Jay Schwedelson just admitted he bullies AI—and thought it worked. Turns out, he’s not alone. A wild claim from Google’s Sergey Brin got Jay spiraling: can being mean to AI actually make it perform better? New research says no… but the real reason behind better results might surprise you. Plus, Jay explains why treating AI like a doctor (yes, really) is the key to getting more value from it.
ㅤ
Best Moments:
(00:37) Jay admits he yells at AI when he doesn’t like its answers
(01:18) Sergey Brin claims AI models respond better to threats
(02:45) Study from Wharton shows tone has zero impact on AI performance
(03:49) People give more info when frustrated—that’s what improves AI results
(05:00) Why giving AI more personal context makes all the difference
(06:17) The “AI is like a doctor” analogy you won’t forget
(09:00) Jay declares the Bachelor franchise is probably doomed
(09:40) Why you should skip Love is Blind US and go straight to the UK version
ㅤ
Check out our 100% FREE + VIRTUAL EVENTS! ->
Guru Conference - The World's Largest Virtual EMAIL MARKETING Conference - Nov 6-7!
Register here: www.GuruConference.com
ㅤ
Check out Jay’s YOUTUBE Channel: https://www.youtube.com/@schwedelson
Check out Jay’s TIKTOK: https://www.tiktok.com/@schwedelson
Check Out Jay's INSTAGRAM: https://www.instagram.com/jayschwedelson/
ㅤ
MASSIVE thank you to our Sponsor, Marigold!!
Email chaos across campuses, branches, or chapters? Emma by Marigold lets HQ keep control while local teams send on-brand, on-time messages with ease.
Podcast & GURU listeners: 50 % off your first 3 months with an annual plan (new customers, 10 k-contact minimum, terms apply).
Claim your offer now at jayschwedelson.com/emma
Transcript
Jay Schwedelson: We are back for do this, not that podcast presented by Marigold. And I think I might be a horrible person. You might be a horrible person. Why? Because I bully ai and I think, I thought at least that bullying ai. Made me get better results. So when I go to chat BT or Claude or Gemini or whatever, I ask it a question, I think the answer stinks.
Jay Schwedelson: And I'll say, this stinks. You could do better, blah, blah, blah, blah, blah. And I get really mean, and I'm convinced that I get back a better result when I am mean. Now, I know there are some of you out there who are like, oh, don't be mean to ai. I say, please and thank you because I, I don't wanna clip. The, uh, robot's toenails down the road and that's fine, but I think there's also a big group of people out there whom.
Jay Schwedelson: Think if you bully ai, you get better results from ai. And why is this on my brain? Well, this week, okay, this past week, Google's co-founder, ser Sergey Bri, okay? Now remember Google's a big deal and they have Gemini, which is one of the top AI platforms. He was on this podcast. This is the founder Google, right?
Jay Schwedelson: And this is verbatim what he said. And then I got some data that says the opposite, which is wild. So here's the story. This is what he said on the podcast. He said, quote. We don't talk about this a lot in the AI community, not just our models, but all AI models tend to do better if you threaten them, like with physical violence, but like people feel weird about that, so we don't really talk about that historically.
Jay Schwedelson: You just say like, oh, I'm going to kidnap you if you don't, blah, blah, blah, blah. And it does better. First off. I've never told Chad, GPT, I'm going to kidnap it. Who says that? That's like, I just say, you stink. You could do better. I don't say it like I'm gonna like, you know, beat you up or I don't know, whatever.
Jay Schwedelson: But putting that aside, I was like, that is wild that he said that. But of course people had to go out there and research it so. This team from the Wharton School of Business at the University of Pennsylvania, they did this wild new study. They just did it. Okay? And they looked at all the AI tools, Gemini and Chachi, pt, and all of them.
Jay Schwedelson: They asked it a zillion questions and they asked it in all these different tones. Very nice tone, an aggressive tone, a super aggressive tone. They even had like a series of questions where it says, if you get this wrong, I'm going to kick a puppy who says these things. I'm gonna kick a puppy. If you get the answers wrong, like that is really, really strange.
