Ever had that moment when you walk back into your parents’ house and suddenly feel like a teenager again?

Well, today, we’re diving into that curious phenomenon of how our brains work, especially when it comes to our memories and perceptions.

We’re unpacking a fascinating intersection between human psychology and artificial intelligence, exploring how our brains create “maps” based on past experiences, and how these maps can shape our interactions with others.

The wild part? These same processes are mirrored in the algorithms that power AI! We’ll chat about how both humans and AI rely on filters to make sense of the world, but there's a crucial difference—humans have the ability to update their maps and break free from outdated patterns, while AI is stuck in the past.

So, grab a cuppa and settle in as we explore the magic and limitations of our minds compared to the machines we’ve built!

Takeaways:

1. In this episode, we explore the fascinating intersection of human psychology and AI, revealing how our brains process reality similarly to algorithms.
2. We discuss how our perceptions are shaped by predictive models, using the example of returning to a childhood home to illustrate outdated maps of ourselves.
3. The podcast highlights that both humans and AI use filters to process information, but AI lacks the metacognitive ability to question its own biases.
4. We delve into the implications of these cognitive shortcuts, particularly in the context of social media and corporate decision-making.
5. The discussion emphasizes the importance of self-awareness in updating our internal maps, contrasting human flexibility with AI's rigid programming.
6. Finally, we encourage listeners to experiment with their interactions, imagining engaging with others as if meeting them for the first time, shaking off historical biases.

Chapters:

1. 00:01 - The Impact of Early Relationships
2. 01:26 - Understanding Human Perception and AI
3. 06:11 - The Cognitive Filters of AI: Deletion, Distortion, and Generalization
4. 08:11 - Understanding AI Limitations and Biases
5. 10:48 - The Shift from Historical Patterns to Authentic Connection
6. 16:46 - The Power of Resetting Your Perception

Links referenced in this episode:

1. linkedin.com/newsletter/start-with-ai

I want you to try something right now.

Just take a second and picture a parent or maybe a carer, someone who showed up for you consistently when you were growing up.

Yeah.

And pay close attention to what just happened in your brain there.

Right.

A face probably appeared.

You might have heard a specific tone of voice or pictured a certain room from your childhood.

You might have even felt a sudden warmth or, you know, if we're being honest, maybe a familiar tension.

Exactly.

And the really fascinating part is that you didn't consciously choose to assemble all those specific details.

They just arrived as a package response.

Welcome to the Deep Dive, by the way, I'm your host and I'm thrilled to have our resident expert here to help break this all down.

It's great to be here.

And that arrival you mentioned, it happens in absolute milliseconds.

What's profound about that packaged response is that the person you just pictured isn't actually who they are today.

Right.

It's a version of them.

Precisely.

What you are seeing and feeling is who your nervous system has classified them as.

It's a compiled output based on years of historical data, all filtered through every pattern you've ever registered about them.

So your brain essentially ran a predictive model, and you experienced that model as reality.

And that dynamic right there is the core of today's Deep Dive.

We're pulling from this really compelling LinkedIn newsletter titled Start with AI, written by

Yes.

And our mission today is to explore this crazy intersection between human psychology and artificial intelligence.

We're looking at how our neurobiology processes reality using the exact same mechanisms that power modern AI.

And maybe more importantly, we are going to isolate the one defining capability that humans have, which algorithms just completely lack.

It's such a critical piece of writing because it forces a mirror up to our own internal programming.

I mean, before we can effectively analyze how we train our machines, we kind of have to understand how we train ourselves.

Right.

The architecture of human perception actually provides the blueprint for the artificial systems we are currently deploying at scale.

Okay, let's unpack this.

Yeah.

Because the text offers a highly relatable scenario to illustrate this architectural mirroring.

Consider the experience of walking back into your parents house.

Oh, this is a perfect example.

Right.

You are a fully independent adult.

You're running your own life, paying taxes, all of that.

Yet within minutes of crossing that threshold, you find yourself being treated like you're 17 years old again.

