Feeling a bit lost when it comes to AI?

You’re definitely not alone!

In this episode, we dive into the common struggle of navigating new tech, especially AI, and how it can leave us feeling a bit like a deer in headlights.

We explore a brilliant piece by Heather, a seasoned tech veteran, who breaks down why learning AI often feels overwhelming and how we can shift our perspective to embrace that learning curve as part of the journey. Instead of treating it as a hurdle, we’ll chat about how this curve is actually the work we should be doing.

So, if you’ve ever felt like you missed the memo in the tech world, stick around as we help you bridge the gap from pushing buttons to genuinely understanding the tools in your hands!

This Deep Dive podcast is AI generated from the Start With AI Newsletter on LinkedIn - linkedin.com/newsletter/start-with-ai

Detail

Unpacking the complex feelings of inadequacy that often accompany learning new technologies, this episode takes listeners on a journey through the challenges of engaging with AI. The speakers kick things off by sharing their own relatable experiences of staring helplessly at screens, feeling as if everyone else has already mastered the technology while they are left floundering.

This common sensation of being overwhelmed is particularly pronounced with AI, which frequently feels like a foreign beast. They introduce Heather's perspective, a seasoned expert in the tech world, who believes that the key to conquering these feelings lies in reframing our approach to learning AI. Instead of viewing the learning curve as a barrier, Heather suggests that it is, in fact, integral to the work we must do.

The conversation delves into how traditional technology has conditioned us to think of ourselves as flawed when we encounter difficulties, contrasting this with the fluid and unpredictable nature of AI.

The hosts discuss the importance of ‘weaponizing ignorance,’ a concept introduced by Heather, which encourages learners to retain a beginner's mindset. This approach not only helps in mastering AI tools but also fosters a culture of documentation and reflection on one's learning process, preventing the isolation that often comes with tech challenges.

As they wrap up, the speakers present the TOTE (Test, Operate, Test, Exit) loop as a practical method for engaging with AI. This feedback loop emphasizes experimentation and adaptation over rote memorization, which is increasingly ineffective in a world where AI is constantly evolving. They challenge listeners to embrace the learning process and view mistakes not as failures but as essential steps in their growth.

The episode concludes with a powerful reminder that learning to direct AI is akin to conducting a jazz band, where flexibility and responsiveness are key, marking a significant shift in how we approach technology in our lives.

Chapters:

• 00:01 - Navigating Technology Challenges
• 03:59 - Understanding Unconscious Competence in Learning AI
• 06:54 - Transitioning from Traditional to AI Systems
• 09:28 - Navigating the Transition from Traditional Tools to AI
• 13:20 - Understanding the TOTE Loop in AI
• 19:13 - The Impact of AI on Human Resilience

Takeaways:

• Feeling stuck while learning AI is super common, so don’t sweat it too much!
• Traditional tech tends to make us feel incompetent, while AI requires a mindset shift.
• Navigating AI tools is like conducting a jazz band, it’s all about collaboration!
• The learning curve in AI isn't just a hurdle, it’s part of the job itself!
• When engaging with AI, expect some trial and error, it’s all part of the process!
• Understanding the mechanics of AI tools prevents you from falling into blind usage traps!

Companies mentioned in this episode:

• Start with AI
• descript
• Claude

You know that.

That really quiet sinking sensation in your stomach when you're just staring at a screen?

Oh, yeah.

Trying to figure out some new piece of technology and just completely hitting a wall.

Exactly.

I mean, maybe you're watching a tutorial or reading some documentation.

It just seems like, you know, the rest of the world is breezing right past you.

Right.

They all got the memo and you didn't.

Yeah.

You sit there completely stuck on, like, step two, just wondering if you somehow missed a secret onboarding meeting for the modern world.

It's an incredibly isolating feeling, honestly.

It really is.

You just feel, well, incompetent.

And if you felt that recently, especially around artificial intelligence, you are definitely not alone.

Not at all.

It is arguably the most universal experience in the professional world right now.

Yeah.

And I think the friction really comes from the fact that in almost every other area of your life, you know you're highly capable.

Right.

You're doing complex things every day.

Exactly.

You solve complex problems, you manage teams, you navigate difficult situations.

But you put a slightly altered interface or a blank prompt box in front of someone.

