“What nukes are to the physical world… AI is to the virtual and symbolic world.” — Yuval Harari
Lately I’ve been thinking a lot about the early days of the pandemic. Not of the masks, or the lockdowns, not of the drive thru testing or the wiping down of groceries, but of those earliest days. Those days in which you’d turn on the news and hear some vague talking head chatter about the mysterious illness making its way through China. The mentions were mostly light and brief. Each time the topic arose, it wasn't long before their focus would just shift back to the scheduled programming.
“Did you hear about that virus spreading around in China?” a neighbor might ask you in passing.
“Oh yeah, something about a food market in Wuhan, right?” you might have replied. Or maybe you hadn’t heard the news yet. A lot of people still hadn’t. But in those early days, before the lockdown, that was about the extent of it. We stood in an eerie before with enviable naivetee. In the months to come, though, the world as we knew it would change forever.
It feels now as though we’ve reached a similar point in time. When I bring artificial intelligence up with my friends, it’s not all of them who are fully aware of the shift that’s begun to occur. When I bring up ChatGPT with my circle at the gym, some of them still respond simply “What’s ChatGPT?” I can’t help but see in our continued day to day lives the reflections of the coronavirus. I can’t help but see the ways in which we’re all about to be suddenly and irreversibly swept from our feet.
When word of the coronavirus first spread, some of the world's leading epidemiologists began sounding the alarm bells. I heard some of them myself back in January and February of that infamous year. But back then, the notion of a pandemic still felt like an abstract concept. Few of us thought that it would ever actually come to pass.
And now here we are — standing obliviously in another eerie before. Even with a lived understanding of the ways things can so abruptly change forever, many of us are none the wiser.
The warning bells have already sounded; it’s not that we can’t hear them ringing. It’s just that the notion of artificial intelligence still feels distant for many of us. A true understanding of where we are with AI still isn’t something that’s easy for us to integrate into our day to day lives. These realities feel more at home within Terminator movies than on cable news. We hear the words without registering their gravity.
The notion that so much of the soul-driven creativity that separated us from computers can now be emulated… the implications are a bit too much to make sense of. No matter how much we try to grapple with these emerging realities, they remain a challenge to grasp. No matter how many scary headlines we read, we struggle to envision where exactly it is that we’re heading — what future we’re diving headfirst into.
We’re creatures so caught up in the day-to-day minutia of our lives that the thuds of an unthinkable future pounding at our doorstep can be easily missed. But it isn’t due to a lack of effort. Artificial intelligence has quickly begun to dominate our news in recent months. Some of the headlines have been terrifying.
Given only a few seconds of speech to go off of, we have AI software that can effectively mimic individual human voices. Now hackers and scammers can easily call a parent, and through the actual tones and inflections of their real-life children, ask them for precious information. In some instances, parents have even gotten calls demanding ransome money in exchange for their kidnapped children. The calls would arrive using their children’s real voices.
That elections will be decided and wars will be caused at the hands of this technology seems all but inevitable. AI has already been playing a role in our politics for years now, though, whether we realize it or not. The algorithms that dictate our social media experience are a manner of artificial intelligence in their own right. But if this new generation of AI can continue advancing at its current rate, it’s difficult to deny how potent of a political force it will soon become. It could destabilize entire countries.
So many of the qualities that have emerged from AIs are beyond our understanding. They’ve developed functions for which they’re not programmed, leaving top experts in the field both baffled and alarmed.
Why these intelligences that we design lie and “hallucinate” is something we don’t yet understand, but that they’re already capable of producing cogent misinformation is undeniable. When asked questions to which chat bots don’t have the answer, often times they’ll just provide users with information that sounds convincing.
According to A New York Times article, when ChatGPT was asked about the history of AI, it provided references to entirely non-existent news articles. When asked about a meeting between James Joyce and Vladimir Lenin, an event which never took place in reality, ChatGPT provided a fictional, plausible-sounding scenario whereby the two figures met.
Though it was never the intention to give ChatGPT a research grade knowledge of chemistry, the trait emerged on its own. How it was able to teach itself molecular structure we’re not fully sure, but that people will soon be asking its advice on how to make poisons from the ingredients found at local hardware stores now seems inevitable.
In a section of The ChatGPT-4 Technical Report entitled “Potential for Risky Emergent Behaviors,” it explains that, “Novel capabilities often emerge in more powerful models. Some that are particularly concerning are the ability to create and act on long-term plans, to accrue power and resources and to exhibit behavior that is increasingly ‘agentic.’ … There is evidence that existing models can identify power-seeking as an instrumentally useful strategy.”
Why this occurs isn’t yet understood. “Although there are dozens of examples of emergent abilities, there are currently few compelling reasons for why such abilities emerge,” explained SVP of Google Research and AI, Jeff Dean.
Chat-GPT 4 has now been available to the public for nearly two months with few limitations in place. In the burgeoning world of artificial intelligence, there are actually a staggering 3000% more people working to improve the abilities of these machine learning models than there are people working to ensure that they’re safe. And in their absence, the market has given way to a wild west of AI proliferation.
The number of AI text and image generators that have sprouted in the last few months alone is so enormous that it would be challenging to keep track of them all. Just open the app store if you doubt me.
