Rise of the machines & the fallacy of exponential thinking

     Unless you have been hiding under a rock for the past little while, you must have surely heard about artificial intelligence and the risks it poses to mankind. In fact I wont fault you for hiding under a rock after listening to some of the absolutely bombastic eventualities as described by some rather esteemed people of our generation. Most of the arguments made against AI are based on two fundamental premises. The first being that AI is growing at an exponential pace or close to it and that the risk of something happening in the near term (5-10 years) is real and definite, unless we control it today. The second premise is that since AI has the potential to become more intelligent than us, we have no surefire way of predicting how it will behave and therefore cannot use past technological developments as much of a basis because we’ve never created anything that has the ability to, wittingly or unwittingly, outsmart us. Nevertheless there is a fundamental flaw in this thinking. That flaw is "the fallacy of exponential thinking".

     To truly understand exponential thinking, one has to understand linear progression, and by that extension linear thinking. In mathematics linear progression is when something changes or progresses in fixed increments from one stage to another, and has a starting point and an ending point. An example is counting from 1 to 10. In this example you start from 1 and increment by 1 every time until you reach 10. Another example is increments of 5 (5,10,10,20 . . .). We have been doing this since we were in primary school and we easily understand linear progression. Linear thinking therefore is a process of thought following known cycles or step-by-step progression where a response to a step must be elicited before another step is taken. This is so ingrained in us that we automatically believe that  the world around us operates linearly and that all growth follows a linear pattern. That is until we first hear about exponential progression.

    Exponential progression on the other hand is an increase in number or size, at a constantly growing rate (eg:- 1,2,4,8,16,32 . . .). Compound interest is a good example of exponential growth. It seems easy enough at first, but to quote Albert Alan Bartlet "The greatest shortcoming of the human race is our inability to understand the exponential function".

      Lets use an example to better understand exponents. Lets assume you were to eat the ever popular M&M candies but with a twist. If you were to eat M&M candies that increment linearly for a month, on day 30 you would have eaten 30 candies weighing 27.44 gms with a total caloric intake of 90 calories. At the same token if you were to eat them growing exponentially on day 30 you would have eaten 536,870,912 candies weighing 491 metric tons and a total caloric intake in excess of 1 billion. Most people would never have guessed that on Day 6 the number of candies on the exponential side crossed the count of candies on the linear side. This is the power of exponents.

	Linear				Exponential
	M&M's	Calories	Weight (gm)		M&M's	Calories	Weight (gm)
Day 1	1	3	0.91		1	3	0.91
Day 2	2	6	1.83		2	6	1.83
Day 3	3	9	2.74		4	12	3.66
Day 4	4	12	3.66		8	24	7.32
Day 5	5	15	4.57		16	48	14.64
Day 6	6	18	5.49		32	96	29.27
Day 7	7	21	6.40		64	192	58.54
Day 8	8	24	7.32		128	384	117.08
Day 9	9	27	8.23		256	768	234.16
Day 10	10	30	9.15		512	1536	468.33
Day 11	11	33	10.06		1024	3072	936.65
Day 12	12	36	10.98		2048	6144	1873.31
Day 13	13	39	11.89		4096	12288	3746.61
Day 14	14	42	12.81		8192	24576	7493.22
Day 15	15	45	13.72		16384	49152	14986.44
Day 16	16	48	14.64		32768	98304	29972.89
Day 17	17	51	15.55		65536	196608	59945.78
Day 18	18	54	16.46		131072	393216	119891.56
Day 19	19	57	17.38		262144	786432	239783.12
Day 20	20	60	18.29		524288	1572864	479566.23
Day 21	21	63	19.21		1048576	3145728	959132.47
Day 22	22	66	20.12		2097152	6291456	1918264.93
Day 23	23	69	21.04		4194304	12582912	3836529.87
Day 24	24	72	21.95		8388608	25165824	7673059.74
Day 25	25	75	22.87		16777216	50331648	15346119.48
Day 26	26	78	23.78		33554432	100663296	30692238.95
Day 27	27	81	24.70		67108864	201326592	61384477.90
Day 28	28	84	25.61		134217728	402653184	122768955.80
Day 29	29	87	26.53		268435456	805306368	245537911.60
Day 30	30	90	27.44		536870912	1610612736	491075823.21

     Seeing a table like the one above re-enforces the fact that if AI is truly growing exponentially it poses huge risks for us. Fortunately for us, exponential growth works great on paper, but almost always runs into physical barriers in the real world. For example try folding a piece of paper in half more than eight times. Its starts to become very hard. the thickness of your average A4 size paper is roughly 0.1mm.  If you can fold it 10 times it will be approximately 2 inches high but if you somehow manage to fold it 20 times..... wait for it....  it will be 172 feet high and lastly if you fold it 30 times it will be roughly 53 kilometers high. You can see where this is going. The problem is that we just don't have the capability to fold paper that high or eat that many candies in a sitting. AI is going to hit hardware limitations and processing limitations and to get it to reach those next steps (the 20th fold of paper or the 20th day of eating candies) we will need to spend enormous resources (time, money, personnel, e.tc.). At some point we will ask ourselves why we are doing this at all. We will ask ourselves if it is truly worth diverting all these resources to achieve some utopian AI fantasy, when 80% of the worlds people still live on under 10$ a day.

      Let me make it clear. I am not against AI, especially "Specialized AI". It is what allows us to have self driving cars, enable businesses to predict online purchases and allow credit card companies to detect fraud. What irks me is when rich, famous people hijack the podium and use their money to tell us about the dangers of general AI as depicted in science fiction novels and popular movies. Humanity is far from the "general AI" of the HAL 9000 variety. Before we reach there humanity will come to the conclusion that its just not worth it.