How Computers Think #72

00:00:00: You know that feeling when you open up a brand new completely blank document on your screen?

00:00:05: Oh,

00:00:05: yeah.

00:00:05: It's just this stark white emptiness.

00:00:08: right and Right there at the top left corner is That little blinking cursor

00:00:12: right.

00:00:13: Just pulsing

00:00:14: pulse Pulse pulse.

00:00:18: it Is um it waiting for you.

00:00:22: It almost feels like an interrogation.

00:00:23: I really does, its demanding input

00:00:25: Exactly!

00:00:26: You sit there staring at the machine expecting to provide some kind of momentum but on a hardware level The Machine is just sitting in perfect silence.

00:00:33: Its idling running this continuous background loop Just waiting for a hardware interrupt from your keyboard.

00:00:38: Yeah physical keystroke

00:00:40: Right and that stark blinking void Is exactly what we were plunging into today.

00:00:45: Because, well our listener provided a source document that consists of exactly seven words.

00:00:50: Seven words which I mean forces it completely different kind of analysis than we usually do here.

00:00:54: Yeah!

00:00:54: Completely

00:00:55: different.

00:00:55: We aren't synthesizing you know massive chapters of data today...we are essentially deconstructing a microscopic fragment of text?

00:01:03: I have to admit when i first look at this..I thought were getting punked.

00:01:07: Oh!!

00:01:07: I totally though was joke too right?

00:01:09: or someone just hit copy paste on Like a title field by accident and submitted it to us.

00:01:15: Yeah, like a clerical error

00:01:17: But let me just read the entirety of our source material for this deep dive.

00:01:20: here It is quote excerpts from how computers think.

00:01:25: How?

00:01:25: Computers Think Wow that Is it.

00:01:29: That Is The Entirely Of The Text.

00:01:31: Seven Words.

00:01:32: It's incredible.

00:01:32: Continuing a title and then just a repetition of that exact same title, right?

00:01:37: So is this a profound philosophical statement or are we literally just staring at a broken data string?

00:01:43: Well the mechanical reality Is it might very well be a clerical error like script pulling a metadata tag into body text

00:01:50: field.

00:01:51: Right A glitch.

00:01:52: But whether it was intentional Or a glitch The text exists as an artifact.

00:01:57: now And analyzing that artifact actually reveals an enormous amount about the assumptions baked into our relationship with technology.

00:02:05: Well, that's interesting... Yeah because when you have zero-body text to rely on The framing of the title carries all the weight!

00:02:12: The premise itself is loaded with assumptions about us About machines and about that semantic space between the two.

00:02:18: Okay

00:02:19: Let's unpack this.

00:02:20: we have to start with the verb right?

00:02:21: That title is how computers think.

00:02:25: Think.

00:02:25: yeah by using the word think.

00:02:27: Isn't the source immediately applying a biological conscious trait to his system of silicon and copper?

00:02:33: Oh, absolutely.

00:02:34: It's personification right out-of-the gate

00:02:36: because I was gonna say it feels like calling a car's engine its heart Mm-hmm.

00:02:41: But that analogy doesn't actually work does it

00:02:44: not really now?

00:02:45: Because a hard is just a mechanical pump pushing fluid And an engine has a mechanical pumped pushing pistons.

00:02:51: They are mechanically similar.

00:02:53: Right The physics or parallel

00:02:54: Yeah, but a brain is entirely different from a CPU.

00:02:57: I mean the cpu operates on absolute binaries true or false one-or zero Yep

00:03:01: voltage or no voltage

00:03:02: exactly while a brain operates on chemical gradients and noise in parallel neuroplasticity.

00:03:09: So using the word think here feels i know fundamentally deceptive.

00:03:13: it Is deceptif?

00:03:15: And It's a deception that actually dates back to The very origins of computer science.

00:03:19: really how far back are we talking?

00:03:21: like all the way Back?

00:03:23: When pioneers like John von Neumann were first establishing computer architecture, they deliberately borrowed biological terms.

00:03:30: Just to make it easier to explain?

00:03:32: Exactly!

00:03:33: To make these incredibly abstract room-sized machines comprehensible to the public and other engineers... They called data storage memory.

00:03:42: Oh

00:03:42: wow right!

00:03:43: Memory.

00:03:43: I never even think about that as a metaphor anymore.

