Why I am skeptical of AI doomsday predictions

“Is the Gods that create a monster, monster themselves or are they something much worse” — Loki, the Norse God.

Published in

Stacked Squares

10 min readJul 7, 2019

In Greek mythology, Hephaestus, the God of fire and metalworking, created golden robots and machines which are capable of undertaking tasks by having the intelligence to understand what the Gods want. We can see from this ancient story the fascination humans have in creating intelligent machines. The same fascination can be seen in the creation of the Mechanical Turk¹, which beat the likes of Napolean Bonaparte and George Washington in chess. Though it was an elaborate hoax, we can see how fascinated humans are with the idea of creating artificially intelligent entities. Today, Hephaestus’ golden robots and Chess automaton (Mechanical Turk) are a reality. But we have shifted our goals further. According to Turing’s Test², the machine must be able to hold a conversation with humans and humans should not be able to discern between the machine and humans, to determine whether or not a computer is capable of thinking like a human being. Google duplex³ prima facie achieves that.

Disclaimer: I’m not an expert in AI. Machine Learning²¹ is what is colloquially referred to as Artificial Intelligence.

I’d love to hear your opinion. You can comment below the article or send private comments if you are reading on a desktop. I’ll be editing the article even after publishing, to evolve. I’ve added relevant links and additional resources in the appendix.

Under the strawberry moon (18th June) on a deserted bridge, a couple of my friends and I were having random discussions. Discussion inadvertently (Hmm…..maybe not) moved to the subject of Artificial Intelligence. One of my friends mentioned a TED talk by Sam Harris⁴ (which I have embedded below), where Sam talks about the future with super-intelligent machines. Some of the arguments I presented to him is the basis of this article.

Sam Harris | Can we build AI without losing control over it? |TED Talk

There are many similar videos which predict AI taking over the world. My main problem with such predictions is that all of them consider that there will be no improvement in human intelligence (or improvements won’t be fast enough) and doesn’t even consider the possibility of new dimensions of consciousness humans could attain. I think it is a discounting view of what humans really are capable of.

Square peg for a round hole

Alexander Luria, a Soviet neuropsychologist conducted a series of experiments to compare the intellectual activity across different cultures. One of the experiments was to see whether the subjects can think in hypothetical situations. I’m inserting an excerpt from Chapter 4 of his book “The Making of Mind”⁵.

Another syllogism was presented: “In the far north, where there is snow, all bears are white. Novaya Zemlya is in the far north, and there is always snow there. What color are the bears there?”
“There are different sorts of bears.”
The syllogism was repeated.
“I don’t know. I’ve seen a black bear. I’ve never seen any others … Each locality has its own animals: if it’s white, they will be white; if it’s yellow, they will be yellow.”
“But what kind of bears are there in Novaya Zemlya?”
“We always speak only of what we see; we don’t talk about what we haven’t seen.”
“But what do my words imply?” The syllogism was again repeated.
“Well, it’s like this: our tsar isn’t like yours, and yours isn’t like ours. Your words can be answered only by someone who was there, and if a person wasn’t there, he can’t say anything on the basis of your words.”
“But on the basis of my words, ‘in the north, where there is always snow, the bears are white,’ can you gather what kind of bears there are in Novaya Zemlya?”
“If a man was sixty or eighty and had seen a white bear and had told about it, he could be believed, but I’ve never seen one and hence I can’t say. That’s my last word. Those who saw can tell, and those who didn’t see can’t say anything!”
At this point a young Uzbek volunteered, “From your words it means that bears there are white.”
“Well, which of you is right?”
The first subject replied, “What the cock knows how to do, he does. What I know, I say, and nothing beyond that!”

In this era, one would be hardpressed to find an individual who cannot answer a similar question. Even though individuals were able to think in hypothetical for a long time, but now almost every human can do it.