Oh, man, forget it.

Right.

Suddenly that baseline confidence just totally evaporates.

Okay, let's unpack this because that is exactly the mission of our deep drive.

Today we are looking at this really fascinating document titled Tech Takes Time.

And that's the good news.

It's such a great piece.

It really is.

It's written by a woman named Heather, who is a 35 year veteran of the tech industry.

I mean, she's been a developer, a technical author, a trainer, and she's actually the founder of Start with AI.

Yeah.

She has an incredible amount of experience.

She does.

And she put together this incredibly insightful breakdown of why learning AI feels, well, uniquely overwhelming.

And more importantly, how we can systematically stop feeling stupid when we sit down in front of it.

Which is what everyone wants to know.

Right.

And her core premise is that learning AI isn't just a matter of, you know, climbing a steeper learning curve.

No, it's much bigger than that.

Yeah.

She argues it actually requires a fundamental perspective shift in how we relate to software altogether.

And that perspective shift is literally the dividing line between thriving right now and just burning out.

Yeah.

Our goal today is to help you, the listener, bridge that gap between merely pushing buttons.

Because someone told you to.

Which we all do.

We all do it.

But bridging the gap between that and genuinely understanding the architecture of the tool you hold in your hands.

Because right now people are just trying to use their old mental models.

Right?

Exactly.

And we are seeing it backfire on a massive scale.

Okay, so before we dive into the mechanics of AI and why it feels like this totally alien landscape, we actually need to look at the baseline first.

Right, the traditional tech.

Yeah.

Like, why does traditional technology, that regular software we've used for decades, why does that already make really smart people feel incompetent?

That is the perfect place to start.

Either shares this story in the source material that just perfectly isolates this.

She talks about being a founding user of a piece of software called descript,.

Which is a video and audio editing tool.

Right, Ed.

And she knows it inside and out.

But then the company rolls out this massive interface upgrade and they host a webinar to walk their power users through all the changes.

Sure.

Pretty standard.

So she watches the webinar, sits down to use the tool, and just completely fails to get it to work.

And keep in mind, we are Talking about a 35 year tech veteran here, not someone who's opening a laptop for the first time.

Exactly.

So she's frustrated.

She goes back, watches the replay, fails again.

Oh, wow.

Yeah.

So she watches it a third time, like hyper focused.

And she finally realizes the issue wasn't her comprehension at all.

The presenter was simply failing to state the transition keystrokes out loud.

His hands were moving faster than his mouth.

Exactly.

What's fascinating here is the psychological mechanism at play, which is known as unconscious competence.

The presenter had used his own software so, so deeply and built such dense neural pathways around it that he had effectively forgotten what he knew.

That happens all the time.

His brain just chunked the complex sequence of actions into one single thought.

Right.

He tells the audience to execute step A, and his fingers just automatically perform steps B and C in the background while he moves on to talk about.

Step D. It makes me think of those cooking shows, you know, where a master chef is throwing ingredients into a bowl and they casually say, now just fold in the cheese.

Yes.

Oh my gosh, yes.

And you are standing in your kitchen freezing up, thinking, what does folding mean?

Mechanically, like, am I stirring it?

Am I flipping it?

Am I folding it like a piece of paper?

Right.

Because to the chef, folding is an atomic unit of action.

It's one thing, but to a learner, it is a glaring absence of instructions.

That captures the dynamic perfectly.

And we really have to look at the underlying architecture of traditional software to understand why that missing instruction is so devastating.

Okay, break that down for me.

Well, traditional technology is entirely deterministic.

It is built on rigid binary if, then statements.

If I do this, then the computer does that.

Exactly.

If you do not execute the exact sequence of required inputs, the system returns nothing.

It does not guess your intent?

No, it's very literal.

So when an unconsciously competent trainer skips one tiny keystroke, they inadvertently create a trapdoor in a deterministic system that beginner steps right on it and falls through.

Man, a trapdoor.

That is such a good way to put it.

And the really tragic part that the text highlights is what happens next in the beginner's mind.

The internal reaction.

Yeah.

When they fall through that trapdoor, they don't blame the deterministic architecture of the software.

And they rarely blame the trainer either.

No, of course not.

They just freeze, stare at a screen that refuses to cooperate and quietly internalize the failure.

