TEDx Talk Rationale

 Rationale

Singularity, What Does It Mean?

Several of my reviewers have no technical or math background.   They had a lot of trouble understanding the word singularity.   I couldn’t drop the word or the entire technical world would not recognize my topic because all of them have heard about the singularity and already have opinions about it.

TedXEustis, unlike other TedX events, had a strict rule.  No visuals of any kind.  I dearly wanted to use this visual that I prepared a year ago as my visual.  They say a picture is worth a thousand words.   I didn’t have a thousand words to spend, I had only a sentence or two.  I wanted to use that curve to illustrate how technology would change and the same curve to depict how unemployment would change   Que sara sara, I had to make do without it.

Of course singularity is an exaggeration.  I’m sure that Von Neumann knew that.  He meant is at a metaphor.  It will feel like a singularity.  It will approximate a singularity.   We need a visual metaphor to emphasize that this future will be very different than the rapid change we experience today.

Why a singularity?

OK,  this is the core of the subject.   I have to confess being less than forthcoming in the talk.  Not all jobs are created equal with respect to the singularity.   Science and engineering (STEM) jobs relate to future technical progress, so replacing these jobs with machines creates a positive feedback leading to runaway growth, whereas other kinds of jobs have no direct effect.   Critics of the singularity idea will call this section hand waving because it can not be proved.

Suppose we have a national laboratory.   100 science researchers work there.  What do they do all day?  Science yes, but also they fetch coffee, they write papers, they review papers, they travel, they go to conferences, they help organize conferences, So, in a 40 hour week is there 1 hour per of science to advance the state of the art?   4 hours?  12 hours?  I don’t have an actual number in mind, but it is subsantially less than 40.

So why can’t we have machines that assist on doing those non-science tasks?   Actually we already do.  Office automation began in the 1970s.  In the 60s, we had 1 secretary per 4 engineers plus a typing pool with 35 typists to support 300 engineers.  The machines can’t go to the restroom for us or do some other thing.  But it is not crazy to think that machines doing non-science work make us productive enough to have 50 researchers at the lab instead of 50.   Then as machines become more capable, 40, 30, 20, 10 and so on until in theory we have one brilliant scientist doing the work of 100, but surrounded by machines.    

If the researcher has an idea, the machines can easily scan the literature (including historical literature in all languages) to see if someone has done it before, or something closely related.   If the researcher wants to validate his idea with experiments or observations, machines with access to the raw data from all past experiments and observations can data mine that raw data to see if it provides evidence without running a new experiment.   If new data is needed, the machines might help design the experiment, they can round up resources, labor, write the software, oversee execution, gather and organize the results, and compare them with the researcher’s hypothesis.   In other words, they can do an awful lot as assistants, with a human being remaining at the top of the pyramid.  None of those assistant roles require human-like intelligence or consciousness.   AI perhaps, but nothing resembling human thinking.

Now, suppose the researcher has a breakthrough.  Hooray, a really new concept!  How long do you think it will take for other scientists to accept the idea?   Plate tectonics took about 40 years.   Special relativity about 10 years.   Max Plank proposed Plank’s Constant in 1900, but even at the time of his death in 1947 he still didn’t accept quantum mechanics.   The fight over Darwinian Evolution still rages.   We still have Flat Earthers.  I don’t know what the average time is for ideas to be accepted, but I’m guessing it is measured in decades.    As I said in the talk, “People drag their feet.  Machines don’t.”  So in less time than it takes to say “Hmmm, let me think about that.”, an AI could have accepted tectonics, relativity, QM, evolution, and much more as the theories best matching the evidence.   Then as the scientist inhales, the AI moves on to the next steps following those conclusions. The machines can also apply new scientific theory to find engineering applications to make new machines and new products.   Ideas for how to improve AIs can be accepted as valid and propagated to every AI on the planet in milliseconds. If the new idea turns out to be bad, it can be reverted in milliseconds.   I don’t think that the singularity comes by the machines firing their bosses, it will come by them moving so fast that we humans are left eating their dust.

What happens when the scientist realizes that the machines are making him irrelevant.   He could shut the machines off, but if he did so, now he needs to re-hire his 99 colleagues and to brush up on non-science skills that have atrophied.   His laboratory would also become non-competitive with other labs that still use AI.   He could let the lab go out of business, but then he would have no job.   The jobs issue becomes entangled with decisions.

So, we can come to the point where machines bring us to the singularity, but still leave human figureheads in control at the top of each pyramid.   In all of this I repeat --- no human-like level of intelligence is required from the machines.   AI’s yes, but not human comparable intelligence.

I think the best word that captures what I tried to describe is subversion.  When I think of the singularity, I think of machines taking over by subversion.   But even subversive effects do not require rational thinking on the part of the machines.   The machines just do tasks, and gettting tasks done so efficiently subverts human overlords. Machines don’t need a strategy.

Singularities Are Far From Linear

Most of us think linearly.  If the singularity is 27 years away in 2047, when are we halfway there?   About 13 years, or 2033, correct?   Not correct.  Not if our future is nearly singular.    The halfway point in terms of technology and in terms of unemployment might not happen until 2045 or 2046.  Obsolescence of money, and abandonment of the work ethic would happen then, if at all.  All the remaining changes will then happen at what seems to us in 2020 as the very last minute. The last minute is a good metaphor for a nearly singular event.  I’m confident that is exactly what Von Neumann meant when he chose the word singularity.

There was nowhere near enough time in the TED talk to address nonlinearity.

Social Stigma

Actually, my belief is much simpler than I made it sound in the talk.

Past the magic number of 50% unemployment, the unemployed become the majority.  The majority elects politicians to make laws, Advertisers pander them. Entertainers pander them.   They decide who to honor or dishonor.  Whatever social stigma remains will not be directed at the majority.  That sounds like a no brainer to me.

Money & Jobs

One reviewer told me to “get a grip on reality.”   He had no problem with immortality, but a world without money just sounded silly to him.

Here is the image I had in my mind when I wrote that talk.   Do you remember the Sorcerer’s Apprentice scene from Disney’s Fantasia?  Mickey used magic to get a broom to do his work carrying water.  But then the broom became many brooms until Mickey almost drowned in excess water. (Watch the video here.)  I thought, “In a world where we are drowning in stuff (goods and services) who needs money?

So I visualize a machine.  Take a robotic surgeon machine.  It may or may not have AI.   Now visualize a robotic factory that produces these machines without human help.  The a robotic factory factory that produces robotic factories without human help.   Very quickly we become like Mickey Mouse and the water.  In that kind of world, who needs money?

Why would the factory owner/manager allow the factory to continue if they don’t get paid by customers?   Well, what is the alternative.   Would you sell the factory?   What would the buyer give you, money?  That’s a weak incentive in this Mickey world.   Real estate is the only thing I can think of that will remain scarce enough to have trading value.   But even if you sell the factory for real estate, that would put you out of a job.   We associate jobs with fame, influence, and power.   Those things alone may be a powerful incentive to keep your factory running even if no money comes in.   In other words, the money issue and the jobs issue are entangled.

So, like I said in the talk, when supply significantly exceeds demand, money becomes obsolete.  I believe that to be rational and realistic.

 

Immortality

There is no direct connection between prolonged life expectancy and the singularity.   In my opinion, it is just one example of the benefits that can come from advanced technology.    Genetic research already uses AI, but the AI is just a productivity enhancing tool; it is not central to the subject of the research.