Another point which is always assumed in similar doomsday predictions is, they reduce intelligence to just information processing and the speed of doing the same. I don’t think that is the case at all. Human intelligence is not merely information processing (What it actually is? I don’t know how to define that). For a moment, let us consider human intelligence is just information processing. Processing capabilities of machines can be calculated in FLOPS⁶. In the case of humans, we don’t have a proper instrument/metric to measure our information processing capabilities. Human intelligence is, generally, measured only in terms of the output produced by humans, not the actual amount of information processed.

Let me try to calculate the amount of information processed by our brain.

By moderate estimates, the megapixel count of the human eye⁷ is 576 megapixels (My opinion is that the human eyes are better than that).

The biggest camera sensor has a resolution of 247 megapixels⁸.

The RAW image of 247 megapixels is at the very least 300 MB⁹ worth of data.

The RAW image of 576 megapixels sensor is nearly 1 GB¹⁰ worth of data.

Films typically run at a 24 fps (frames per second).

Even though human vision is much smoother than that, let us take 24 fps at 576 megapixels.

At 24 fps, human eyes capture data amounting to 24 GB every second.

We can discern the difference between screens with refresh rates of 90 Hz and 120 Hz.

So we can definitely consider that the human eye has the capability of capturing at least 90 fps. At that rate, the amount of data transmitted to our brain is equivalent to at least 90 GB every second.

This is just from the eyes. We have other four senses too. I don’t even know how to calculate the information gathered by other senses. At what bitrate is audio captured by ears? How to objectively quantify smell and taste (without arbitrary metrics)? For skin, there are millions of never endings. How to quantify data captured through the skin? I don’t know.

Human brains not only fetches the information, but it also processes and filters out the information we don’t need, identifies objects, classifies them, condenses all the past memory of that object to a sentiment/an emotion (Just like AI is supposed to do). Other than guesstimation¹¹ (estimation by guesswork), there is no method to calculate the actual processing power of a human brain. By that estimate, it’d cost $1 billion per hour to run a system with the same processing power as a human brain.

I’m not trying to be pedantic with all the numbers. All I’m trying to say is that we don’t really have a handle on what the human brain is capable of doing.

****************************************************************

Looking at our evolution to Homo Sapiens from another vantage point, one could say that we somehow improved by becoming significantly more vulnerable than our ancestors. When humans evolved, in almost every manner, humans became more vulnerable than their ancestors. We climbed down from the relative safety of a tree to the ground. We lost the speed in movement, that our ancestors were capable of. Our body is weaker than those of our ancestors. Standing up straight on two legs exposing our vulnerable body parts, especially neck, put us at a significantly higher risk. Look where humans are now. We are the apex predator in the animal kingdom.

Throughout human history, one thing we have excelled at is pulling the rug from under what challenges us. Fire is a destructive force and used to be worshipped. We harnessed it to serve our purposes. Lightning killed many humans. Now we use the principles of the same to power almost everything we use. Steam was an inconsequential product while boiling water. We harnessed steam to pull goods weighing in tonnes across land and water. Steam engines didn’t provide quick enough transportation. We built aircraft to travel across continents faster than the speed of sound. It is inhospitable conditions outside the Earth’s atmosphere. We built a full-fledged station out there. Pathogens were killing us in millions. We figured out we could build resistance to them by willfully injecting the very same pathogens in dilution (let’s say).

AI (machine learning) can create something different from what already exists. AI created a different sport¹². Not a completely new one. AI created a different language. Not a new method for information exchange.¹³ AI can draw like Picasso. Can AI create a new Picasso? Can AI system pull a new Banksy¹⁴? AI-powered systems may be able to dance like Micheal Jackson. Can it create a new Micheal Jackson? Could AI have invented Penicillin²²?

I doubt.

During the process of canonization of saints of the Catholic church, there is a role of Devil’s advocate¹⁵ (to check for confirmation bias). Even the Devil gets an advocate in the Vatican. I think it’s only fair to have someone arguing the case for humans. Even the people who reject AI doomsday predictions, do not make a case for humans capability. I believe that we should be embracing the full potential of AI. At the same time remembering Frankenstein’s monster as a cautionary tale.

