WEBVTT 00:00:00.000 --> 00:00:03.220 OK, so we started using advanced AI because, 00:00:03.259 --> 00:00:05.860 well, it promises speed, right? And a technically 00:00:05.860 --> 00:00:09.099 perfect draft every single time. But now there's 00:00:09.099 --> 00:00:12.080 this strange thing happening, that perfection. 00:00:12.429 --> 00:00:15.449 It's actually stripping the writing of its its 00:00:15.449 --> 00:00:18.109 human voice and ends up sounding robotic Forgettable 00:00:18.109 --> 00:00:20.489 really and that's the really tricky part now 00:00:20.489 --> 00:00:22.870 is an article that sounds perfectly, you know 00:00:22.870 --> 00:00:26.129 safely AI generated That's almost worse today 00:00:26.129 --> 00:00:28.910 than something a bit clumsy, but clearly written 00:00:28.910 --> 00:00:32.890 by a person worse how so because that robot voice 00:00:32.890 --> 00:00:36.909 signals Well, it signals a lack of honesty a 00:00:36.909 --> 00:00:39.090 lack of care readers. They want a connection 00:00:39.090 --> 00:00:42.060 not just an algorithm spitting out text Welcome 00:00:42.060 --> 00:00:44.079 to the deep dive. We get it. You need to gain 00:00:44.079 --> 00:00:46.240 knowledge fast, but you also want it to stick, 00:00:46.280 --> 00:00:48.240 to be effective. Yeah. So our mission today is 00:00:48.240 --> 00:00:50.619 to figure out how thoughtful writers like you 00:00:50.619 --> 00:00:55.179 can use AI's power without letting it erase your 00:00:55.179 --> 00:00:56.960 unique personality. Exactly. We're going to move 00:00:56.960 --> 00:00:59.039 past the usual, oh, just write better prompts, 00:00:59.619 --> 00:01:01.939 advice. That's part of it, sure. But we're really 00:01:01.939 --> 00:01:04.060 focusing on the editing phase, the human touch. 00:01:04.269 --> 00:01:08.209 We found five specific kind of sneaky robot signs 00:01:08.209 --> 00:01:11.269 that show up in AI drafts. And we've got practical 00:01:11.269 --> 00:01:13.370 ways to fix them. And it's interesting, digging 00:01:13.370 --> 00:01:15.430 through the sources, we realize these patterns, 00:01:15.829 --> 00:01:17.849 they're not just machine things. Yeah, I know. 00:01:17.989 --> 00:01:21.790 They're really signs of weak, maybe even lazy 00:01:21.790 --> 00:01:25.609 human writing that the AI just amplifies because, 00:01:26.230 --> 00:01:29.480 well, they're common, statistically safe. Right, 00:01:29.780 --> 00:01:32.180 so learning to spot and fix these actually makes 00:01:32.180 --> 00:01:35.900 you a stronger writer overall, AI or not. Okay, 00:01:36.019 --> 00:01:37.819 so where do we start? That initial feeling when 00:01:37.819 --> 00:01:39.799 you first use these tools? Oh yeah, that initial 00:01:39.799 --> 00:01:42.019 magic moment. It's powerful. You put in a prompt, 00:01:42.180 --> 00:01:44.659 chat GPT or whatever, and boom, a whole article. 00:01:44.980 --> 00:01:47.579 1500 words, outline looks great, grammar's perfect. 00:01:47.680 --> 00:01:49.700 You think, okay, this is it, the ultimate shortcut. 00:01:50.269 --> 00:01:52.709 I still wrestle with that myself, honestly, that 00:01:52.709 --> 00:01:54.569 feeling. After churning out a bunch of articles, 00:01:54.689 --> 00:01:56.730 I look back and notice, yeah, they were professional. 