00:03:46: It's totally normalized now.

00:03:48: They call processing pathways neural networks long before modern AI.

00:03:53: But the mechanism of a computer processing data is entirely distinct from biological cognition.

00:03:58: Right,

00:03:58: because there's no actual recalling happening just fetching data form physical sector on your drive?

00:04:03: Yes!

00:04:04: A machine operates on Boolean logic.

00:04:06: Tiny physical switches, logic gates that are either open or closed... it lacks the biological architecture to experience anything.

00:04:14: It's

00:04:14: not feeling the data?

00:04:15: No Not at all.

00:04:19: Certainly doesn't think in any human sense of the word.

00:04:21: It executes instructions linearly based strictly on voltage changes.

00:04:26: Okay, but the title doesn't say how computers execute voltage Changes.

00:04:31: no it doesn't.

00:04:32: its as think and That linguistic choice has to change the user interface right?

00:04:37: It changes How we interact with this thing.

00:04:39: Oh, it changes everything about our relationship To the output.

00:04:42: because if I conceptualize a computer is something that calculates I treat it like a highly advanced abacus.

00:04:50: Right, A tool?

00:04:51: Yeah,

00:04:51: Tool!

00:04:51: So if its spits out the wrong number i assume there's flaw in formula or entered something wrong.

00:04:57: But if conceptualize as something that thinks Doesn't fundamentally alter the hierarchy of trust.

00:05:02: It shifts user from treating machine as a tool to treating agent

00:05:06: An Agent?

00:05:07: Wow

00:05:08: The way we name a process basically governs our psychological reality.

00:05:12: When we assume the machine is thinking, We project our own internal lived experience onto whatever it outputs.

00:05:18: Right

00:05:19: because know what thinking feels like to us.

00:05:21: Exactly!

00:05:22: Human-thinking involves lived experience, context, nuance, physical sensation... ...we naturally assume that thought requires consideration of broader contexts.

00:05:32: So when system that's framed as thinker gives an answer We are far more likely to defer it's authority.

00:05:40: Because we assume that has considered variables we might have missed!

00:05:44: Oh man, this perfectly explains the terrifying phenomenon you see on news…

00:05:49: Which one?

00:05:49: Where people blindly follow their GPS directions right into a lake... Oh yes

00:05:54: or off of collapsed

00:05:54: bridge?

00:05:55: Yes they assumed machine is thinking about the

00:05:57: route Right.

00:05:58: think its looking at window with them

00:05:59: Exactly They project human level situational awareness onto device in dashboard.

00:06:05: They assume the GPS knows it's raining, or knows that bridge is out.

00:06:09: Or you know... knows a lake isn't a road.

00:06:12: They surrender their own spatial reasoning.

00:06:14: Completely!

00:06:15: Because they trust in machine thinking.

00:06:16: The mechanism behind this misplaced trust Is basically massive failure to understand the routing algorithm itself.

00:06:23: How so?

00:06:24: What does the GPS actually do?

00:06:26: Well.. A GPS isn't looking at world at all.

00:06:29: It is evaluating vector data.

00:06:32: It uses path-finding algorithms like Dijkstra's algorithm or A star.

00:06:37: I've heard of A Star.

00:06:38: Yeah, they're very common in computer science and what they do is assign numerical weights to different segments of a graph.

00:06:45: it calculates the path of lowest mathematical resistance based strictly on the static map data it has stored in its memory.

00:06:51: So it literally just doing math.

00:06:53: there's no map in its head right?

00:06:55: Just a web of numbers.

00:06:56: If the map data incorrectly labels a boat ramp as a road, The algorithm just assigns it away calculates the shortest distance and issues instruction.

00:07:05: Turn left!

00:07:07: There is zero comprehension of water danger or physics.

00:07:11: This system is merely evaluating numerical weights in completely closed loop.

00:07:15: But because interface feels so conversational You know, with the pleasant voice telling you to turn left and five hundred feet.

00:07:20: Exactly!

00:07:21: The human user fills in all those cognitive gaps with their own biases... ...the verb THINK creates this very dangerous illusion of comprehension.

00:07:30: So bringing it back to our source material.. ..the title How Computers Think is essentially bypassing entire philosophical debate about machine consciousness.

00:07:38: Right It doesn't ask, do they think?

