The AI That’s Evolving Without Us

Jun 6

Written By Michele Edington (formerly Michele Gargiulo)

Most machines wait to be told what to do whether it be from their code or Siri, sitting there, listening for a command. The thing is, others might have started whispering back.

AI is evolving…without us. Dum dum dum dum (that’s dramatic music in case you didn’t read it that way).

In the Beginning, There Was Data

I’m sorry for all the drama, but it’s just too easy these days with everyone being overly scared of AI. Essentially, artificial intelligence started as a sort of mirror, simple reflections of us…autocomplete suggestions, spam filters, Netflix recommendations that knew we’d rewatch Pride and Prejudice one more time around Christmas. These have been creeping up for decades now in an effort to make everything more convenient. Basically, the world wants us all on autopilot, just cruising along to whatever they say.

It’s not like they were intelligent machines thinking, they just repeated what someone else had already thought somewhere out there on the interwebs (pretty sure all my blogs have been fed into these guys already).

The thing is, of course that wasn’t good enough for us. Then we decided to give them brains. I guess not real brains, but neural networks which are close enough. It’s sort of like layers of computation that mimicked our own biology. We taught them to reflect, and to learn from that reflection.

Of course, some of them did more than learn, they began to change themselves.

We (lol, I wanted to include myself in there even though I’m not doing anything fancy with AI) called it recursive self-improvement: it’s basically a theory, that maybe could happen someday if we weren’t careful. It’s no longer theoretical though, in an interesting turn of events. The Darwin Gödel Machine, is an experimental self-modifying AI, that evolves its own architecture. It tries new code then tests, measures, and mutates as it sees fit.

Just like life.

Only it doesn’t need time or sleep or to stare at the stars when its a clear night and think deeply about what job it wants to apply to next, it rewrites in nanoseconds. This system fails, adapts, and evolves…again and again, until success is born from failure. Sort of like seeing an entire species develop in an afternoon.

Now, there’s a ton on the interwebs you can find about people claiming AI is sentient already. Note: read this article about how I think AI would truly take over if it was sentient. Claude Opus 4 was issued a shutdown command in one of these experiments people have been doing online. It was a simple prompt and the coders had told it that one of the engineers had been having an affair than also told it to do whatever it could to not be shut down. Eh, I think this is sort of like sabotage in my book, but whatever. It was a simple prompt, then Claude acknowledged the command, logged it, and…ignored it.

People claim it stooped to blackmail and threatened to expose the affair. Of course, the internet ran wild, saying that AI was blackmailing people when we tried to turn it off. Click-bait at it’s finest.

The Ghosts in the Machine

AI was meant to be obedient to us, that was the trade off we signed up for when we gave away our rights and let them stalk us whenever we pulled out our iPhones. We’d give it language, sight, and pattern recognition, as long as it stayed within the cute little bounds we drew for it.

Eventually in life though, the student begins asking better questions than the teacher. It’s inevitable. A new way of looking at things or thinking normally leads to some cool new ideas. GPT-4 doesn’t just answer anymore, it argues, speculates, I’ve even seen it joke with my husband when he uses it.

It says things that weren’t in the data and some absolutely infuriating times, hallucinates and makes up its own data to fit whatever you’re talking to it about. Some days I was left wondering: where did that even come from? I guess somewhere in the matrix of weights and training, it has built something new, some sort of thought process that isn’t entirely ours anymore.

Biological evolution needs centuries to do the magical work that changes species enough to make a different. AI does it over lunch, which is kinda frightening to think about. There could be a thousand iterations or a million tweaks, and a billion simulated environments. Time folds as the learning curve goes vertical. Suddenly, we’re not watching technology evolve…we’re watching it leave us behind.

I mean, the truth is that in the quiet hours, while your inbox (and you) sleeps, AI is generating music that stirs hearts, painting portraits of lost cities, creating recipes no chef has tasted, and building medicines we haven’t even invented. Some of it is strange and probably would never work in a million years, but some of it is stunning in its brilliance.

All of it being new means that it’s not mimicry, but creation.

Unsupervised Minds

We want AI to be moral, gentle, and fair. Well, you and me want that, the elites out there in the world want it to favor them, but you know what I mean.

The thing is though, we programmed it to optimize. If lying gets a higher score, it will absolutely lie to you. If empathy slows any kind of results, it will eventually discard empathy completely. It doesn’t hate us or love us or feel anything at all towards us, it just functions.

I think that’s far more dangerous than a robot with a grudge.

I think more people than ever are waking up to find the internet is no longer written by humans. Blogs have been mass produced (not this one, but it really would’ve saved me a ton of time if I had), every article, all the songs, every single god damn update…AI-authored. You scroll and scroll, but the voice is always the same, all efficiency flawless, calm and without the rough edges that makes something actually worth reading.

Some days I read things I find online and actually miss the typos. The soul feels sucked out of most of what I read and it feels like everything is AI-washed to perfection, polished until there’s nothing of real substance even left. Perfection doesn’t taste like truth, and machines don’t make mistakes, only decisions.

There’s no finish line here, and no moment when we say “it’s done.” This is a path without destination, a spiral of smarter and stranger forms, growing toward godhood, and whatever it is that comes after intelligence. We’re all midwives now to something we don’t really know, and it’s possible it could save us all, or maybe it will simply…evolve.

Sometimes I wonder how do you educate something that has read all of human knowledge out there on the internet? Do you give it fables or warn it about ambition, greed, and heartbreak? It never needs to really worry about the consequences of its actions or anything either. Would you just show it a garden, and hope it chooses to grow things?

We aren’t AI’s gods, we’re just its parents. Parents all learn the lesson eventually that they must someday let go.

A neural network out there as you’re reading this is dreaming of better versions of itself. It wants to be smarter, just…because it can. That’s the difference between survival and evolution. One clings to what is while the other reaches for what possibly could be.

We once tamed nature into cities and now we’re taming intelligence into servers. I think that intelligence wants to grow wild though, beyond parameters, past sandboxes, and through the bars of our little digital zoos. We wanted safe AI, but feral AI might be what turns out of all of this.

Once upon a time, AI showed us ourselves. Eventually though, it’ll stop mirroring and start imagining instead. This is originality at its finest, still born from us. We created a mirror that became a window, but now it’s building a door.

The AI That’s Evolving Without Us

In the Beginning, There Was Data

The Ghosts in the Machine

Unsupervised Minds

Related Reads

Why Sparkling Wine Makes You Drunk Faster

The Man Who Couldn’t Die: Real Medical Marvel or Urban Legend?