00:01:56.829 --> 00:01:59.170 They were neat. But they didn't sound like me 00:01:59.170 --> 00:02:02.370 anymore. The soul was gone. Where were the messy 00:02:02.370 --> 00:02:05.170 sentences, the random thoughts, the jokes? Only 00:02:05.170 --> 00:02:08.729 I find funny. And to fix that, we've got to understand 00:02:08.729 --> 00:02:11.370 why the AI writes like that, what's actually 00:02:11.370 --> 00:02:13.669 happening under the hood. Right. So a large language 00:02:13.669 --> 00:02:17.199 model. Think of it simply as a system predicting 00:02:17.199 --> 00:02:21.219 the next most likely word. It's read, like everything, 00:02:21.520 --> 00:02:23.639 billions of examples. Every textbook, every news 00:02:23.639 --> 00:02:25.560 article, every comment section. Exactly, so it's 00:02:25.560 --> 00:02:28.340 just choosing the statistically safest word to 00:02:28.340 --> 00:02:31.060 put next, what comes up most often in that context. 00:02:31.460 --> 00:02:34.300 And the result of that mechanism is, well, safe 00:02:34.300 --> 00:02:37.020 writing, predictable writing. Yep, common patterns, 00:02:37.439 --> 00:02:40.419 staying neutral. And it loves balanced rhythmic 00:02:40.419 --> 00:02:42.500 structures, loves them. Because they're everywhere 00:02:42.500 --> 00:02:44.580 in the training data, right? Billions of times. 00:02:44.620 --> 00:02:47.580 Yeah. So the outcome is smooth, like frictionless. 00:02:47.740 --> 00:02:50.960 But there's no surprise. No real motion. OK. 00:02:51.199 --> 00:02:53.759 So here's a question, then. If the AI defaults 00:02:53.759 --> 00:02:56.000 to these safe repetitive structures, what's the 00:02:56.000 --> 00:02:58.759 core job, the essential task, for the human editor 00:02:58.759 --> 00:03:01.479 looking at that first draft? Our job is, well, 00:03:01.479 --> 00:03:04.060 it's to break those patterns, deliberately. To 00:03:04.060 --> 00:03:06.000 inject life, maybe some friction, definitely 00:03:06.000 --> 00:03:08.490 surprise, back into the text. Okay, let's dig 00:03:08.490 --> 00:03:10.770 into the first couple of traps. These are structural 00:03:10.770 --> 00:03:13.189 ones. We'll start with sign one. The three -part 00:03:13.189 --> 00:03:18.110 rhythm trap. Ah, the rule of three. Yes. AI absolutely 00:03:18.110 --> 00:03:20.849 adores it. You see it everywhere. Slogans, commands, 00:03:21.030 --> 00:03:24.930 lists. It's fast, simple, and powerful. That 00:03:24.930 --> 00:03:27.490 kind of thing. It feels organized. It does feel 00:03:27.490 --> 00:03:29.669 powerful initially, but I see this, and yeah, 00:03:29.669 --> 00:03:32.689 I fall into it too sometimes. It sounds authoritative 00:03:32.689 --> 00:03:35.789 maybe, but use it too much, and the whole piece 00:03:35.789 --> 00:03:38.789 starts sounding like a... like a pre -programmed 00:03:38.789 --> 00:03:40.909 speech, like someone reading cue cards. It just 00:03:40.909 --> 00:03:43.030 flattens everything. The rhythm gets monotonous. 00:03:43.069 --> 00:03:45.650 Totally. It's like music where every single beat 00:03:45.650 --> 00:03:48.430 is exactly the same. Thump, thump, thump. So 00:03:48.430 --> 00:03:50.789 the fix is, well, it sounds simple, but it needs 00:03:50.789 --> 00:03:53.710 intention. You got to break that rhythm. Stop 00:03:53.710 --> 00:03:56.150 using three equal sentences or three equal phrases 00:03:56.150 --> 00:03:58.229 all the time. OK, that sounds good. But, you 00:03:58.229 --> 00:04:00.349 know, when you're on a deadline, reaching for 00:04:00.349 --> 00:04:04.189 that rule of three is just, it's fast. How do 00:04:04.189 --> 00:04:06.389 you actually force yourself to slow down and 00:04:06.389 --> 00:04:09.340 vary it? in a real workflow. You have to consciously 00:04:09.340 --> 00:04:12.520 edit for cadence. Try this. Write one really 00:04:12.520 --> 00:04:15.840 short, punchy sentence. Almost blunt. Then follow 00:04:15.840 --> 00:04:18.519 it immediately with a really long, complex one. 00:04:18.819 --> 00:04:21.519 Something detailed makes the reader pause and 00:04:21.519 --> 00:04:24.879 think. Ah, change the pace. Exactly. Then maybe 00:04:24.879 --> 00:04:26.899 end with a median length one. You're creating 00:04:26.899 --> 00:04:29.379 a natural flow. Kind of like how a person actually 00:04:29.379 --> 00:04:31.699 breathes when they talk. Not like a drum machine. 