00:07:41: No it boldly assumes the premise as a settled fact.

00:07:48: It

00:07:56: completely disarms your skepticism.

00:07:58: You aren't even invited to argue?

00:07:59: Exactly, you weren't invited about whether the machine possesses an inner life... ...you are simply instructed to observe mechanics of

00:08:08: it!

00:08:08: It forces a human reader into submissive posture.

00:08:11: Yes

00:08:11: ready to receive this so-called technical explanation artificial cognition.

00:08:16: But that illusion of hidden wisdom completely falls apart when you look at the structure of what actually follows the title?

00:08:22: It shatters, completely!

00:08:24: Let's look at a literal framing in this source document.

00:08:27: We established the title but first line says excerpts from how computers think colon.

00:08:32: The colon is doing alot heavy lifting there.

00:08:34: And then the second line, The only actual data provided in the whole document is simply...the phrase repeated.

00:08:40: How computers think?

00:08:41: Right!

00:08:42: We have a colon promising an explanation followed by literal tautology.

00:08:46: The semiotics of the colon are crucial here.

00:08:48: In written language A colon acts as a logical operator.

00:08:52: Make it equal sign Sort-of.

00:08:53: yeah It signifies that what follows will prove or explain Or list elements of what preceded it.

00:09:03: Okay, I see that.

00:09:04: And the phrase excerpts from it amplifies this promise right?

00:09:08: It implies a curated selection of deep insights pulled from a much larger comprehensive body of work.

00:09:14: So we are primed for revelation.

00:09:16: Yes and instead We get a reflection.

00:09:19: The

00:09:19: mirror

00:09:19: excerpts form how computers think.

00:09:22: colon colon

00:09:23: How computers thing its off since more function is basic recursive loop

00:09:27: Like a logic error where the function just calls itself indefinitely?

00:09:31: Exactly.

00:09:31: You asked for the program for excerpts and it points back to title.

00:09:35: I know we can dig into philosophical meaning here, but isn't that highly probable?

00:09:39: this is just a glitch?

00:09:40: Oh almost certainly!

00:09:41: A backend variable assignment when someone accidentally set the excerpt field equaling the title field.

00:09:47: From a purely technical standpoint, yes it is almost certainly a variable assignment error in whatever content management system generated the page.

00:09:56: But you know what?

00:09:57: That glitch... ...is exactly what makes this document so illuminating.

00:10:01: Why's that?

00:10:03: Because the error lays bare at the mechanical literalism of machine.

00:10:07: Okay explain that!

00:10:09: To a human reader, a tautology repeating the premise as the conclusion is a failure of communication.

00:10:14: Right

00:10:14: it's frustrating!

00:10:15: It's empty.

00:10:15: if I ask you a question and you just repeat my questions back at me i'm going to get annoyed.

00:10:19: Exactly but...to the computer executing the script.

00:10:23: there is no contradiction There is no emptiness And there is not frustration.

00:10:27: Its doing its job

00:10:28: Precisely.

00:10:29: The system was instructed to populate field labeled excerpt with data string.

00:10:35: It pulled the nearest available string, which just happened to be the title and it executed the transfer.

00:10:40: The machine state is perfectly valid.

00:10:42: The Machine is perfectly happy with this outcome!

00:10:44: It doesn't have emotions.

00:10:46: but yes...it resolved a task…the glitch reveals that computer does not care at all about semantic meaning of words..It only cares about moving data from point A-point B.

00:10:57: So the machine did its job perfectly based on parameters given.

00:11:02: The fact that the human user finds the output meaningless is entirely irrelevant to this system.

00:11:08: Completely

00:11:09: irrelevant!

00:11:09: It's

00:11:10: a perfect demonstration of the gap between our human anticipation and meaning, And a machine's literal execution of a task.

00:11:18: Look at it through the lens on information theory Specifically, Claude Shannon's foundational work back in the nineteen forties.

00:11:24: Oh shannon he was like a father of digital communication right?

00:11:26: He

00:11:27: is!

00:11:27: And information theory data and meaning are entirely decoupled

00:11:31: De-coupled but they don't rely on each other

00:11:33: Exactly.

00:11:33: Data is simply measure of entropy Of bits being transmitted across a noisy channel.

00:11:39: Whether those bits form a Shakespeare sonnet or just random static To transmission channels it all just data.