They just decide, well, I am simply not a tech person.

Because we have been conditioned over 40 years of personal computing to believe that the machine is always right.

If the machine does not work, the human must be flawed.

But Heather, because she writes technical training for a living, fights this by intentionally weaponizing her own ignorance.

I love that phrase, weaponizing her ignorance.

It's brilliant.

She forces herself to work with systems she does not fully understand yet, maintaining a really rigid beginner's mindset.

So she doesn't go on autopilot.

Exactly.

Every time she struggles, she documents the missing keystrokes.

She maps the trapdoors so the next person doesn't fall in.

Okay, so if traditional tech is a trapdoor of missing keystrokes because it is so wildly rigid, AI feels like the floor is just made of sand.

Yes.

It shifts under your feet.

Right.

So how do you even learn a tool that literally changes its own interface?

Well, this is the crux of the issue.

We are moving from deterministic systems to probabilistic systems.

Probabilistic?

Yes.

AI does not run on rigid if then statements.

Large language models, at their core, are prediction engines.

Okay.

They are mathematically calculating the probability of the next most likely word or token based on the vast context of your prompt and their training data.

Right, so it's guessing in a highly advanced way.

Exactly.

And because of this probabilistic nature and parameters like temperature, which introduce randomness to keep the output creative, the system is fundamentally fluid.

Meaning, you know, you could type the exact same prompt on Tuesday and Wednesday and get two completely different outputs.

Yes.

The underlying model might have even been updated overnight without you knowing.

Correct.

For decades the path to mastering software was rote memorization.

Go to file, then edit, then preferences.

Right.

Memorizing the steps.

But AI dismantles that model entirely.

Heather argues that you no longer drive software.

You direct it.

Wait, if the veteran says traditional tech is hard because it needs exact steps, but AI completely lacks exact steps, aren't we just trading one kind of trapdoor for a totally invisible one?

That's a really common fear, actually.

I mean, how is the listener supposed to grasp something that changes every time they touch it?

And that is exactly why the shift is required.

You can't rely on rote memorization anymore.

You have to fundamentally change your relationship to the tool.

I want to make sure I am grasping the practical difference here.

Driving traditional software is kind of like playing a piano.

Okay, I like that.

You press the middle C key, the hammer hits the string, you get a middle C note.

It is mechanical and perfectly repeatable.

Yes, every single time.

But directing AI feels more like conducting a jazz band.

Oh, that's good, right?

You can give them the tempo, the key, the sheet music, but the solo the saxophonist plays on Tuesday will not be identical to the one they play on Wednesday.

You aren't pressing keys.

You.

Your guiding intent.

That is an excellent way to conceptualize the shift.

When you direct a jazz band, you have to actively listen to what they are playing in the moment and adjust your direction accordingly.

You have to respond to them.

Exactly.

You can't just wave your baton in a pre memorized pattern and expect a masterpiece.

Yeah, that would be a disaster.

But this transition from piano to jazz band is deeply unsettling for people who have spent their whole careers being rewarded for memorizing the piano key.

Oh, yeah.

They feel like they lost their superpower.

They feel that shifting ground, they get overwhelmed and they start desperately looking for a way to turn the jazz band back into a piano.

Which brings us to a part of the source material that genuinely made me wince.

As someone who follows data security pretty.

Closely, I know exactly what you're going to say.

Because people want the rigid, predictable steps back, they start looking for shortcuts.

And Heather tells a story about a client who was utilizing ready made AI skills.

Yes, the instruction files, Right?

These are essentially prepackaged files or autonomous agents that you just drop into your workspace to make the AI perform specific tasks automatically.

Basically trying to add buttons back to the software.

Exactly.

The client was using them heavily, but admitted they had absolutely no idea how they work.

Not at all.

They couldn't read the files, they couldn't modify the Instructions, and they certainly couldn't evaluate if they were secure.

They just hit download and let them run.

And this happens constantly because we are carrying over our habits from the deterministic era.

We are so used to downloading an app update, trusting the developer and letting it run in the background.

But AI is different.

Very different.

We need to define what these new AI tools actually are.

When we talk about an AI agent or a tool like Claude code, we aren't just talking about a passive chatbot sitting in a browser window.

No, we are talking about granting an LLM actual agency.

Exactly the issue.

