Human intelligence — is it more hype than help? | Opinion

Everyone is talking about intelligent agents these days, but I have to admit I'm not totally convinced. The whole world is raving about how important and unique these things are, and I'm just not sold.

I'm talking, of course, about "H.I.", which stands for "Human Intelligence."

Sure, humans have scaled brains, but there's nothing fundamentally new about the hardware. It's just bigger and more complicated.

The hype, alone, isn't such a big deal. However, folks are acting on faulty beliefs, and it's causing real harm. Humans keep getting hired for critical tasks, such as driving vehicles, despite mounting evidence that more reliable alternatives exist. Courts continue accepting human testimony even though research shows eyewitness accuracy hovers around 40%.

So, is human intelligence all it's cracked up to be? Let's take a look at a few of the most common issues with humans that might make you think twice the next time you're trying to decide between using a human or using a trusted piece of machinery.

The hallucination problem

The mainstream media hasn't covered this as much in recent decades, but if you lived through the '60s or the '90s, then you know all too well that humans can hallucinate.

What is a hallucination? Well, the human brain is complicated. It's really a black box in that even our best science still cannot decipher what's going on in there with 100% accuracy. And the most minor things can make a human hallucinate, including tiny molecules that are sometimes found in moldy bread or naturally growing mushrooms and even some kinds of weeds.

Although some humans hallucinate due to intrinsic hardware issues, any human can hallucinate, if given the wrong inputs. Even without special inputs, humans routinely generate false memories, invent details to fill gaps in their knowledge, and genuinely believe things that never happened.

Psychologists call such phenomena "confabulation," but that's really just a fancy word for "unreliable."

Confidently wrong at scale

More troubling, humans are often most confident when they're most wrong. Studies show that human certainty has essentially zero correlation with human accuracy. A human can be absolutely certain they locked the door, turned off the stove, or sent that email — and be completely mistaken.

This wouldn't be such a problem if humans acknowledged their unreliability, but they don't. They often double down or get defensive. At worst, humans can knowingly deceive others in a process called "lying."

In legal settings, that witness who swears they saw the defendant? Research suggests they're wrong nearly 60% of the time in high-stress situations. But we've built an entire justice system around trusting them anyway.

Severe context window limitations

Humans can only hold between four and seven pieces of information in their working memory at once. Try explaining anything moderately complex to a human, and they'll nod along for the first few points, then their eyes glaze over. You can actually see the hardware failing.

This is a real problem in business contexts — email messages must be extremely short, ideally conveying only a single point or ask. Try adding more than one query or piece of information, and often the human will respond only to the first one.

Humans need information repeated, contextualized, repeated again, and then someone needs to send a summary email.

Mandatory maintenance cycles

Humans require approximately eight hours of daily downtime for maintenance. It's a process that experts call "sleep," but let's be honest about what it is: forced shutdown for essential repairs and garbage collection. Skip this maintenance cycle even once, and human performance degrades measurably. Skip it repeatedly, and humans start hallucinating even without funky inputs.

Humans also need regular refueling every few hours (i.e. "meals") and other tedious little updates throughout the day: coffee, snacks, bathroom breaks, and on and on. The average human spends about 16 hours per week just on basic maintenance and fueling operations.

Critical security vulnerabilities

As sophisticated as some claim them to be, humans are remarkably susceptible to social engineering attacks. A sufficiently motivated bad actor can manipulate most humans with embarrassingly simple techniques: phishing emails, phone scams, sob stories, appeals to authority, manufactured urgency.

The human bias toward trusting other humans creates systemic vulnerabilities. Our economy loses billions of dollars annually to fraud that exploits these weaknesses — cons that wouldn't work on even a moderately sophisticated spam filter. Yet humans fall for this stuff constantly.

Worse, humans exhibit confirmation bias, anchoring bias, availability bias, and about 180 other documented cognitive vulnerabilities. These aren't bugs we can patch. They're fundamental features of ancient architecture.

The alignment problem

Perhaps most troubling is humanity's chronic misalignment between stated values and actual behavior. Humans say they prioritize family, then work 80-hour weeks. They claim to care about the environment while driving alone to the gym. They express concern about privacy, then post their entire lives on social media.

This isn't hypocrisy, exactly — it's a deeper design flaw. Human reward systems frequently incentivize behaviors that undermine human goals. Humans optimize for likes instead of connection, test scores instead of learning, promotions instead of meaningful work.

But seriously

I'm not actually arguing we should replace humans with machines for everything. I'm pointing out that the same criticisms we level at AI apply to the squishy meat bags we're already trusting with much more important tasks.

We're all biased by training data we call "upbringing." Our brains confabulate, and our memories are severely limited. We require constant maintenance, we're vulnerable to manipulation, and we pursue goals in ways that undermine our stated values.

The question isn't whether AI is perfect. The question is can AI reliably perform better than a human at a given task.

Because here's what nobody wants to say out loud: the biggest risk with AI might not be that it's too different from human intelligence. It might be that it's too similar.

Gabriel Cassady owns an AI consulting business in Springfield and writes regularly about AI and technology. Learn more about Gabriel and his AI work at GabrielCassady.com .

This article originally appeared on Springfield News-Leader: Human intelligence — is it more hype than help? | Opinion