The Way That AI Works

We can begin with a brief 30,000-foot overview of how contemporary generative AI and LLMs are functionally designed. This sketch applies to many popular LLMs, including ChatGPT, GPT-5, Claude, Gemini, Grok, CoPilot, and others. Keep in mind that the sketch is merely at a high-level and makes broad depictions. For a more in-depth explanation of how modern-era AI works, see my discussion at the link here.

Consider these four key steps:

• (1) AI models are usually trained on vast corpora of existing human knowledge, typically found by scanning human writings throughout the Internet.
• (2) During this initial data training phase, the AI models algorithmically steer toward statistical regularities.
• (3) Statistical regularities tend to favor common patterns over rare ones.
• (4) Therefore, AI-generated content tends toward average or consensus views.

When I give talks about the latest AI trends, I often use a classic example about statistics to illuminate the nature of statistical regularities. Suppose we had a room equally filled with professional basketball players and with professional horse jockeys. The basketball players are all around 7 feet tall, while the horse jockeys are around 5 feet tall.

What is the average height of those in the room?

The statistical answer is that the average height is 6 feet. In one sense, yes, that’s correct. On the other hand, it doesn’t do a suitable job of conveying the true heights in the room. If someone told you that the occupants averaged 6 feet in height, and you knew nothing else about the people, you would be unlikely to suspect that they are at heights of 7 feet and 5 feet.

Using AI Often Entails Averages

When you give a prompt to generative AI, the chances are that you will conventionally get an averages-based response. You might not realize that’s happening. The answer will undoubtedly seem convincing and sensible. Nonetheless, to some extent, you are getting an averages-oriented response.

The usual default for responding to user prompts tends to lean this way:

• Conventional reasoning is displayed by the AI.
• Moderate viewpoints are computationally preferred by the AI.
• Familiar narrative structures are used by the AI.
• Common stylistic patterns appear.
• Mainstream explanations are the norm.

Those are the averages that we see daily when using generative AI and LLMs.

Big Trouble Afoot

Let’s assume that the use of AI continues to increase. Billions of people across the planet become heavily dependent on using AI for all sorts of needs. They use AI to get answers to personal questions. They use AI to get answers to work-related questions. Everything they do or decide is in some manner significantly aided by dipping into AI.

Mull over what will happen if people at scale are continually and exhaustively being fed by averages-based answers to their questions.

The worry is that society will gradually shift toward a semblance of intellectual homogenization. We will all stepwise gravitate toward the same lines of thinking. People will no longer think outside the box. They are being insidiously tilted toward the average. Unfortunately, people won’t even realize what is occurring. It isn’t readily apparent as to what is happening, since it doesn’t stand out and it is happening on a widespread basis to everyone at roughly the same time.

This is reminiscent of the tale of fish in the fishbowl. When asked about the water in the fishbowl, the fish are puzzled by what you mean, since they all take their circumstances for granted and hardly recognize they are immersed in water.

The Vicious Downward Spiral

You’ve perhaps seen blaring headlines warning us that we are doomed to be inundated by AI slop. The idea of AI slop is that AI is producing vast volumes of averages-based outputs, and those outputs are getting posted onto the Internet. At this pace, the Internet will become dominated by AI slop rather than by human writing. See my analysis of these claims at the link here.

This Internet incursion is worrisome for humans since we all make extensive use of the Internet. Thus, not only will your use of AI get averages-based responses, but when you access the Internet, the predominant information you will find there will also be averages-based.

You are utterly surrounded by averages in the content that you will be consuming.

Human minds are presumably going to become incredibly dulled. Without exposure to something other than averages, your mind will only know about and think of averages. The problem worsens since the latest AI will be trained on the Internet, therefore converging increasingly toward averages. It is a vicious cycle. Average content is produced by AI, posted to the Internet, and becomes the fodder for the next generation of AI.

Dismal, abysmal, highly disturbing.

Thinking Mindfully About The Issue

Now that we’ve gotten one side of the argument on the table, let’s go ahead and give reasoned attention to the other side. This is something that is rarely mentioned. The usual aim seems to be to get people worked up, garner views, and put people in the grip of a frenzy. I suppose it is also possible that some plainly do not know there is another side and fail to do their proper homework on the weighty matter.

Here’s a crucial consideration worthy of rapt attention:

• Generative AI can generate varied perspectives on demand, and users are not forcibly locked into the averages-based responses per se.

Allow me to give an analogy to the use of traditional mass media. A person opts to read a particular mainstream newspaper as their daily diet of news. The newspaper is likely shaped toward a predominant overarching perspective. In that manner, the person reading the newspaper is essentially getting a somewhat intellectually homogenized view.

If the person opts to only read that one newspaper, they are going to be immersed in that one view. They might not realize there are other views to be had. They are getting a single perspective on the world around them.

The person decides to read a different newspaper. Assuming that this newspaper differs from the one they are used to reading, they are now getting exposure to alternative views. They are breaking out of the mold or rut they are in.

AI Providing A Range Of Views

The beauty of contemporary AI is that, though being accused of being intellectually homogenized, the AI can depict more viewpoints than users would otherwise normally see. It can happen at the push of a button. You barely need to lift a finger to get this to happen.