An agent is a probabilistic model that has been given tools.

It has been given access to APIs, to your file system, to your terminal, or, you know, to your company's CRM, which is huge.

It is.

It uses reasoning loops to decide on its own when to pull data, when to rewrite a file, or when to send an email.

Here's where it gets really interesting and honestly, highly problematic.

If you download a ready made AI skill from a third party and plug it into your system without understanding it, you are handing over the keys to your digital life to a black box.

You have no idea what it's doing.

You don't know the prompt instructions inside that file.

What if the instructions are poorly written and the agent hallucinates a command to mass delete a database?

It happens.

Or worse, what if it's a malicious file designed to quietly exfiltrate your contact list through a prompt injection?

If we connect this to the bigger picture, doing this with AI today is functionally identical to the early days of the Internet.

Oh, the Wild west days.

Right?

When people would just download random executable files from total strangers and run them on their desktop without a second thought,.

Just clicking on anything.

Exactly.

Modern IT departments would shut that down instantly today.

Yet because these AI tools are marketed as productivity shortcuts, professionals are just bypassing all their critical thinking.

They are so desperate to avoid the learning curve of directing the AI, they are just installing blind automation.

They just want the buttons back.

But because they skipped that fundamental perspective shift we talked about, they surrender their judgment entirely.

You cannot safely operate a probabilistic tool with execution privileges if you treat it like a predictable traditional app.

And that reality leaves a lot of people feeling trapped.

Right.

Rote memorization fails because the interface constantly changes.

Blind shortcuts fail because they are massive functional and security risks.

So you're stuck.

Exactly.

Yeah.

So the logical question is, what is the actual mechanism for learning this effectively?

That is the million Dollar question.

That is what I was hoping we could get to.

How does the listener sit down tomorrow and actually build this skill without burning out?

Well, Heather introduces a really powerful reframe here.

The common mindset is that the learning curve is the tax you pay before you can do real work.

Like you learned Excel's interface so you can finally get to the accounting.

Right.

But in the world of AI, you have to invert that completely.

The learning curve isn't the obstacle keeping you from the work.

The learning curve is the work.

Okay, I want to dig into how that actually functions in practice.

Because the text brings up a concept called the 2 OTE loop.

Yes.

TOTE.

It originates from NLP.

Wait, I need to pause you there.

When we talk about AI, NLP almost universally stands for natural language processing.

Right.

Like the actual technology behind parsing human text.

Oh, right.

But in this context, the source material is referencing psychology, isn't it?

Good catch.

Yes.

In this specific instance, Heather is referring to neuro linguistic programming.

Okay, got it.

It's a psychological approach to communication and personal development.

And the concept she pulls from it is TO I.

It stands for test, operate, test, exit.

Okay, so walk me through what a tote E loop actually looks like when I am sitting at my keyboard.

Sure.

It is a continuous feedback cycle.

You try an approach.

Say you write a specific prompt or give an agent a certain set of instructions.

Right.

That is your first test.

Then you operate.

Meaning you let the AI generate the output or take the action.

Okay.

Then you test again.

You evaluate the result.

Did it hallucinate?

Did it nail the tone?

Or was it completely off base?

Usually the first time, yeah, it's a bit off.

Right.

And if it failed, you adjust your parameters and loop back to the beginning.

You only exit the loop when the output finally matches your goal.

I am going to push back on this a little bit.

Because looking at that cycle, isn't tote just a fancy academic term for basic trial and error?

It sounds like it, doesn't it?

Yeah.

I mean, how is that a massive perspective shift?

We use trial and error all the time.

The distinction lies in who is being trained.

In traditional deterministic software, if you are relying on trial and error, it means you have failed to learn the system.

Ah, because you don't know the right buttons.

Exactly.

You are just clicking menus blindly until something works.

The software isn't changing.

You're just stumbling around the piano hoping to hit a C chord.

Right.

But in probabilistic AI, the toady loop is the active process of steering.

Because the model interprets language, context and intent.

Your adjustments are actually refining the model's current operational state.

Oh, that makes sense.

You aren't failing to find the right button.

You are actively collaborating with the system to narrow down the probability of a correct answer.

Ah, I see.

Trial and error in old tech is a sign of incompetence.

But the to loop in AI is the actual definition of competence.

Perfectly said.

