Storyless-centric machines!

""Note: I wrote this article more than a year ago, but I just watched Yuval Noah Hariri's talk 4 days ago on AI and the future of humanity, and many ideas actually overlapped. So I am posting it here now"":

“This article is super interesting, but the introduction needs a story”, said my friend, whom I shared the first draft of this article with. Then she added, “the insight is very much thought-provoking but the introduction needs a better story for the readers”. A good story is not a need for the readers only. In fact, all of us, humans, need stories to understand and make sense of the world around us. For example in the business world, ”What's the difference between a good brand and a great brand? A story! Right? We all know the answer to this one. Even in leadership and managing teams, one key element of good leaders is clear communication and the ability to unify a shared vision for their teams, in other words, forming a productive story.?

The fact that we believe in stories and engage in storytelling sets us apart from all other species. As Yuval Noah Harari explained in his (masterpiece) book Sapiens, humans would not have been able to cooperate together with flexibility and in huge groups, if it were not for the power of storytelling. No other species can do that, except us (humans). Even if other species can see and deal with objective reality much better than us in most cases. Harari said we had to invent and believe in stories such as the story of money, the story of countries' borders, laws, and many other stories. Without them, we wouldn't have been able to build and manage this "civilization." A story is our most fundamental way and basic building block of human learning. But what about machine learning? And how will we as story-centric creatures coexist with such storyless intelligent machines that we created? This is the “story” that I am concerned with in this article!

Stories are the fundamental in how we comprehend reality and life's patterns, influencing our decision-making processes. They serve as a means of receiving and conveying information, extending beyond fictional narratives to encompass various forms and subjects, including scientific theories. The human inclination to impose stories in different contexts stems from our need to make sense of life. For example, although machines excel at pattern recognition and correlation without the need for stories, we often desire them to generate story-centric outputs. We still want them to use their capabilities to create stories for us. Stories such as an AI piece of music, a well-plotted movie script, or the IBM-powered debater machine that performs debates on stage with humans. These make good stories for us, or more precisely, necessary stories for us to understand AI! Since we can't understand it, or communicate our understanding of it in any other way. On the other hand machines can't understand stories or, in other words, don't need stories to learn, or communicate. Well, they don’t need stories to understand realities, because they have facts!!

Stories are vital for human communication and understanding, providing a comfortable means of conveying complex information and connecting facts to form meaningful patterns. Ultimately, stories shape our perception of reality, and when shared by a large group, they can become the accepted reality for that collective.

But what is the problem with stories? Actually, the problem is the idea of needing a story in the first place! Every human can have a different perception of reality "based on their story". However, if a large group of people believe in the same perception of reality, wouldn't this make it a reality for this group??Harvard University has a very famous course about storytelling and public narrative. So the problem is our natural tendency to link the facts and realities in a story-like form in our heads. Even in the most pragmatic and sophisticated domains, we still are incredibly susceptible to stories. It affects us and our judgments in good and sometimes in bad ways. I remember in January 2019. I had the pleasure to go with some of my colleagues to EPFL in Lausanne during the Applied Machine Learning Days event, where Gary Kasparov (the Russian World Chess Champion and grandmaster) was there giving a talk. And he was talking about his experience playing and losing against deep blue in 1997*. During the event, he was asked if there is a clear difference between playing against an AI and humans, and he said yes, it feels different. He clarified that with a human player, he could quickly identify if the player is more of an attacker or a defender, or if the player is way more skillful with the horse over other pieces, for example, or if the player has a specific style. While playing against a machine, you can't spot any of that. Every single move is simply a new game for the machine, with no particular style or strategy to spot. Thinking about what he said now, I honestly believe Kasparov wanted to say that he couldn't form a story about his opponent in the case of playing against a machine! He couldn't develop a story of the opponent's style or of the game! But our brains will make up a story anyways to function, even if it is not a correct story. Maybe Kasparov's story at that moment was as small as asking himself will I be the first chess grandmaster to lose to a machine on public TV?! Because of this, every mentally strong person can keep the chatter in their head under control and can talk positively to oneself. To ensure a positive and productive narrative because we are unable to go storyless, so better to make a story with a positive narrative then!

The question here is, don't these stories distort realities sometimes (many times actually)? Some will argue it may help, on the other hand, to have a reference "story" to lean on, and it might be useful in some situations. But only with humans, not with machines!?

And this is precisely the point. AI doesn't understand stories, and more precisely and importantly, doesn’t need stories to recognize or communicate patterns. It can see through the facts it is supposed to see, avoiding any possible blurriness or distortion caused by any form of a story. Some would still blame AI for being biased in not recognizing darker skin pedestrians in an autonomous driving algorithm, or for selecting more men CVs than women CVs on a job applications screening algorithm. However, the reality is that AI isn't biased. It just magnifies our biased stories of reality in a very unbiased, efficient, and objective way. And we still make stories out of it, which plots and confirms our storyline of biased and scary AI, rather than blaming our biased historical data where the algorithm learned from in the first place.?

AI can do only one thing, but it does it really well; this thing is pattern recognition (or mathematical likelihood, if you may). Yes, it can do it on a gigantic multi-vector and multi-layered scale, that can enable the algorithm to not only analyze, but also operate and predict realities. However, it is still the one thing, which is the one superpower of AI, pattern recognition ability. It's not memory, storage, security, or computing; these are the superpowers of other technologies, but for AI, the superpower is certainly the extremely sophisticated pattern recognition ability. But to be more precise, the true superpower of AI is pattern recognition with no stories needed!?

In the (excellent) book Life 3.0, Max Tegmark expected and explained many futuristic scenarios in what role AI will play and how it will coexist with humans. Tegmark had to plot stories for us to understand these scenarios; why? Well, he needed to do this for us, the readers, to make it more exciting because we can't do it in any other way than stories. Despite that, most of these scenarios are fact-based predictions, our human way of wiring; these facts are simply through stories, which can ensure an understanding of the scenarios but certainly can't guarantee a natural and factual outcome.??

On the contrary, AI's storyless factuality is the core of its intelligence and superiority! Apparently for humans, being story-centric creatures was and still is our leading superpower among other species, yet looks like it might be our narrowest bottleneck and main weakness among machines.?

The critical question is, 100 years from now, will we be able to write a good story of our collaboration and coexistence with machines, or will machines, based on pragmatic facts, put an end to a humankind story that they never understood in the first place?!??

?

*Deep-blue computer that won against Gary Kasparov is not considered a proper AI algorithm by many experts. It was more of a high processing and computing machine at the time that was able to make decisions based on probability calculations. Unlike AlphaGo or Alpha Zero that were developed by Deepmind in 2016 and 2017

#aiforgood #ai #deepmind #yuvalnoahharari #storytelling #story #humanity