Imagine this. Assume that you ask a question of AI, and in so doing, you are essentially telling the AI to do this (by its conventional default):

• User default prompt: “Give me the answer as a mainstream viewpoint.”

You don’t realize that’s what you’re telling the AI because it is the hidden standard default. But a user doesn’t have to allow the default to rule the game. You can easily instruct the AI to give a response from a different perspective.

Consider this:

• User instructive prompt: “When you give me an answer, do so with a mainstream viewpoint and also include a completely different perspective that is unlike the mainstream viewpoint.”

The AI would likely give you the averages-based response first, and then include an outlier response. You can then compare the two responses. This enables intellectual expansion.

Going back to my analogy about the room filled with basketball players and horse jockeys, the AI would essentially give you the answer about the occupants being 6 feet tall (that’s the averages-based answer) and also tell you that some are 7 feet and some are 5 feet in height. This would hopefully catch your eye and spark you into the realization that the average’s answer was an average answer.

Escaping The Bubble

I would wager that most users probably do not realize they can change up the default AI settings regarding shaping responses. They assume that whatever the AI maker has tuned the AI to do is the best or proper way of doing things. Ergo, one notable problem is that people don’t widely realize they can push AI toward providing a wider array of perspectives.

This is a prime example of why education and training about AI must be undertaken across-the-board. If people learned basic AI literacy skills, they would know that they don’t need to sit still and be fed the averages-based responses of AI. Users can fight back. I’ve covered over fifty keystone prompting skills that people can use to get more out of AI and escape from the default AI bubble; see the link here.

Another consideration is that policymakers and lawmakers should consider establishing new AI laws that ensure people are aware of their AI options. For example, when logging into AI, the AI could directly tell you that the default setting involves averages-based responses and then ask if you would like to change the default. This would force AI makers to put this option at the front and center of users’ eyeballs and allow people to decide what they prefer.

For more about the burgeoning rise of new AI laws, see my coverage at the link here.

AI As Amplifier Of Intellectual Curiosity

In the dense forest of AI as a mass bland propagator, we need to also recognize that AI can be a shining light or amplifier of intellectual curiosity. The case can be made that AI is going to improve human intellect. This stands in stark contrast to the doom-and-gloom predictions.

Here’s the deal. You can use AI to quickly and at a low cost explore unfamiliar viewpoints. Suppose someone has asked a question about societal policies that might be devised to cope with a pressing national issue. If you read only one newspaper to figure this out, you are likely cloaked in one viewpoint.

Instead, via AI, you can ask and compare perspectives from a multitude of angles. What are the economic ramifications of the policies? What does history tell us? How might those who oppose the prevailing policy tend to argue in opposition? An entire intellectually challenging and invigorating debate can be undertaken at your keyboard. Nice.

I especially relish the AI persona capability of generative AI, which allows you to have AI simulate the kinds of responses that a specific person or type of person might express; see my detailed explanation about AI personas at the link here.

My usual go-to is to ask AI what Abraham Lincoln would say. Please know that if you do use AI personas, such as Honest Abe, you are getting a computational simulation that might or might not aptly represent the nature of that person. Always be on your guard. This is yet another one of those AI literacy aspects that is sorely needed for those who make frequent use of LLMs.

History Repeats Itself

Those of you who were around during the initial days of the Internet, and the World Wide Web or Information Superhighway, will recall that many critics were clamoring and forewarning that the world would become intellectually normalized by the advent of online postings. We would, at scale, become dominated by a single viewpoint. The rise of Big Brother had a means or mechanism to finally take hold of our minds.

It was a ruinous future ahead of us.

You can certainly say a lot of justifiably disparaging things about the Internet nowadays, but I don’t think that claiming it has made us into a unified singular viewpoint is an especially viable argument. If anything, the range of inquiry is much larger than it had ever been in a paper-based world. The worries about a long-tail effect of everyone seeing the same info and passively becoming the same are no longer the crux of worries when it comes to the downsides of the Internet.

The Internet has expanded access to niche interests. You might carp that this has caused massive fragmentation and splintering. The odds are that large-scale use of AI could be headed down that same path, assuming that people realize they can tap into AI and go beyond the default setup.

The World We Are In

All told, you can assert that generative AI has the potential to provide these upsides:

• Ready access to a multitude of viewpoints that otherwise would have been arduous or costly to find.
• Makes it easy to explore niche domains.
• Can enable persistent personalized intellectual journeys.
• Lowers the traditional barriers of accessing and interpreting highly specialized knowledge.
• Serves as an incentive toward originality and differentiation if people opt to go that route when using AI.

AI is a dual-use proposition. When used well, the benefits are tremendous. When used insufficiently, the downsides are ominous. It is up to society to help guide which direction we go. Could we become AI-led mental zombies? Yes, we could allow ourselves to fall into that pit of despair. Could we instead become intellectually stronger and more capable as humans due to AI? Absolutely, though this requires our outright effort to ensure that such a happy-faced world arises.

As per the famous remarks of Abraham Lincoln: “The best way to predict the future is to create it.” We must take heed of that sage advice when it comes to AI. Get creative.