And worse, you catch yourself reacting like a 17 year old.

Exactly.

The author notes that this Isn't just some quirky regressive family habit.

It's what we call a map.

Yeah.

Their internal map of you was largely finalized decades ago.

For a multitude of reasons.

That specific neural pathway just hasn't required a major update since.

Sense.

And the inverse is also true, isn't it?

Absolutely.

The version of them that you brought through the front door.

That's an outdated map too.

You are interacting with a historical data set of your parents while they are interacting with a historical data set of you.

So the actual present day territory is entirely obscured by these predictive maps.

Yes.

And the newsletter roots this map making process in Neuro linguistic programming or NLP, which has been studying this phenomenon for 50 years now.

To really understand how we construct these maps and eventually how AI does the exact same thing, we have to look at the three primary filters our brains deploy to manage information.

Because the sensory data of the world is just too vast to process in real time, our nervous system has to use these filters for pure efficiency.

But that same efficiency is exactly what locks us into outdated patterns.

So the first of these filters is deletion.

Deletion is a raw survival mechanism.

You simply cannot encode everything you experience.

Your brain acts as this ruthless editor, selecting what to retain and what to discard based on existing relevance.

The author uses a really striking example for this.

If a carer was there for you 200 times but let you down three times, which data points do you instantly retrieve when you picture them?

Right.

Whether you retrieve the care or the betrayal tells you everything you need to know about your active filtering system.

The memory you hold isn't an objective history.

It's a heavily pruned dataset where contradictory evidence has been completely deleted.

To maintain a cohesive narrative, you curate

the memory to protect the model that is so wild.

Which brings us to the second distortion.

Yes.

Distortion takes the data that actually survives the deletion process and actively reshapes it to fit the patterns you already expect to see.

Think about predictive coding in the brain.

Like if you receive an ambiguous text from a colleague, you might immediately interpret it as passive aggressive.

Or a brief silence on a conference call instantly reads as disapproval.

Exactly.

The hostility wasn't in the text and it wasn't in the silence.

Your brain distorted neutral data into negative data because your internal model was already primed for conflict.

The meaning basically arrives preloaded.

If your map dictates that a specific environment is adversarial, your perceptual system will literally warp incoming signals to confirm that hypothesis.

Which naturally leads to the third filter.

Right?

Generalization.

Correct.

Generalization is the mechanism that takes a single isolated experience and extrapolates it into a universal law.

Like one failure becomes I am always

terrible at this, or one breach of trust becomes nobody can ever be relied upon.

These generalized rules run quietly in the background.

They categorize every new input before it ever even reaches your conscious awareness.

So every single person you interact with today is running this three part model.

The reality you navigate is just the output of your own processing.

It's a highly efficient representation, a necessary one even.

But it is never the territory itself.

Here's where it gets really interesting, because for the first time in human history, we have outsourced these three filters to external systems.

We have the newsletter draws a direct mechanical line between these human cognitive shortcuts and the architecture of artificial intelligence.

We haven't just built calculators, we have built automated versions of our own cognitive biases.

AI systems delete, distort, and generalize using the exact same frameworks.

Let's actually look at the technical translation of those filters, starting with AI deletion.

Okay, let's hear it.

The training data fed into any neural network is not the totality of human knowledge.

It's not objective reality.

It is a highly curated, pruned selection of human output.

Massive swathes of potential context, nuance, and historical perspective are literally deleted from the data set before the training phase even begins.

So the model's foundation is built on absence just as much as presence.

Exactly.

Then we have AI distortion, which the tech industry commonly refers to as hallucination.

Right.

When an AI model lacks the precise data to answer a query, it doesn't just stop and say, I don't know.

No.

It navigates its vector space to find the nearest plausible mathematical fit.

It effectively bridges the gap by distorting the data.

It does have to synthesize a confident, yet entirely fabricated response.

It reshapes reality to complete the pattern, just like a human interpreting an ambiguous text message.

And finally, generalization.