00:11:45: I get it.

00:11:46: The machine successfully transmitted the bits for the characters and how computers think.

00:11:51: It achieved perfect data fidelity, but meaning is something that only happens at the destination

00:11:59: when a human reads it?

00:12:00: Yes!

00:12:01: When a human mind decodes those bits and integrates them into a web of context... ...the source text we have contains almost zero bits of data.

00:12:12: But the meaning it forces a reader to generate by confronting that empty loop is immense.

00:12:17: Because it forces short circuit in readers brain, we approach text expecting machine explain itself

00:12:24: us.

00:12:24: We want comforting

00:12:26: paragraphs Demystifying black box and instead machines feed their own prompt back.

00:12:32: It reminds me how modern algorithmic echo chambers work.

00:12:36: That is a highly relevant comparison.

00:12:38: Right, like.

00:12:38: think about how recommendation engines function on social media or search platforms.

00:12:42: They use vector embeddings to map your preferences.

00:12:45: Exactly!

00:12:45: they track what you click.

00:12:47: How long you linger in post What share?

00:12:50: The algorithm isn't thinking of what will fulfill you or challenge.

00:12:53: intellectually

00:12:53: Not at all.

00:12:54: It's simply calculating the mathematical proximity between past behavior and available content inventory.

00:13:01: You input desire for information and the system outputs a slightly repackaged version of your existing biases.

00:13:08: It's so wild when you put it that way, we type a query into a search bar hoping to discover something new... ...and the algorithm just hands us back a mirror!

00:13:15: It gives this back our own title and calls an excerpt?

00:13:18: Exactly!

00:13:19: Just A equals A scaled up to societal level.

00:13:22: We think are communicating with this vast knowledgeable entity but were actually talking about reflection on our inputs.

00:13:29: In that friction The friction between our expectation of narrative resolution and the reality of this algorithmic reflection is where the true value of this source document lies.

00:13:39: The seven words?

00:13:40: Yes, the seven words!

00:13:42: If this had been a standard ten-page technical essay on how logic gates process binary code... ...the reader would have simply absorbed the data passively.

00:13:51: We would've consumed it like a textbook.

00:13:52: Right…The human mind would've followed the narrative structure Set up exposition conclusion without having to do any heavy lifting.

00:14:01: But because the document provides nothing but a recursive loop, it acts almost like a linguistic

00:14:06: trap.

00:14:06: It refuses to do the work

00:14:07: for you by boldly stating the premise that computers think and then flatly refusing to elaborate.

00:14:14: The burden of definition is thrown violently back onto the reader.

00:14:17: You are forced to look at the blinking cursor Yeah...you have

00:14:21: to look in white space And decide what computer thought actually looks.

00:14:26: if even exists

00:14:28: deep desire for technology to possess a recognizable inner life.

00:14:33: We want it be alive,

00:14:34: we want the machine think because if its just complex arrangement of logic gates evaluating mathematical weights then our immense reliance on it feels incredibly isolating.

00:14:44: It's lonely!

00:14:45: We wanna collaborate not just calculator

00:14:47: Exactly.

00:14:48: The text tantalizes the reader with the word think, promising an encounter within artificial mind and then it immediately shuts the door providing only the mechanical unfeeling reality of a closed loop

00:15:01: which brings us full circle.

00:15:02: we started by asking if this seven-word document was just a glitch or profound statement.

00:15:08: And it turns out,

00:15:09: its both.

00:15:10: It turns the glitch is a profound statement.

00:15:13: The absolute absence of this essay Is the most accurate representation Of machine logic we could have possibly asked for?

00:15:20: IT REALLY IS!

00:15:21: I want to thank our listener For sending in something so absurdly concise Because it forced us To look past that text entirely and really examine the architecture Our own expectations.

00:15:31: It has been a surprisingly demanding exploration of the space between human cognitive bias and literal machine execution.

00:15:38: And as we conclude, I want to leave The Listener with one final unprompted thought to mull over.

00:15:43: long after this audio ends We have spent time examining the void on the blank page and that recursive loop of a machine merely repeating its own parameters.

00:15:55: If a document confidently titled How Computers Think provides absolutely zero data, no explanations and no text beyond its own repeated title.

00:16:04: Does the act of thinking actually shipped entirely from computer to human reader who is left sitting there actively trying to decipher blank page?