In her book, Coming of Age in Samoa, anthropologist Margaret Mead¹⁶ details her encounter with a fishing tribe which was on the verge of extinction. The tribe was not able to cross the ocean and catch fish the way they used to do. The tribe used to make large special canoes, which they used to fish and travel to nearby islands. Over the course of time they, as a tribe, forgot the craft of making these canoes, which caused the existential threat they were facing. They were at the brink of extinction because they forgot their craft, forgot what made them exceptional.

Nowadays we seem to be replacing our brain with a smartphone and that’s a very troubling development. Let me put forth an observation, which I think is pertinent. When I was young, when a dispute arose about something among friends, even with objective (verifiable) facts, everyone used to argue, trying to prove their side. Everyone attempted to make the most logical argument they could construct in their favour. Afterwards (if it was an objective fact in dispute) they’d find out from the library or by some means the truth of the matter. The person who was wrong will get to analyse what was wrong in his/her reasoning, try harder to get things right and build better arguments for the next time. Nowadays if there’s a dispute, even with matters which are subjective in nature, people will whip out their phones and what Google says in its card or the first search result that is considered the truth. We are losing our craft to think critically and form arguments.

If you search for the term Loki, the panel on the search result shows that Loki is a marvel character. There is no mention of him being a Norse God in the panel. When Thor is searched, the 2011 movie comes up. At least when one searches for Cyclops, the panel shows information about the mythical characters, not the Marvel hero. If one has a rough idea of how Google search ranks webpages and know that the knowledge panels can be owned by individuals or businesses¹⁷, one would at the very least be skeptical about the search results (same applies to Wikipedia and other websites). It is quite interesting to know that children of tech employees in Silicon Valley are discouraged extensive use of technology²³.

! Spoiler Alert: The following paragraph contains details about Spider-man: Far from Home.

In the latest Marvel movie the protagonist, Spider-Man, has to wade through the most surreal Augmented Reality¹⁸ field fighting weaponized drones (which cannot be seen within the AR field). He has to rely on the knowledge that the field he’s in is an illusion and “The Peter tingle” (Spider-sense) to smash the drones and defeat the antagonist. We have our own versions of “Peter tingle”. Twenty years ago, there was no proof needed beyond photographic evidence. But now, most people are not easily hoodwinked by any photograph. We get a feeling when the photograph may be fabricated (unless it cements the confirmation bias). Similarly, we’ll develop “Peter tingle” for deepfakes¹⁹ and whatever that may come in the future. I believe we, humans, will be able to shift the paradigm against whatever challenges us.

Cherrypicking a statement from the movie, Jurassic Park²⁰ “I’m simply saying that life finds a way”.

****************************************************************

Appendix

You can read my previous post here, where I narrate how I hacked a national aptitude test with the information I gained by observing and figuring out what it meant.

¹Miscellaneous contributors, The Turk, Wikipedia

²Margaret Rouse, Turing Test, TechTarget

³John Callaham, What is Google Duplex and how do you use it?, Android Authority

⁴Sam Harris, Can we build AI without losing control over it (Transcript), Singjupost

⁵Alexander Luria, Chapter 4: Cultural Differences in thinking, Book: The making of mind

⁶Miscellaneous contributors, FLOPS, Wikipedia

⁷Dr Roger Clark, Notes on the Resolution and Other Details of the Human Eye, Clarkvision Photography

⁸Sebastian Anthony, Canon unveils 250-megapixel prototype DSLR camera sensor, arsTechnica

⁹ ¹⁰Peter Forret, Megapixel Calculator, toolstud.io

¹¹Nick Bostrom & Anders Sandberg, Whole Brain Emulation: A Roadmap, Future of Humanity Institute, Faculty of Philosophy & James Martin 21st Century School, Oxford University (PDF)

¹²Jon Fingas, AI developed a whole new sport, Engadget

¹³Roman Kucera, The truth behind Facebook AI creating a new language, Towards Data Science (Medium)