle 2 orders of magnitude or more less often than transistors in a processor.

The last issue is software. How do you design software to make use of memristors unique capabilities?

If memristors were working perfectly today, I can tell you exactly how: they'd make 'cells' of each logic gate type, but composed of memristors rather than transistors, and just use existing software. This means using existing design paradigms.

Existing design paradigms, I have tons of experience with. There is zero possibility of a singularity occurring with existing hardware/software combinations; for one thing the capability to reprogram is nonexistent given that programming itself is already extremely difficult.

An example of this can be seen in comparing processors in the 1990 era vs. 2000 era.

There are some fundamental differences - principally in the 1990s there was still debate on RISC vs. CISC, but beyond this the basic function of processors in 1990 were identical with 2000. The more modern processor were different in only 2 ways: sheer scale as in more transistors which were smaller and faster - along with more memory, and pipelines.

Scale is straightforward, but pipelines are basically shortcuts through the processor for specific cases of instructions. A 1995 in consumer hands processor might have 2 pipelines, today there are 14 in the i7 core and 31 in the last Pentium 4 cores. Each pipeline makes 'guesses' as to what is coming next, and if right jumps over part of the normal processor flow.

You can see that this doesn't lend itself well to scalability: putting in 100 pipelines is well beyond the realistic design capability of anyone, anywhere right now. Pipelines make extravagant use of cheap transistor real estate, but at a cost of huge amounts of heat overhead thus there is a very strong law of diminishing returns.

If people can't do it, even with ever burgeoning software, there's a chicken or egg deal going on which only the super-omniscient Borg of the future can create the hardware/software which gave birth to it.

Note the dates above are 'designer' dates; add 2-4 years for the actual in production dates.

One positive point for memristors: one of the bigger secondary issues in 2006 (when I left the industry) was leakage. Basically transistors fundamentally squirt a few electrons into the substrate; enough of these transistors and the substrate stops acting as a voltage ground and problems ensue. It is one reason why AMD used an SOI process even though it was more expensive and furthermore handicapped their ability to increase volumes (and thus doomed them vs. Intel).

I'd assume memristors wouldn't have this issue.

[thriftyandboringinohio]

Re: The Singularity

I think you're generally right - except for the 200 year part (and you did say "might"). This "people are living longer" stuff always gets to me, so it's nothing against you. If Ramesses II could live to 87, then there is no magic now. They may prevent a heart attack or infection from being terminal where it otherwise would have been, but the upper bound of aging seems to be fairly consistent through time.

I took that from Kurzweil, just going by memory without looking it up again. He's predicted that children today might live to be 200 years old as miraculous advances in medicine become available during their lifetime. A couple years ago I became aware of his singularity concept and looked through it casually.

This point of his stuck with me because it would apply to my children if he's got it right. As I recall, he postulates nanobots in the bloodstream doing microsurgery; gene repair; cheap and easy replacement organs; the whole nine yards, right up to transferring ones consciousness into a machine for immortality.

Your point is a great one. Although average ages go up with advances in medicine, the upper age limit for people who manage to live very long stays around 100 years maximum, unchanged since reliable records have been kept.