Why the Singularity Isn’t Coming
What is the Technological Singularity?
The technological singularity is the idea that technologically is advancing exponentially, and at some point these gains will compound so much that there will be a massive explosion in human and machine intelligence, immortality will be achieved and humans will obtain a godlike type of status. In a sense, it’s almost a form of technological religion. One of the main proponents of the technological singularity is Ray Kurzweil.
One thing to note is like many other singularists, Ray Kurzweil puts the date of the singularity to be near his potential lifespan. He believes that the singularity will happen by 2045. This would make him about 97 when it happens. It seems that many singularists fear death and want to believe that the singularity is within their reach. That is why why most prominent singularists put the date of the singularity near their potential lifespan.
Exponential Functions are Really Sigmoid Functions
The problem of many singularists is that they mistake sigmoid functions for exponential functions. What is a sigmoid function? A sigmoid function is one that starts off with an exponential trend but then with growth slowing down and flattening. It’s like an exponential trend followed by a logarithmic trend.
There’s a good article I found by procrastilearner on Steemit where he explains how most functions that appear to be exponential are actually sigmoid.
Sorry Kurzweil But The Singularity Is Not Near - Steemit
"The Singularity."Image credit: Pixabay.com link CC0 license Ray Kurzweil is a famous futurologist that has... by…
For example, the machine gun was invented in the mid-1800’s. The first machine guns were heavy and clunky affairs that jammed often. Later models that were lighter and more efficient but overheated quickly because they fired ammo more rapidly. Later machine gun models were improved on the production side until eventually the AK-47 was invented which can be made quickly and cheaply. This gun was invented in 1945 and improved over the following decade.
In the 70 or so years that it has been around its design is still very much the same as in those early days. The machine gun is therefore a good example of a technology’s advancements eventually slowing down. New models are only incremental improvements over previous models.
Like the case of machine guns, when looking at most advances over a long enough time, they appear to be sigmoid instead of exponential. In fact, sigmoid functions are more aligned with how the world works. There can’t be exponential functions because they would break the laws of physics. There isn’t infinite space or energy to utilize.
Population growth is an example from biology where we can see where an exponential function would hit its limits. Bacteria in a petri dish can grow very fast, doubling every 20 minutes. But eventually, they run out of food, and the rate of growth declines and they die off.
Human population growth works in a similar way. The population rose very rapidly, growing 8x larger since the year 1800. However, this is not an infinite trend and birth rates are now dropping rapidly. Families are getting smaller and humans have realized that they need to live more sustainably if they want to remain on the planet without suffering the fate of bacteria in a petri death. Thus, the population is expected to level out around 10 billion. We won’t be having another 8x growth from 8 billion to 64 billion.
Aircraft speed records are also something that show a sigmoid function. Early on in aviation as engineers learned more, they could quickly set new records. However, since 1976, no new records have been set. Due to heat, friction, and limits in reducing drag, new improvements in aircraft could only possibly yield minor improvements. In this case though, no institution has even tried to break the record set in 1976. This is because supersonic spy-planes fell out of favor due to the ability to spy with satellites.
While innovations are limited by the sigmoid curve, sometimes there’s also no reason to improve on them anymore, because new innovations can work in that previous field better.
Computing is something that seems to be truly exponential for many, but in reality, we are reaching a point where it is getting hard to shrink transistors any more. This is because there are problems with heat dissipation and quantum tunneling of electrons when transistors get to small. Transistor size is also limited by the size of the Silicon atom.
Why is Moore’s Law ending? Computers are no longer improving at the rate Moore’s Law suggests while the costs of producing more advanced semiconductor chips are rising instead of falling.
While processing speeds are still trending close to Moore’s Law in some cases, it’s simply because companies are spending significantly more on research and development to try and meet the expectations of Moore’s Law. But we can see in the actual pricing of newer graphics cards and processors that things clearly haven’t held up on the pricing end. The cost of new computer hardware these days is extremely expensive, and most new graphics cards coming out are not affordable to consumers. Moore’s Law isn’t just about processing power, but the expense to obtain it. In the past, the cost for double the processing power after a certain period was almost constant, but those times are long past now.
Eventually, companies like Intel and AMD will have to accept that exponential improvement is done, as they’re hitting the limits of what they can do with silicon wafers. There have been new technologies suggested to continue to innovate, like using germanium or light-transistors, but these innovations will start at their own points and also follow their own sigmoid based curves.
Sigmoid functions are inherent to our world, while there’s not evidence of true exponential functions appearing anywhere. It seems that Ray Kurzweil and other singularists exaggerated exponential trends in an attempt to believe in the singularity.