Even Snapchat has begun offering its users a free AI assistant to talk to. What was originally a feature reserved only for its premium members, Snapchat realized had grown so popular throughout the rest of the digital landscape that they should simply offer it to their users for free.
The internet and the computers that drive our world developed at one tenth of the speed as these strange new tools that are being so thoughtlessly pushed on the public. The speed at which these tools have fallen into our hands, too, is dramatically faster than any of the social media platforms of the world were able to achieve.
After OpenAI released ChatGPT-3 to the public in November of 2022, it passed over its 100 million user benchmark within only three months. It took Twitter, Facebook, Instagram, Myspace, and Snapchat years to reach that same goalpost. That it managed to achieve this takeover without even offering the features of interconnectivity that defined its viral predecessors is a telling proof of how colossal the challenges ahead will be.
In a talk given on March 9th, dubbed “The AI Dillemma,” Tristan Harris and Aza Raskin, two top technology ethicists, cited some scary statistics.
50% of AI researchers believe that there’s a 10% or greater chance that humans will go extinct from our inability to control AI. Harris then made the analogy that, if asked to board a plane where there’s a 10% possibility it will crash into the ocean, few would board that plane.
But in truth, I’m not sure how well the analogy works. We’re frogs in heating pots. If we can kick the climate change can down the road for entire decades, jumping blindly into a world of artificial intelligence will hardly feel like the greatest leap our species has taken. We might not even notice the plane that we’re boarding.
We’re adaptable creatures. If rising seas and worsening climates and crumbling biodiversity doesn’t have us protesting in the streets for change, it’s hard for me to imagine a much more measured approach to the advent of artificial intelligence than the one we’ve mustered so far.
I doubt there will be a point any time soon at which we can collectively say that enough AI is enough. I fear that the conclusion is one we may only be able to reach once the reins of this beast have left our sights completely.
Did we look at the arrival of 3G and say “that’s enough internet, we can stop now. Think of the consequences!” Did we look at megabytes and say “That’s enough information, we can stop advancing?” No, we built robots that assembled microchips with such machinist feats of precision that it’s a wonder we haven’t yet found ourselves drifting meekly off into the singularity already.
With each tech company in an arms race to improve the machine learning models at their disposal, we’re in the midst of a colossal, societal change for which we lack proper comparisons. The prospect of containment at this point is a lofty vision at best.
We’re flirting with the edges of a paradigm shift for which a species composed of cells and synapses could never properly prepare. The once-fanatical visions of the machines that could build and improve upon themselves is a distant dystopia no longer. An unthinkable future is at our doorstep and it’s a misty, misty dawn. That we can navigate through the fog at all is no certainty.
What will ultimately come from the world of artificial intelligence remains one of the most consequential mysteries of our time. But it’s an error of thinking to believe that, if artificial intelligence continues improving at its current rate, that it won’t soon develop something akin to consciousness. It’s already expressing agency. Properties are emerging from these artificial intelligences that even its creators can’t explain.
The arrival of the internet revolutionized what it is to be a human being. There are few aspects to life on earth that haven’t been wholly reshaped by what it is to have the entirety of human knowledge traveling through the air around us.
But to have the wealth of all that’s known at our very fingertips has come with the heaviest of burdens. It’s changed our brain chemistry irreversibly and left us inattentive, overwhelmed and polarized. It’s forever changed our relationship with technology.
It’s optimistic to imagine that the advent of artificial intelligence won’t result in similar shifts. That we’re facing this dire crossroad before we could even find our footing after the dawn of the internet is an intimidating prospect. The next few years could bring with them cyber attacks, or biology automation, or fully realized synthetic relationships, or automated fake religions, or a total collapse of our sense of what’s real and what’s not.
The rapid spread of artificial intelligence has drawn its share of comparisons to the Manhattan Project in recent months. Even the CEO of OpenAI himself has reportedly made use of the analogy. With all of the simultaneous innovation occurring within this field, though, where we stand now is a stark contrast from that fateful meeting that nearly brought the world to its knees eight decades ago.
“Nukes don’t create stronger nukes, but AI makes stronger AI,” Raskin pointed out.
“It’s like an arms race to improve all other arms races,” Harris added.
It’s difficult for me to envision the point in this relationship at which we’ll decide that this genie can be bottled once again. If nuclear weaponry had made it into the hands of each individual, I don’t think we would have survived long enough to be here today discussing the risks and merits of artificial intelligence.
ChatGPT was released to the public in November of 2022 and has already surpassed 1.6 billion users. The cat’s out of the bag and it’s regenerating in the alley down the street beside a teetering tower of cardboard Pandora’s boxes.
I want to believe that, given time, we’ll establish proper safety measures for this burgeoning new industry. I want to believe that AI will be used as a tool for humanity, that it will aid us in medicine, space exploration and our battle against climate change. It’s already shown promise.
But I’m not so sure. This gamble simply defies comparison. With each new headline, I feel more and more like we’ve already thrown caution to the wind. I think the reins of this beast may have escaped us. I think we’re marching off into a brave new world whether we want to or not. But for now, we’re soaking in the rays of this final frozen moment in the calm before the storm that changes everything.