A large language model is fundamentally a generalization engine operating at an unprecedented scale

through token prediction weights.

It analyzes massive data sets to establish universal probabilities about which words or concepts belong together.

Right.

It looks at a billion isolated text strings and creates hard and fast rules to predict the next output.

It is pattern recognition, enforcing a synthesized norm across all future interactions.

We have literally engineered machines in our own psychological image.

What's fascinating here is the singular defining difference between the two systems.

While both humans and AI rely on these filters to process reality, an AI entirely lacks a metacognitive layer.

Metacognition being the ability to think about your own thinking?

Exactly.

An AI possesses no mechanism to step outside of its own parameter weights and observe its processing.

It cannot ask itself if the output is a result of a biased milter or an accurate reflection of the territory

to ground that in a practical reality.

The newsletter highlights the well documented failure of Amazon's AI hiring tools, and we are sharing this purely to illustrate the mechanics of the technology as presented in the source text.

The premise was highly logical.

Train an algorithm on a decade of successful hire data to identify future top performers.

But because of the historical demographics of the tech industry, the training data was overwhelmingly male.

The initial data set had already run the deletion filter.

Women were statistically absent from the model of what constituted a successful hire.

Right?

So the algorithm did exactly what it was designed to do.

It generalized.

It learned that male associated traits were

predictors of success and began actively penalizing CVs that included the word women's, such as an applicant noting they captained the women's rowing team.

The critical takeaway here isn't just about corporate sexism.

It is about the mechanics of the failure.

There was no malicious coding involved, but

because the AI lacks that metacognitive layer, there was no pause.

There was no internal mechanism that could flag the pattern and say, hey, this.

This parameter seems to be mirroring historical bias rather than optimizing for talent.

It simply ran the generalized rule blindly.

And if we connect this to the bigger picture, the implications stretch far beyond automated HR screening.

Yeah, the author expands on this to argue that most of our socioeconomic systems are actively being run by unexamined maps.

Entire corporations routinely miss massive market shifts because contradictory market signals are deleted by the organizational filter before they ever reach the executive suite.

Boardrooms are operating on a map of the industry from five years ago, distorting

new competitor data to fit their legacy models.

And you see that same mechanic weaponized in social media Algorithm.

These platforms are applying deletion, distortion and generalization to our digital lives at a staggering scale.

Consider how your feed operates.

The algorithm deletes any content that falls outside your established engagement history.

It distorts your perception of consensus by mathematically amplifying content that provokes high arousal visceral reactions within its vector space.

And it generalizes your past clicks to lock you into a predictive loop, quietly shaping your future behavior.

We are no longer just navigating our own internal biases.

We have outsourced our mapmaking to opaque algorithms that run without our Metacognitive oversight.

The map has essentially replaced the territory.

For millions of users.

The feed is accepted as the absolute reality, and the architectural filtering that constructed it remains completely invisible to the consumer.

I want to pivot to the second half of the source material now, because the author, Heather, grounds all of this heavy theory in a deeply personal somatic experience.

She shares a story about her father.

It's a very powerful story.

She describes him as a formidable man with a dominant presence.

Consequently, their relationship for most of her adult life was defined by resistance.

They were, in the context of our discussion, two very rigid maps occupying the

same room, each entirely convinced that the other's map was the fundamental problem.

Right?

And that dynamic is the defining characteristic of a prolonged interpersonal gridlock.

Both individuals are operating almost entirely through their distortion and generalization filters, reacting to

historical predictive models of each other rather than the present moment.

Well, her father was eventually hospitalized, and Heather experienced a strong intuition that he wouldn't recover this time.

On her drive to the hospital, she was stopped at a red light.

And in that forced physical pause, she made a deliberate conscious choice.

She decided to completely drop her map.

She chose to let him be exactly who he was without the historical weight of how she thought he should be.

She suspended the predictive coding.

She stopped requiring the territory of her father to match the idealized map she had carried for decades.

She specifically notes that she didn't sit in her car rehearsing a monologue or planning some big confrontation.