00:04:31.759 --> 00:04:34.040 That makes a lot of sense. Connecting ideas, 00:04:34.139 --> 00:04:37.459 showing depth, not just listing things. OK, let's 00:04:37.459 --> 00:04:40.779 talk sign two, overusing parallel structures, 00:04:41.199 --> 00:04:45.060 the not this, but that trap. Oh, this one is 00:04:45.060 --> 00:04:47.160 everywhere, especially in business writing. It's 00:04:47.160 --> 00:04:48.800 not about working hard. It's about working smart. 00:04:49.060 --> 00:04:51.139 Right. This isn't just marketing. It's storytelling, 00:04:51.439 --> 00:04:53.160 stuff like that. Yeah. It gives you that strong 00:04:53.160 --> 00:04:56.439 contrast. AI likes it because it seems definitive, 00:04:56.579 --> 00:04:58.939 easy to process. But the real issue here, the 00:04:58.939 --> 00:05:01.600 deeper one, is that this binary framing, black 00:05:01.600 --> 00:05:03.240 or white, this or that, it makes the thinking 00:05:03.240 --> 00:05:10.060 feel, well, flat. Simplistic. Real life is messy, 00:05:10.360 --> 00:05:12.540 right? It's full of complexity, contradictions, 00:05:12.720 --> 00:05:15.620 all those shades of gray. If you only offer opposites, 00:05:15.660 --> 00:05:18.639 you're missing the human part. Absolutely. So 00:05:18.639 --> 00:05:21.680 the fix here for the human editor is to intentionally 00:05:21.680 --> 00:05:23.899 embrace that complexity, even the contradictions. 00:05:24.100 --> 00:05:26.160 OK, give me an example. Look at that leadership 00:05:26.160 --> 00:05:29.160 example from the source. The AI draft might say 00:05:29.160 --> 00:05:31.420 something flat -like. Leadership is not about 00:05:31.420 --> 00:05:34.600 taking power. It is only about taking responsibility. 00:05:35.120 --> 00:05:36.839 Sounds neat, right? Yeah, sounds like a slogan, 00:05:37.100 --> 00:05:40.660 very clean. But the human rewrite, it accepts 00:05:40.660 --> 00:05:42.519 that while power and responsibility are kind 00:05:42.519 --> 00:05:44.339 of tangled up in reality, you can't always separate 00:05:44.339 --> 00:05:47.279 them so cleanly. Right. So the rewrite adds a 00:05:47.279 --> 00:05:49.120 personal perspective. It might say something 00:05:49.120 --> 00:05:51.139 like, sometimes leadership is actually about 00:05:51.139 --> 00:05:53.860 being quiet, about listening hard, and realizing 00:05:53.860 --> 00:05:56.240 that power isn't something to just avoid, but 00:05:56.240 --> 00:05:59.399 it's this. This volatile thing you have to manage 00:05:59.399 --> 00:06:02.259 ethically, moment by moment. Ah, okay, so the 00:06:02.259 --> 00:06:05.279 AI gives you the perfect, simple slogan. And 00:06:05.279 --> 00:06:08.079 the human gives you a thoughtful tension. Something 00:06:08.079 --> 00:06:10.720 to chew on. Okay, probing question time. Beyond 00:06:10.720 --> 00:06:12.939 just editing the draft, how can we prompt the 00:06:12.939 --> 00:06:16.819 AI initially to maybe avoid these simple opposites 00:06:16.819 --> 00:06:20.560 and repetitive rhythms? Ask for complexity. prompt 00:06:20.560 --> 00:06:23.279 it to explore connections, inherent contradictions, 00:06:23.639 --> 00:06:26.079 maybe even specify using varied sentence lengths 00:06:26.079 --> 00:06:28.699 from the start. Okay, so those first two were 00:06:28.699 --> 00:06:31.779 about structure. The rhythm of the writing. Now 00:06:31.779 --> 00:06:33.759 let's get into the traps that really kill the 00:06:33.759 --> 00:06:37.860 voice itself. Starting with sign 3. Using too 00:06:37.860 --> 00:06:40.500 many big reveal tricks. Yeah, that sort of canned 00:06:40.500 --> 00:06:44.279 Q &A rhythm AI falls into. That fake drama. Like 00:06:44.279 --> 00:06:46.980 what? Like, what's the secret to productivity? 