Paul's Blog

Bookmark this to keep an eye on my project updates!

View My GitHub Profile

Two Mysteries Solved

Rain lashed against the window, the rhythmic drumming a familiar lullaby to the old dog, Milo. Curled beside the fireplace, he watched the embers dance, their flickering glow painting memories on the backs of his eyelids. He remembered chasing butterflies as a pup, the exhilaration of the hunt, the sweet scent of summer meadows clinging to his fur.

A twig snapped outside, pulling Milo from his reverie. Ears perked, he lifted his head, instantly recalling the countless times that sound meant a raccoon prowling for scraps. His body tensed, not in blind fear but in calculated caution. Years of experience had woven a tapestry of memories, each thread a past encounter, each knot a learned response.

He didn’t need to see the raccoon to predict its next move. It would circle the house, drawn by the scent of dinner, then attempt an entry point - the back door, most likely. With a sigh, Milo thumped his tail against the floor, a preemptive warning.

His human, Sarah, stirred on the couch. “Something wrong, boy?” she mumbled, her voice thick with sleep.

Milo whined softly, his gaze fixed on the back door. Sarah followed his line of sight, understanding dawning in her eyes. She rose slowly, the familiar creak of the floorboards another thread in the vast tapestry of their shared memories.

Together, they moved as a unit, a silent symphony honed by experience. Sarah grabbed a flashlight, the beam a comforting presence in the dark. Milo crept towards the back door, using his knowledge of the house’s layout to position himself for a strategic intercept.

The raccoon, as predicted, was there, its beady eyes reflecting the flashlight’s beam. A growl rumbled from Milo’s throat, a sound born not just of instinct but of the countless past confrontations that had taught him the language of dominance. The raccoon, recognizing the futility of its attempt, slunk away, melting back into the storm.

Milo relaxed, a satisfied rumble escaping his chest. He hadn’t needed complex commands or spoken language. The dance between him and Sarah, the memories woven into their very being, had been enough. That quiet victory was a testament to countless shared experiences. There, in the powerful interplay of memory and prediction, flickered the essence of intelligence.

What Our Best Friends Can Teach of Intelligence

Milo’s ability to react intelligently provides one of the most fundamental insights into intelligence. As we consider the growing impact of artificial intelligence on our everyday lives, we need to understand what makes something truly intelligent.

Current Machine Learning (ML) systems use the stimulus-response approach. They excel at rote tasks but cannot do high level reasoning nor anything that requires planning or novel problem-solving.

What is “real” intelligence? Is it simply a matter of complex behavior? Or is there something more to it? I believe the most convincing definition of intelligence was given by the co-creator of the Palm Pilot, Jeff Hawkins, who posits you need two things to be intelligent: memory and prediction.

Hawkins argues that the true definition of intelligence is not determined by how an organism acts. Instead, it’s the ability to form memories of past experiences and use those memories to predict future outcomes. This concept goes beyond just reacting to stimuli.

This reframing of intelligence has profound implications for our pursuit of machine learning & intelligence. Traditionally, ML has focused on mimicking complex human behaviors such as playing chess, translating languages, and recognizing objects. However, Hawkins’ definition means we need to move beyond replicating actions and instead focus on building machines that can remember and predict. Using memory and prediction, we can develop much more flexible and creative machine intelligence.

TLDR

Intelligence is not behavior; it’s memory plus prediction.

Ben

The insistent trill of Ben’s alarm ripped him from a dream of scaling a precariously balanced stack of breakfast buffets. Groaning, he fumbled for the offending device and silenced it.

Ben was a creature of habit. Shower, shave, and the same worn-in Oxford shirt. In the kitchen, the dance began—toast the bagel, a smear of plain cream cheese, then a dollop of ruby-red strawberry jelly. Next came the pantomime of whisking eggs, the sizzle as they met the hot oil. Coffee, dark and strong, brewed while he devoured his breakfast masterpiece.

Fueled by caffeine and carbs, Ben headed out the door. As he rounded a corner, a figure materialized from an alleyway, shrouded in a lab coat that seemed to defy the summer sun.

“Benjamin Davies?” the figure rasped, voice like sandpaper. Ben stopped, surprised.

“Uh, yeah, that’s me,” he replied cautiously. The figure stepped closer, revealing a face dominated by outlandish goggles and a wild mane of white hair.

“Splendid! A candidate for the Chronogastro Experiment!” the figure declared, brandishing a device that looked like a futuristic toaster crossed with a bicycle helmet.

Before Ben could react, the figure zapped him with a beam of light. A wave of nausea washed over him, then receded as abruptly as it came. Blinking, he stared at the scientist, who was now packing away his contraption. “There,” the scientist chirped, “memory of your breakfast successfully erased! Now, go forth and contribute to the advancement of scientific knowledge!”

Ben, utterly bewildered, mumbled a thanks and continued on his way. Reaching his office, he was greeted by the usual morning banter.

“Morning, Ben! Coffee?” Sarah, his co-worker, asked, holding up a steaming mug.

“Sure, thanks,” Ben replied, accepting the mug. “Actually, I don’t think I had breakfast this morning. Feeling a little rushed.” A chorus of surprised gasps filled the room. Ben froze—hadn’t he just devoured his bagel masterpiece minutes ago? A strange sense of unreality washed over him. Had he dreamt the entire breakfast? Or was this some sort of elaborate prank?

It was a reminder, a quirky one to be sure, of the messy beauty of human experience. Memories, even the seemingly mundane ones like breakfast, were the building blocks of our day, our narrative.

Shrugging it off, Ben sipped his coffee, the bitterness strangely comforting.

TLDR

Our memories are a big part of our consciousness.

Consciousness Defined

A fundamental question has not yet been answered when trying to engineer machine intelligence: What is consciousness? We need to lock down the definition before we can progress.

Hawkins proposed a definition that makes the most sense. Here’s the gist:

There are four elements which combine to form consciousness: Self-model, memory, senses, and qualia. A self-model is another way to describe self-awareness. It’s basically an internal representation of yourself, your actions, and your place in the world. In a mental simulation, you are yet another object that can act and be acted upon. Memory is the storage of input, i.e., people, places, things, experiences, symbols, etc. Senses are the ability to perceive the world, whether through organs, cameras, microphones, or lines of code. They are essential for navigating and interacting with your environment. Qualia is the subjective experience of things—the redness of a rose, the tang of a lime.

These factors easily apply to human consciousness. For example, when you go to sleep, you have almost no memory, sense, or self-awareness. Your qualia is limited to your dream world. We call this a state of unconsciousness. So, these elements accurately explain consciousness, but they also provide a perfect blueprint for creating synthetic consciousness.

Here’s the beauty of this approach: We can focus on building these core functionalities without descending into the philosophical quagmire of consciousness. We can scale or throttle all these aspects of consciousness until synthetic consciousness emerges in the state we’ve imagined. In other words, we have an unproblematic formula for consciousness.

Imagine an AI with a phenomenal self-model, a bottomless memory bank, and an uncanny ability to perceive the world.

TLDR

We can build scaleable, intelligent synthetic consciousness.

Part 4: Architecting Synthetic Intelligence

TOC