She verily shifted her internal state.

She loved him completely without the conditions of her generalizations.

And the outcome of that silent internal shift is arguably the most important data point in the entire newsletter.

When she walked into his hospital room, before she had even crossed the floor or spoken a single word, her father looked at her and immediately told her he needed her to know how much he loved her.

He articulated a vulnerability that simply hadn't been present in their adult relationship.

Heather hadn't altered his single external variable.

She didn't engineer a debate or attempt to persuade him.

She utilized her metacognitive layer to recognize her own filter, and she chose to release it.

And shifting her internal map instantly altered the relational dynamic.

It gave him the space to step outside of his own historical map.

AI cannot do this.

An algorithm cannot experience a red light moment.

It cannot register the somatic weight of a useless historical pattern and consciously decide deeply within the body to discard it.

It cannot update its parameter weights out of intuition or a sudden capacity for grace.

That profound agility is strictly human territory.

It is a capacity available to you, the listener, in any moment of stillness you choose to capture.

That is the definitive human advantage in an increasingly automated landscape.

We possess the unique architecture required to rewrite the foundational code while the program is actively running.

The newsletter brings this realization directly into the professional sphere, too, focusing on practitioners, coaches, therapists, leaders, people who do what the author calls the work underneath the work.

Because when you are managing an employee or counseling a client who is fundamentally stuck, the presenting complaint is rarely the real structural issue.

The real work requires digging beneath the surface to identify the outdated filter that is still executing on ancient data.

You are searching for the deletion that solidified into an identity.

For instance, a professional who routinely deletes evidence of their own competence, operating under the generalized rule that they are an imposter.

Or you are looking for the distortion that has become entirely self, confirming the employee who interprets neutral feedback as a personal attack, effectively guaranteeing the hostile environment they expect.

A highly advanced AI could easily transcribe a coaching session and flawlessly highlight the linguistic markers of those NLP filters.

It could point out every single time a client says always or never.

But identifying the filter is only 10% of the process.

The actual work is creating the precise psychological conditions that allow the client to notice the filter themselves.

It's about deploying the right question, pacing the silence, and holding the emotional space necessary for a human being to bear the friction of a paradigm shift.

AI cannot hold space.

That distinction is exactly why the author highlights Sue Knight's ongoing work in this field.

Knight has spent decades mapping these specific NLP territories.

and Heather is actually co hosting a session there dedicated specifically to AI integration.

The curriculum focuses heavily on developing this sophisticated human capability, training practitioners to recognize when they or their clients are trapped

inside a predictive filter, and mastering the somatic and linguistic tools required to actually break out of it.

So what does this all mean?

If we synthesize the technical architecture of AI with the neurobiology of human perception, the newsletter leaves us with a rather stark reality.

Yes, massive algorithmic systems currently hold highly detailed predictive models of human you constructed entirely from your historical data and past behaviors.

But so do your closest friends, your colleagues, and your family.

Both the neural networks we engineer and the social networks we inhabit rely on predictive models of who we used to be.

This raises a really important question that the author poses at the end of the who gets to update the map?

Right?

An artificial intelligence is permanently bound by the constraints of its training data and its vector space.

But you have the metacognitive capability to pause.

You can actively observe your own filtering systems, suspend your historical generalizations, and choose to engage with the actual territory right in front of you.

To build on that, I want to leave you with a final thought experiment to carry through the rest of your week.

I love this idea.

We know that the algorithms running on your devices and the people operating in your life are continuously running predictions based on your past actions.

But what would happen if you unilaterally decided to wipe your own training data clean for just 24 hours?

A complete reset?

What if, starting tomorrow, you interacted with every colleague, your partner, or your oldest friend as if you were encountering them for the very first time?

If you completely dropped your historical maps, your internalized generalizations, and your predictions of

how they will behave, the changes could be profound.

How might their responses unexpectedly transform simply because you gave them the space to be who they are today, rather than who your filters expect them to be?

That is the superpower you have.

Use it.