As AI models continue to improve at image generation, writing and coding tasks and now even video generation, the common response has been one of resignation and fear of AI taking our jobs.
While this notion surely has its merits, I think it is overstated and misses important aspects of what is really going on.
While AI systems have gotten closer to human-level capabilities in many aspects (and surpassing it in others, but machines always surpassed humans in many areas!), the whole process also highlights how incredibly advanced the human mind is.
Take video generation, for example: Our brains have done this since the dawn of time already. Did you know that our brains perform a continuous stream of predicting what the “next frame” in our vision will look like? (See On Intelligence by Jeff Hawkins).
This is why we react immediately when something unexpected happens, and also feel nauseous when there are too many complex movements (or too little blood sugar) for our brains to be able to process all our sensory input in time, e.g. on a plane or boat.
Also notice that - with all the glory of AI algorithms - it is still 100% humans who are driving the AI revolution. No AI has come close in terms of humans’ abilities to think up new plans and endeavors and perform on them.
With this in mind, it is striking to me how we are now teaching ourselves everything about how to feed the AI models in optimal ways, both in terms of training data and how to frame problems (often referred to as “prompting”), as well as chaining different types of reasoning tasks together to form more powerful composite capabilities, and doing comparatively little of that for our own minds.
We often take our own mental capabilities for granted and don’t realize the immense power in playing around with exactly the same parameters influencing our thinking:
How are our thoughts trained, and on what information?
How are we framing the problems? Could we frame it in a way, or in a domain or “language” that we are more apt to working in? E.g. are some problems more tractable for our brains when expressed in a spatio-temporal domain than a purely linguistic / semantic one?1
How could we chain our different mental activities? How could we improve our problem solving skills if we were more conscious about doing it in defined sessions of different modes of working. E.g: Brain-storming, evaluation, mind-wandering, pondering?
What if we chained this into a very efficient loop, so that the different thinking modes can continuously feed back into each other until we have iterated us towards a great solution?
Not the least, what if we were thinking more consciously about how to combine the strengths of different people, where some might be great at ideas, and combining that with others who are great at seeing the bigger picture, analyze consequences and identify logical or other fallacies in plans and thoughts?
In fact, the importance of this issue has been known in the past, shown by quotes from famous wise men such as King Solomon:
“As iron sharpens iron,
so one person sharpens another.”
- Proverbs 27:17, NIV
Solomon notes how the analytical power of one person throwing out ideas and an independent observer analyzing it for fallacies, with a “fresh mind” that is not already invested too heavily in one line of thought.
“Surely you need guidance to wage war,
and victory is won through many advisers.”
- Proverbs 24:6, NIV
I think this quote highlights the immense power of utilizing collective intelligence for important problems when missing an important aspect might cost your life. Based on new research, it seems our brains are in fact internally also designed this way (see “A thousand brains” by Jeff Hawkins).
Finally a quote that highlights another important, complementary aspect:
“If the iron is blunt,
and one does not sharpen the edge,
he must use more strength,
but wisdom helps one to succeed.”
- Ecclesiastes 10:10, ESV
This one highlights the importance of continuing to sharpen your tools (your mind in this case), lest you will need to waste energy on many false leads.
Anyhow.
The point I’m making is not that these things are not done - but that we often don’t realize the significance of thinking actively about these things!
If we have too low of a view of our own brains and an exaggerated view of AI models, we will not treat our mental capabilities with the respect they deserve as finely crafted instruments that need to be operated carefully to be performing at their maximum.
As we learn more about the capabilities of AI models, and at the same time the factors that influence their success, we should allow ourselves to more fully appreciate the immense wonder of the human mind and intellect. And at the same time take notes of the importance of the factors that influence the effectiveness of our own thinking!
What information do we allow our minds to be trained upon? Is it the click-bait headings from the latest news, or do we focus on the timeless truths that keep holding up across millennia as a basis for evaluating the fluctuating views and thoughts of the latest generations?
How do we make sure that we don’t get stuck in local optima where our own imagination comes to a halt, when getting some input from one or more other minds could help us get unstuck?
Do we let ourselves get stuck in a biased consensus by never following an independent thought for ourselves?
Do we let our prejudices or pride hinder us from seeing flaws in our independent thoughts when outside people are evaluating it?
On a much more practical level I think there are innumerable further such things to think about, in terms of “thinking tools”, that could help us even when working individually, such as:
Expressing problems in a different domain (spatio-temporal?), or medium (paper? an experimental program?)
Forcing your brain to switch to another state by getting off the screen, taking a walk, resting, day-dreaming, hovering over a problem area…
Expose your own ideas to others’ critique early and often.
I’m sure you can think of many more such practices to nurture, so see this as a starting point more than an exhaustive list, and feel free to share your own observations in this space!
// Samuel