00:06:47.540 --> 00:06:51.350 And the next sentence is... It's simply consistency. 00:06:51.810 --> 00:06:54.850 Oh, yeah. I see that. It's trying to create this 00:06:54.850 --> 00:06:56.790 feeling of intimacy, like it's letting you in 00:06:56.790 --> 00:06:59.250 on a secret, but without the writer actually 00:06:59.250 --> 00:07:01.290 having to show any real personality, it feels 00:07:01.290 --> 00:07:03.290 like. Like an awkward script you've heard a million 00:07:03.290 --> 00:07:06.149 times. OK, so how do we fix that? You've got 00:07:06.149 --> 00:07:08.250 to ditch the script entirely. Instead of that 00:07:08.250 --> 00:07:11.069 chief secret, share a real confession or a moment 00:07:11.069 --> 00:07:13.509 of insight you actually earned. Think about that 00:07:13.509 --> 00:07:15.990 procrastination example. The AI draft just gives 00:07:15.990 --> 00:07:18.779 generic advice. Break down your tasks into smaller, 00:07:18.980 --> 00:07:21.000 manageable chunks. We've all heard that. Right. 00:07:21.220 --> 00:07:24.319 Standard productivity tip. Super generic. But 00:07:24.319 --> 00:07:26.779 the human rewrite, it turns that into a personal 00:07:26.779 --> 00:07:29.600 admission. The writer confesses the real struggle 00:07:29.600 --> 00:07:31.939 wasn't just breaking down tasks. What was it? 00:07:32.120 --> 00:07:34.300 It was realizing the fear wasn't about the work. 00:07:34.699 --> 00:07:37.800 It was about how totally unclear the goal, the 00:07:37.800 --> 00:07:40.420 required outcome actually was. That's vulnerability. 00:07:40.779 --> 00:07:42.839 And it's instantly more convincing. Way more 00:07:42.839 --> 00:07:45.100 convincing. OK, moving to another voice killer. 00:07:45.660 --> 00:07:49.439 Sign four. Weird emphasis or fake details? Ah, 00:07:49.480 --> 00:07:52.879 yeah. When AI throws in these random specifics 00:07:52.879 --> 00:07:55.740 to try and sound authoritative or create atmosphere... 00:07:55.740 --> 00:07:58.439 Like the chipped blue mug on the desk. Exactly. 00:07:58.600 --> 00:08:02.379 Or insisting an email arrived at 3 .47 p .m. 00:08:02.680 --> 00:08:06.459 precisely. It does this because, in its massive 00:08:06.459 --> 00:08:09.300 training data, specific details often show up 00:08:09.300 --> 00:08:12.040 in writing that seems authoritative. But we notice 00:08:12.040 --> 00:08:16.100 it, right? As readers, we spot the filler. Immediately. 00:08:16.350 --> 00:08:18.329 If the detail doesn't actually matter, if the 00:08:18.329 --> 00:08:20.089 color of the mug or the exact minute doesn't 00:08:20.089 --> 00:08:21.990 change anything emotionally or informationally, 00:08:22.290 --> 00:08:24.269 it just weakens the main point. It's noise. It 00:08:24.269 --> 00:08:25.790 breaks the flow. And it kind of breaks trust, 00:08:26.069 --> 00:08:27.889 too. You promised authority, but you just gave 00:08:27.889 --> 00:08:31.189 me set dressing. Whoa. Just imagine scaling that, 00:08:31.350 --> 00:08:34.190 though. Imagine the AI processing a billion examples 00:08:34.190 --> 00:08:36.710 and noticing every single irrelevant detail in 00:08:36.710 --> 00:08:39.149 them. Every dust mote, every pen left on a table. 