Wanting Omniscience in AI

Source: Deep Learning on Medium

Wanting Omniscience: The ultimate human quest

The story of why there is no such thing as full-stack AI because you can’t know it all.

Disclaimer: This not going to be your usual article.

AI is a very elusive and interesting field, uncovering the wonders of what computer vision classification network can do with a superhuman degree of accuracy or the elaborate text that language model can generate or understand the semantic meaning of, is fascinating, but truth be told you can’t know it all.

I have been fighting with this dilemma for a while now.

Not sure I can call it shiny object syndrome because most subfields of AI caught my eye and I have an interesting project that I believe it will bring value to the world. Currently, I’m working on one or two projects as you read this article. This was possible only after many years of anxiety attacks, impostor syndrome, shiny object syndrome, procrastination and mindless scrolling about new AI projects, ideas, papers and articles.

“It is during our darkest moments that we must focus to see the light” — Aristotle

It was the lowest point in my AI journey because I got stuck. I was stuck in my own head.

There is one day in specific that truly marked me, it was the day that I felt like I was having a heart attack. It all happened in 2018 while I was mindlessly binge-watching Youtube videos when I came across this 2h Fastai part II 2018 video lecture taught by Jeremy Howard(my AI superhero) someone whom I have a lot of respect for even though we never met, nevertheless he impacted me deeply and helped me become a better coder and AI practitioner. Now, going back to the story, the lecture was about image segmentation and I had this paper that I was trying to implement for months with no success only to see Jeremy blaze through it in minutes and it was in this moment I saw how little I knew about AI, in specific Deep learning, that I felt this cold chill followed by a knot in my stomach and then my heart started to race so much I couldn’t breathe properly, after the incident I was so scared that I wanted to buy a smartwatch to be able to see or predict such an event. I had never felt that way in my life. Later on, I googled the symptoms, only to find out it was a panic attack.

After some time it hit me like a ton of bricks, right after watching a video from another mentor of mine Patrick Bed David(another person I hold great respect for), I realized that what I really need was to design my own curriculum according to what I really like and need in order to succeed in this vast field. I needed to shift my focus and really find out what are the most important and fundamental aspects of the field I needed to learn. And so I did, I devised a year-long life-plan including everything and everybody for 2019.

“Don’t dwell on what went wrong. Instead, focus on what to do next. Spend your energies on moving forward toward finding the answer.” — Denis Waitley

If you ever felt this way, just know it is normal and you can get out of this situation stronger and better.

I found my answer by taking a few steps back so that I can take the double forward.

The Plan

Let’s go into the details of how I did this:

  1. I chose the courses I wanted to take for the year and topics I’m interested in, with the main course being Fastai 18 and the topics were Deep Learning for Computer Vision and Text Generating Networks.
  2. I choose the technical books related to the field, so I can immerse myself and dive deep and get to the tiny details that make everything. I didn’t want to use just another black-box method, tool, framework or architecture. I wanted to know why and understand the core of it.
  3. I came up with projects and pursued them with all the passion I got, no matter how crazy or outrageous the idea was or the means of which I arrived at the thought. You know what I’m talking about, those crazy star-striking ideas you have while brainstorming or when you are daydreaming, exactly, those ideas have a lot of power, don’t just ignore them next time, write them down somewhere and leave it for a while. Great ideas die prematurely because we try to reason logically about everything, 1+ 1=2 that’s what we were thought, but it doesn’t need to be this way and it is not always true. Sometimes 1+1 = 1, think different, outside the comfort of your box, the greatest things are outside those boundaries, trust me.

This 3 simple things compounded over time and culminated into a series of great and fun projects I have pursued and currently am pursuing. It doesn’t need to be the best and make the whole picture, but slowly but surely it will form one.

“The journey of a thousand miles begins with one step.” — Lao Tzu

Of course, everything doesn’t go as planned it goes even better than planned, somehow situations arise and you have to improvise and you end up with far more than you wished for.

“Ask and it will be given to you; seek and you will find; knock and the door will be opened to you.” — Matthew 7:7

I can tell you this, as 2019 is coming to an end, I can look back and see all the blessings that showed up in my year disguised as an obstacle and actually laugh at my past self for being too worried.

Perhaps, I should have dug deeper into one specific topic such as Computer Vision and that knowledge would easily generalize to other areas instead of trying to know everything superficially as I did for most of the year, luckily my accountability plan and the life plan kept me grounded and accountable for not reaching the metrics I set up, so I did my best and turn the boat around. This was the biggest take away I got from all the work, projects and courses I have done in 2019. I’m lucky to say that I overcame it and completed 90%+ of the plans I set up for myself this year, with this article pushing the metrics even further because of the history and emotions it carries.

This article is for you that is struggling right now and panicking over not reaching your goals. Just know it’s perfectly fine, you can take two steps back, make a Life Plan for 2020 with everything you want to achieve, perhaps add specific deadlines, then take some time to recharge enjoy this time with your loved ones because 2020 you are going to write all the wrongs and it will be the year that you will make a blip on the radar of the AI field.

This is the biggest lesson I got from 2019 and I wrote this article for you.

It is a bit different from the usual but I felt like I should share my personal story because there is someone or even you out there that are going through the same experience as me and that can benefit from knowing a way to overcome it.

You got nothing to lose, give it a try.

A note on Omniscience

The state of knowing everything.

credits to google images

We human beings still hold the title of the most intelligent beings on planet earth and perhaps agreeing with some scientists, the only intelligent organisms in the universe(though not for long), we were not designed to store and process every bit information available around us, one cannot know it all. From the 11 million bits of information that we take in every second from the environment through our sensory system, we only consciously process 10–50 bits of information per second while most of it is delegated to subconscious task handlers or simply discarded, according to the book Life 3.0 by Max Tegmark (recommended).

But even if we could know it all, in theory, it would be for only a fraction of a second because the law of imbalance and equilibrium would not allow the variables to stay in equilibrium for long, because circumstances in which that information was held as true changes. Only a few laws are objective, final and sound, such as the law of nature and the universe.

The Law of Imbalance and Equilibrium. In order for anything to evolve to it’s next appropriate level of existence — to grow — it needs to be prodded or stretched out of its state of equilibrium into a state of imbalance at which time the forces of Nature are engaged to create a new state of equilibrium at a new level.

Perhaps our coming descendants’ AGI(artificial general intelligence) will have the storage capacity and processing power to uncover deeper insights than us and front a new era of information processing. If AGI recursively improves itself to a point where it surpasses the intelligence of all humans combined, being able to take technology as we know to a level far beyond human comprehension, maybe then it can help us extend or “break free” from the limitations of our organic bodies and rise to a higher plane of understanding.

While that doesn’t happen it is crucial that you be careful and patient with the process of acquiring, remembering and using information, in any form, be it a book, blog, analogue (conversation and etc…), video or audio.

You cannot be omniscient, you cannot know it all.