Was this essay forwarded to you? Join for Free Here

This past week, OpenAI seemed to pump the brakes on its newest model, Astra (Power, 2026), while Anthropic launched Sonnet 5.5 just days after it had launched Opus 5.5 (Anthropic, 2026; Ropek, 2026).

There seems to be no slowing down in the artificial intelligence race. A large part of the debate surrounds the concept of superintelligence. A few weeks ago, I wrote about a former Anthropic and OpenAI employee warning the end of the world because of the dawn of superintelligence.

Superintelligence has been loosely defined as an intelligent force that surpasses human intellect across every single domain. And as Bostrom would say, superintelligence is "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest" (Bostrom, 2014).

Multiple firms are racing towards building something of this capability. However, there is a fundamental flaw in their pursuits because imperfect, disagreeing humans can only build machines in their own image. And if the machines and models fundamentally disagree, then they can't all be correct.

If you look at the Latin prefix, "super" means above others, but there's still room above "super." I think a lot of what people are talking about is the idea of ultimate intelligence. A lot of conversations seem to stem around a machine becoming the ultimate source of knowledge. Even Trump last week said artificial intelligence should be deemed "super intelligence" (The White House, 2026), and from a nomenclature standpoint, I don't necessarily disagree.

I think the problem is when we use a term like "super intelligence," we assign a godlike wisdom to a machine. Intelligence is not wisdom. There's a massive discrepancy between those two, and the chasm between intelligence and wisdom is so great that I'm worried that it's the very pit that we will all fall in if we continue to put more faith in machines.

I put an identical prompt into Claude, ChatGPT, and Gemini. Claude and ChatGPT were relatively similar, and Gemini provided a completely different response.

I don't think it's that the LLMs even disagree on the data, but how they present the information was substantially different, enough that a human reading it could come to a different outcome or act differently as a result of reading it. It matters because right now, so many people are looking to artificial intelligence, and specifically LLMs and chatbots, as this unlocked pool of limitless knowledge and information.

Researchers at Ghent University decided to see what influence the creators of certain LLMs had on their output (Buyl et al., 2026). The results were staggering: "Our results suggest that the ideological stance of an LLM reflects the worldview of its creators." You could see the fingerprint of each one of these creators on the final product of the model.

xAI's model had a bent towards "national sovereignty, centralized authority, and economic self-reliance," the study showed. Google was "significantly more favorable on average toward political persons annotated with tags related to progressive societal values…"

Last month, Mark Zuckerberg said there is "no such thing as a singular benevolent superintelligence" (Zuckerberg, 2026).

Meta, Google, Anthropic, OpenAI, and xAI are some of the main U.S. platforms running at light speed towards superintelligence. How is it possible they all have different verdicts on what is right? These models are just magnifications of their creator.

When you have flawed humans creating technologies that can scale, it creates some real concerns. One of the questions I get asked most these days is whether or not I think AI will be able to kill us. And no, I don't feel that an AI model by itself has the capacity to do that. With human prompting, I think there are some very real risks. However, it is far more likely that human beings will turn on each other faster than AI will take out humanity.

I believe imperfect makers can build something beyond themselves. Look what we've built: airplanes. We can fly in the sky. That's beyond any human ability. Try as I might as a kid, flapping my arms as much as I could, or using bags or blankets, I could never get airborne. So yes, humans overcome things beyond ourselves all the time. That's part of the creative nature that's innate in all of us. We love progress. We love creating.

But I think the big difference here is our mindset shift. I see more and more people relying on artificial intelligence, or even superintelligence, as a savior that's going to have the right answers to solve the world's greatest problems. But at the end of the day, it's still a machine created by imperfect human beings.

It's been argued that we have enough resources to solve world hunger today, but why haven't we? If the answer was so simple, wouldn't we have already solved it?

There is no such thing as perfect here on earth. We actively work towards making things more perfect. But if we, as humans, are imperfect beings, how would it be possible to create something truly perfect?

I think back to when we were training image recognition models in my startup in 2019. There were certain state driver's licenses that would fail every time because our sample size didn't have enough of those particular states to train the model. We were biased by proximity in that case.

That bias was eventually fixed with more data. The more data we had, the more accurate we were at predicting true and authentic driver's licenses. So you could argue we solved the ability to recognize certain data, but then what do we do with that data? Well, we used it to rank and assign different people to different tasks in a gig economy platform, which is simply just using more data to make decisions.

But at some point, any one of those pieces of data could be flawed and could skew the decision ever so slightly. A model typically can't see the whole picture. It only feeds off of the last thing it was told and builds off of that.

If you're opposed to watching either CNN or Fox because of their political stances, are you considering that the AI you're trusting to give you information is more unbiased than either one of those platforms? Yes, I'll concede that the news is intentionally slanting, whereas AI might not be as overtly propagating its viewpoint. But it still has biases. And I don't just mean political biases.

It reminds me of the early days of Wikipedia when, if you dared to cite Wikipedia in an academic setting, you'd be automatically flunked for using false data. I think because of the confidence and personalization that chatbots convey in conversing with an individual, it feels like it's right, and people trust it and its outputs more than they should.

I think we're already believing AI outputs more than we're willing to admit. We're already numb to its biases because we're using it. It's creating custom, personalized answers that make us feel better about ourselves. Yet it has underlying biases, some significant, but most are minor.

Each week I come to share some thoughts on recent events or big questions as it pertains to generations, the future, and artificial intelligence. But here's the kicker: You're subscribed here to my point of view, which inevitably has biases. I hope I give you research in a digestible way so you can make some informed decisions, but my writing is just as biased as any one of these AI platforms or any news station you subscribe to, because I'm a human, and my own experiences and backgrounds influence the way I share information.

At least in this case, you know the human behind the writing and the thoughts. Do you know the methodology and biases behind your favorite LLM?