Waking Up
…we were finally on our way home, leaving the tall Pyrenean mountain chain behind, as the city of Zaragoza came into view after a long stretch of dry nothingness. We were driving by the military complex Capitan Mayoral in the outskirts of the city when we suddenly heard a loud, echoing “thump”. Immediately after, a mechanical object swiftly raised above the steel-roofed barracks in the distance, describing a perfectly straight upward trajectory. As gravity slowed it down on its way up, it abruptly spun in mid-air, before falling back down. With the precision of a clock, we saw the enigmatic object repeatedly raising above the buildings, almost like a kid bouncing on a trampoline.
As we drew closer, I could finally recognized what we were looking at. I had seen the prototype on social media about two years ago when it was presented to the public. I could now clearly see its sleek white and black frame and its two cameras, which like eyes, appeared as two spooky tiny red dots against a dark background. The outer design was clearly reminiscent of a mechanical jellyfish or an octopus. For a second, I thought to be staring at one of the alien creatures invading Earth in H.G Wells novel “The World of Wars”. But no, it was a TNTCL, the latest droid from the robotic company LESTA.
From up close, I could see now TNTCL was using its hydraulically-activated tentacles to propel itself upwards into the sky. But most impressive was the beautiful spin it was performing at the top, like a ballerina thrown into the air, its tentacles were extending in a star-like pattern before closing down again, repeating this choreography like clockwork, with a loud “thump” as it was flew upwards each time. We stopped the car in awe to contemplate that magnificent display of locomotive prowess.
And that’s when it hit me. A sense of fear rushed through my spinal cord as I understood I was being fooled by the external beauty of that futuristic jelly-fish. The real beauty of TNTCL did not lie in its outer design, in its intricate movement patterns, in its capacity to propel itself upwards, or in using its mechanical tentacles to spin its own body. No, that was just the surface. The real beauty was inside, in its inner design, in understanding that, while the outer design had not changed in the last two years, while the outer design had been frozen, fixed, and it would remain unchanged for the foreseeable future, the inner design was in continuous refinement. That’s where the real engineering feat laid, in that TNTCL had no been programmed to “jump” or to “spin”; all the designers had to do in order to witness TNTCL propel itself upwards into the air was to let the self-reinforcement algorithm do its magic. Like a toddler in the process of getting to know its own body, getting to know the “body” the “mind” has been put onto, what we were contemplating was TNTCL inner’s algorithm at work, “the mind”, slowly getting to know thyself, getting to know the possibilities and boundaries of its own mechanical body. THERE, laid the beauty: that all that intricate choreography had happened sporadically, that TNTCL had learned, on its own.
I woke up startled. My heart was racing as my dream morphed into a very clear and frightening realization: perhaps TNTCL wasn’t real yet, perhaps it was just a product of my imagination, but the possibility of such an intelligence machine acquiring consciousness seemed all too close — or perhaps, even already too late.
You see, so far, expert discussion around artificial general intelligence (AGI) had revolved around ChatGPT-like large-language models (LLMs). Experts were convinced LLMs were the most advanced form of AI, the most intelligent, and therefore those posing the greatest existential risk to humanity. In particular, those with the capacity of what experts termed reasoning, an emulation of the human logic and inference mechanisms of the human brain. This humanity was captivating, but it was a mere trick; not reasoning per se, but an imitation of reasoning: what reasoning looks like from the “outside”, without the “inside”.
Moreover, LLMs lived in a “box”, in a computer, disconnected from the rich sensorial sphere of the outer world. Instead, TNTCL was in touch with reality, and the quantity of stimuli it could process through its various sensors was far beyond simple text. As a result, despite not being able to “talk” or articulate thought, despite being seemingly so distant from the human thought process, not only it could learn much faster, but its learning process followed much closely that of the human brain, by placing real-world interaction at the centre of the equation. So while LLMs were being trained like parrots, TNTCL was being trained like a human.
And while experts still debated, dismissing the existencial risk posed by TNTCL, for me it was clear: the closeness of the rich sensorial experience of TNTCL to that of humans made it the front-runner in the race towards consciousness. But the big problem was now for mankind to “wake up”, to realize, that, our lack of understanding as to how consciousness arises might be our biggest threat in the development of AGI and that, by the time we wake up, it might be too late.
Thanks for reading! Scroll down or hit the follow button for more thought-provoking articles :)