Exponential Computing Doesn’t Lead to Exponential Gains Everywhere
Although gains in computing are going to be limited by the sigmoid function as anything else is, computing is one area where we have seen vast exponential improvement. From 1956 to 2015, there was a trillion-fold increase in computer power. It’s true that we haven’t seen anything like this in other fields. While this is impressive improvement and has led to many cool things like computer gaming, video editing, and computer-run simulations, improvements in computer processing power don’t translate directly to improvements everywhere.
For example, in a few years computer processing power may double, but it doesn't mean lifespan is going to double or housing cost is going to become twice as cheap. There’s limits on the problems that sheer computing power can solve.
A lot of things have remained very stagnant. One example is the case of American suburbs. Though computing has advanced a lot from 1990 to now, many of the homes in the suburbs pretty much look the exact same as they did 30 years earlier. They might have Amazon Ring cameras or some devices on them, but the design of the houses is mostly the same, a lot of the same problems are there such as high car dependency and isolation.
We do have the potential to live longer than we did in the past, but this is not an exponential trend and is something that improves gradually over time. There’s also setbacks, like due to high accessibility to fast food, lifespan is actually staying stagnant or down-trending in some areas, due to increased obesity.
Singularists believe too easily that everything can be solved with more processing power, while when looking at the actual world we can see it’s clearly not the case. We live better than 1965, but not trillions of times better. There are things like human biology that are very complex that we can’t simply treat like a computing system. There’s complex societal divides that haven’t been helped by trillions of times more processing power. It could be said that technology sometimes drives further divides in our society.
I’m a believer that exponential computing power has led to a lot of important advances and that it is an impressive feat for humanity. However, truly improving society takes a lot of work in many fields, and these aren’t all going to operate exponentially. Even in the computing field, not everything operates exponentially. For example, the way computer programs are written themselves is sometimes more of the speed limiter than the actual processing available. We can add more cores, more memory, but we have to figure out how to use these things. Multi-core programming is still something that many developers struggle to do now.
What About Artificial General Intelligence?
Artificial general intelligence is likely something that will eventually come. However, it doesn’t mean it’s going to usher in something like the singularity. Artificial general intelligence would just mean that computers could have minds like animals or humans. They could learn, think, and reason. This would be a big advancement, but nothing like infinite exponential power. It would just be like having more humans around to help. Humans have their own artificial general intelligence already, and our intelligence has improved gradually over generations, not in tremendous leaps.
The human brain also runs much more efficiently than silicon hardware. It’s hard to compare the brain directly to a computer, but in computing terms, it’s believed the brain can solve certain problems with a performance of one exaflop per second (a quadrillion operations) while using only 20 watts of power. A supercomputer operating at 0.2 exaflops uses about 13,000,000 watts. Due to this, it’s possible that it may not even be efficient to use computers for artificial general intelligence. Computers already handle well certain aspects that are difficult for humans like doing math with very large numbers. It may be best to leave artificial general intelligence type of thinking to biological constructs.
To make artificial general intelligence work on the efficiency of biological processes, we would need to create similar hardware structures to organic brains. Neural networks are a direction towards this, but they are still very far from the brains of humans or animals.
How Will Innovation Actually Proceed?
Innovation will probably proceed like it always has. We make discoveries, make big advances in those discoveries over time, and then a slowing down happens as we achieve mastery of that field. We have to find new innovations in order to continue improving things. But we simply can’t get infinite exponential improvement out of any one technology.
Our life has certainly changed significantly from century to century, but this is not to due with infinite exponential growth. It’s simply improvement upon improvements that we make.
The path forward isn’t always smooth. We advance, invent, and sometimes there comes new unforeseen issues we have to contend with. Humanity definitely lives much better than it ever did, but we should be aware of the problems that we face to.
While many dream of the singularity, it’s important to remember that in much of the world, the farmers don’t even have tractors to use. There’s a lot of slave labor going on in countries like Africa in order to obtain the raw minerals needed for technological giants. Billions of people still live in extreme poverty.
It’s scientifically dishonest with ourselves to believe in the singularity or that there will be some kind of technological rapture. It’s also ignorant of many of the real issues that humanity faces. Yes, singularists do talk to some of these problems, but they make them seem too easy to solve. Many seem to hope for artificial general intelligence to save us or to make us immortal. These are nothing more then delusions.
We do not need to create entire philosophies out of the fear of death, or we’ll be making the same mistakes as religion. Let science remain rational, and remember that progress is the culmination of individuals from many different fields. Progress can be fast, sometimes it can be slow. There isn’t always a clear trajectory for the future.
The singularity may not be coming, but by working together, we can continue to innovate and make our world better.