00:08:39.450 --> 00:08:42.549 Yeah. And it calculates statistically. OK, adding 00:08:42.549 --> 00:08:44.850 a detail like this seems correlated with good 00:08:44.850 --> 00:08:47.769 writing. even if that specific detail serves 00:08:47.769 --> 00:08:50.549 zero purpose in this context. So what's our fix? 00:08:50.950 --> 00:08:54.029 It seems straightforward. It is. Replace those 00:08:54.029 --> 00:08:58.230 irrelevant external props with living internal 00:08:58.230 --> 00:09:01.149 details. Focus on the inner perspective. So instead 00:09:01.149 --> 00:09:03.169 of the mug. Instead of the mug or the clock, 00:09:03.309 --> 00:09:06.330 talk about the strange sense of relief you felt 00:09:06.330 --> 00:09:09.330 after sending a really difficult email. Or the 00:09:09.330 --> 00:09:11.149 knot in your stomach before a presentation. Got 00:09:11.149 --> 00:09:13.779 it. The detail has to reveal something about 00:09:13.779 --> 00:09:16.740 your perspective, your internal state, or actually 00:09:16.740 --> 00:09:19.639 move the story forward. Okay, quick check. What's 00:09:19.639 --> 00:09:21.840 the core rule then? How do you decide if any 00:09:21.840 --> 00:09:24.600 detail should stay in? It absolutely has to teach 00:09:24.600 --> 00:09:26.860 the reader something specific about your perspective, 00:09:26.980 --> 00:09:30.000 or it has to push the narrative forward. No free 00:09:30.000 --> 00:09:32.379 rides for details. Perfect. Okay, we're at the 00:09:32.379 --> 00:09:34.860 final sign. And maybe the most important one, 00:09:34.919 --> 00:09:36.899 because it kind of underlies all the others. 00:09:37.440 --> 00:09:41.129 Sign 5. The missing personal fingerprints. Ah 00:09:41.129 --> 00:09:43.570 yes, the deadliest sign. This is when the writing 00:09:43.570 --> 00:09:47.169 is technically perfect. Factually correct, grammatically 00:09:47.169 --> 00:09:50.570 flawless. But it's just too clean, too distant. 00:09:51.009 --> 00:09:53.429 There's no person there you feel you can trust. 00:09:54.169 --> 00:09:56.950 Because... AI can string words together. It can 00:09:56.950 --> 00:09:59.750 synthesize facts, but it cannot generate lived 00:09:59.750 --> 00:10:02.389 experience. It hasn't lived anything. That's 00:10:02.389 --> 00:10:04.669 where the generic output feels so hollow. It 00:10:04.669 --> 00:10:07.950 has no failures, no funny stories, only it remembers. 00:10:08.070 --> 00:10:11.330 No scars, no triumphs, really. So generic advice, 00:10:11.429 --> 00:10:14.529 like that old cliche, content is king. It's just 00:10:14.529 --> 00:10:17.570 noise without your specific context, your story 00:10:17.570 --> 00:10:19.549 behind it. So this is where the human editor 00:10:19.549 --> 00:10:22.429 really has to step up. This is the intervention 00:10:22.429 --> 00:10:25.000 step. Absolutely. You have to aggressively add 00:10:25.000 --> 00:10:26.759 yourself back in. How wow, what does that look 00:10:26.759 --> 00:10:29.419 like? Add your frustrations. Your little side 00:10:29.419 --> 00:10:32.139 comments. Use unique comparisons like maybe writing 00:10:32.139 --> 00:10:34.080 isn't like building a spreadsheet. Maybe for 00:10:34.080 --> 00:10:35.820 you it's like trying to cook your grandma's really 00:10:35.820 --> 00:10:38.480 complicated recipe from memory. Okay, okay. Mention 00:10:38.480 --> 00:10:40.899 cultural influences, movies you love, books that 00:10:40.899 --> 00:10:45.360 changed you. And crucially, admit mistakes. Show 00:10:45.360 --> 00:10:48.049 your vulnerabilities. That active listening example 00:10:48.049 --> 00:10:50.870 was really powerful here. The AI just says something 00:10:50.870 --> 00:10:53.470 sterile -like. Right. Focus on the speaker and 00:10:53.470 --> 00:10:55.769 show empathy by nodding occasionally. Thanks, 00:10:55.889 --> 00:10:59.610 robot. Yeah, super helpful. But the human version, 00:11:00.070 --> 00:11:01.990 that shares the anecdote, right? About the friend 00:11:01.990 --> 00:11:04.090 who actually pulled them aside. Yes. And told 00:11:04.090 --> 00:11:06.730 them