You are not doing it wrong if you have to ask the AI four times to fix a block of code.

That iterative, messy back and forth negotiation is exactly what directing looks like.

And this is where the ground moving beneath your feet actually becomes manageable.

If you expect a prompt to work perfectly the first time, every time a model update will infuriate you, you're going.

To be pulling your hair out.

Exactly.

But if you sit down expecting to enter a toki loop, if you accept that unpredictability is just the baseline material you are working with, the emotional weight vanishes.

This is part of the process.

A bad output stops being a verdict on your intelligence or a sign that you are falling behind.

It is just the first test in the loop.

It takes the ego completely out of it.

Yeah, and you can see that humility in Heather's own approach.

I mean, after 35 years of building software and training others the text notes, she is actively going back through her own AI training curriculum as a student.

It is a remarkable admission for an industry veteran to make.

It really is.

She mentions that her first pass at learning these tools was just survival mode.

She was in live workshops, keeping pace with the room, just figuring out the rote steps to get a result, just.

Trying to get by.

Yeah, but now she is doing what she calls a deeper pass.

She isn't satisfied with just getting the output.

She wants to understand the underlying mechanics so she can adapt when the interface inevitably changes tomorrow.

Because rote memorization can get you through an afternoon workshop.

But deep comprehension is what allows you to survive a major industry shift.

Absolutely.

This is the entire philosophy behind the training she builds.

The goal isn't to hand you a list of a hundred prompt templates to.

Memorize, which wouldn't work for long anyway.

No, they break constantly.

The goal is to build your capacity to evaluate the machine underneath so that when a vendor hands you a ready made skill, you actually have the technical literacy to read the file, understand what APIs it is calling, and decide if it is safe to execute.

So what does this all mean for you?

Listening right now, when you open your.

Laptop tomorrow morning, it's the Big takeaway.

Yeah.

If we synthesize everything we have covered.

Traditional tech demanded precision, which conditioned us to blame ourselves for missing hidden keystrokes.

AI Demands flexibility, which is deeply uncomfortable because the parameters constantly shift.

And reaching for blind shortcuts to bypass that discomfort is just a fast track to compromising your own work and your security.

It's a huge risk.

So the struggle you feel when an agent breaks or a prompt goes off the rails, it isn't a failure of your intellect.

It is just the Toti loop doing its job.

Heather leaves the reader with a very pointed question in the text, and I think it is the perfect lens for examining our own frustration.

Let's hear it.

She asks.

Where have you been?

Treating a learning curve as evidence that you failed when it was only ever the work asking for a bit more time.

Wow.

Evidence that you failed.

We are so incredibly quick to use a moment of confusion as a weapon against ourselves.

We really are.

But looking at the bigger picture of this transition.

The death of exact steps, the reliance on the toady loop, the need for continuous, patient adjustment.

It leaves me with one final thought that goes a bit beyond the software itself.

Oh, yeah.

If AI Is fundamentally forcing us to abandon the comfort of rigid, perfectly repeatable steps, could this daily practice of directing an unpredictable tool actually be rewiring how we operate as humans?

How do you mean?

Well, think about the skills required to successfully direct an AI Right now.

You have to clearly communicate your intent.

You have to assume the first attempt will be flawed.

It almost always is.

Right.

You have to patiently analyze the misunderstanding without getting angry at the screen, adjust your approach and try again.

That's the TOE T loop.

Exactly.

You are practicing patience, iteration, and adaptability every single time you log in.

What if this constant, messy loop of collaborating with uncertainty is exactly the kind of resilience we need in the rest of our lives?

That is fascinating.

Right?

A deterministic mindset makes us rigid.

But learning to direct a probabilistic machine might actually be the best training ground we have for navigating a world and human relationships that refuse to stand perfectly still.

That is a profound takeaway.

We are utilizing the technology to build our own tolerance for ambiguity.

So the next time you are sitting at your desk and that familiar sinking sensation creeps in because the software isn't doing what you want.

Just take a breath.

Take a breath and reenter the loop.

Remember that you aren't staring at a broken reflection of your own intelligence.

You are just standing on moving ground, learning how to direct the jazz band.

Thank you.

So much for joining us on this deep dive.

Keep questioning the buttons, keep looking at the architecture underneath, and we will catch you next time.