bluntly, you know, talking to you is like 00:11:06.730 --> 00:11:09.710 talking to a wall. You just don't listen. Ouch. 00:11:09.950 --> 00:11:12.190 Yeah, ouch. But the friend wasn't being mean, 00:11:12.250 --> 00:11:14.990 they were being honest. And that moment, being 00:11:14.990 --> 00:11:17.629 called out like that, it was painful, but it 00:11:17.629 --> 00:11:19.850 was transformative for the writer. And sharing 00:11:19.850 --> 00:11:22.570 that. Sharing that shows the whole messy journey 00:11:22.570 --> 00:11:24.529 of learning to listen. It's not just advice anymore, 00:11:24.549 --> 00:11:27.129 it's hard -won wisdom, and it builds instant 00:11:27.129 --> 00:11:30.129 trust because the reader thinks, okay, this person 00:11:30.129 --> 00:11:33.929 messed up too, just like me. So the skill we're 00:11:33.929 --> 00:11:36.450 really talking about here, it's not about avoiding 00:11:36.450 --> 00:11:39.450 AI. That ship has sailed. Nope. Impossible now. 00:11:39.649 --> 00:11:42.370 It's about using AI smartly. Use it for the first 00:11:42.370 --> 00:11:44.850 draft, get the structure, the efficiency, but 00:11:44.850 --> 00:11:48.629 then you have to carefully, maybe even painstakingly, 00:11:48.769 --> 00:11:51.730 edit until that draft carries your unique fingerprints, 00:11:52.269 --> 00:11:54.970 the stuff only you could add. So thinking about 00:11:54.970 --> 00:11:58.330 that, if efficiency and structure, getting that 00:11:58.330 --> 00:12:01.730 basic draft down fast is AI's biggest gift to 00:12:01.730 --> 00:12:05.600 us. What's the one essential human thing we need 00:12:05.600 --> 00:12:09.019 to bring to actually transform that draft into 00:12:09.019 --> 00:12:11.000 something memorable, something that resonates? 00:12:11.370 --> 00:12:14.309 We have to be wiser editors, I think, and maybe 00:12:14.309 --> 00:12:17.450 more honest storytellers. We have to insert ourselves, 00:12:17.590 --> 00:12:20.309 our real selves, into every single word. Yeah. 00:12:20.529 --> 00:12:23.090 So the big idea to take away today seems really 00:12:23.090 --> 00:12:25.509 clear. In this world just flooded with algorithms 00:12:25.509 --> 00:12:27.710 and, let's face it, tons of generic content, 00:12:28.470 --> 00:12:30.710 your biggest advantage, your superpower, is the 00:12:30.710 --> 00:12:33.730 one thing AI can never truly copy, and that's... 00:12:33.730 --> 00:12:36.580 Well, it's you. Precisely. It's not whether you 00:12:36.580 --> 00:12:38.940 use AI anymore. That's not the interesting question. 00:12:39.039 --> 00:12:40.340 It's how you're going to use it. Are you going 00:12:40.340 --> 00:12:42.559 to let it lead you down the path to generic sameness, 00:12:42.940 --> 00:12:44.759 or are you going to guide it with your own unique, 00:12:44.919 --> 00:12:47.200 sometimes messy, authentic voice? You decide. 00:12:47.659 --> 00:12:49.559 So our challenge to you listening right now is 00:12:49.559 --> 00:12:52.440 this. Use the tech. Embrace it. But do it in 00:12:52.440 --> 00:12:54.419 a way that makes you the memorable one, not just 00:12:54.419 --> 00:12:56.440 another faceless name in the crowd. Okay, final 00:12:56.440 --> 00:12:58.850 thought then. Something to mull over. Before 00:12:58.850 --> 00:13:01.049 you hit publish on that next piece you drafted 00:13:01.049 --> 00:13:03.629 with AI, ask yourself this seriously. What's 00:13:03.629 --> 00:13:06.490 the one specific personal failure or maybe a 00:13:06.490 --> 00:13:08.870 specific slightly obscure cultural reference 00:13:08.870 --> 00:13:10.730 that only you could possibly share that would 00:13:10.730 --> 00:13:13.549 instantly make that piece sound completely authentically 00:13:13.549 --> 00:13:13.830 human?