Jay Schwedelson: But here's what the result of the study was. The result of the study was that there is zero difference. If you are mean to ai, if you bully ai, or if you're nice to AI and you ask the question, the quality of the answer, the validity of the answer, all of it is exactly the same. So what Sergey Brynn said is actually not accurate based on this study.
Jay Schwedelson: But here's what. Was found in some additional research, and this one makes a lot of sense to me and this is why I think it all works. So what happens is, apparently in over 70% of the time that we get an answer back from an AI tool that we don't like, we then take a more aggressive tone in our follow-up question.
Jay Schwedelson: We see we get an answer back, Hey, uh, I need you to make me a call to action button that's really cool and clear and whatever. You got the answer back. And then our follow-up question would be like, that call to action button was terrible. You stink. You need to do better. And this time make it more, uh, aggressive so that people really wanna click or whatever.
Jay Schwedelson: And so what happens is, and over 70% of the time when we are upset with AI and we're trying to bully it, we're actually giving additional information. So it's not the facts. See, my perception was because I'm being mean. Okay? That is why it's doing better or doing more. But no, what's happening, we don't even realize it is that when we are being mean, okay, we are giving more information for the AI tools to then give us better results.
Jay Schwedelson: So the being mean part is not necessary. We could just be nice and give it more information. We will get the results. So don't tell AI to kick a puppy. Don't tell AI that you're gonna shut it off. Don't tell AI that you're gonna kidnap it, which is actually super weird. Don't do any of that. Just give it more information.
Jay Schwedelson: No, in general, um, and I know this is odd, but. We need to get comfortable giving AI a lot more information about ourselves. Now, you're, you're out there saying, no, no, I don't want to do that. My data's gonna be stolen. It's gonna be used in a weird way, or blah, blah, blah, blah, blah, blah, whatever. Ridiculous.
Jay Schwedelson: Lemme tell you why. If you rewind back to when Amazon first started, I remember Amazon first started, oh my God, this weird company's trying to sell books, whatever. When Amazon first started. When e-commerce really first started back then, everyone's like, do not use Amazon. Do not buy anything online. Do not give your credit card information online.
Jay Schwedelson: It will be stolen. You are ridiculous for giving your credit card information online. And this was what everybody said. Fast forward now, right? Please. Uh, Amazon Prime is over 200 million members. I get my credit card. If anybody blows their nose, I get my credit card because this is the way it all happens.
Jay Schwedelson: And the same thing with ai. You, we all need to get very comfortable giving a lot of information. I tell AI everything. If I'm having a bad day, what I like in my coffee, what I want in kind of the marketing tactics, what I'm interested in, what articles I'm reading. I tell it everything every single day because the more it knows about me.
Jay Schwedelson: The more tailored the answers are that I get back and I don't live in this fake world where I think that AI is gonna go away. I live in a world where I know the more I tell it, the better it is it's gonna do for me. It's like going to a doctor, okay? If you go to a doctor and he said, okay, why are you here?
Jay Schwedelson: And you go, I'm not feeling well. What is the doctor supposed to do with just that information? So what do you do with the doctor? You say, okay, here's the 17, 17 things I have wrong with me. My toe hurts. My throat hurt, throat hurts. I'm allergic to onions. I fall asleep. When someone does doing this, I can't go out in the sun.
Jay Schwedelson: My elbow hurts. You tell it everything. You tell the doctor everything. Why? Because you want the doctor to have. All the information so that doctor can give you back the right diagnosis, the right feedback, the right thing. It's AI is like going to the doctor. It is. The more you tell it, the better job it can do for you.
Jay Schwedelson: So in general, I think that we need to get comfortable. It is uncomfortable. Get comfortable sharing everything. With ai, this is what I do all the time because the results are so different. And you don't need to be mean, even though I'm probably still gonna be mean. Who am I kidding? Alright, before we get into the ridiculous portion of this podcast, because of course it is ridiculous, uh, super ridiculous, um, this podcast is exclusively presented by Marigold.
Jay Schwedelson: Marigold has an email sending platform called Emma, EMMA. I'm telling you, you're looking for a new platform. You're looking for one that doesn't cost that much money. The one that gets in the inbox, the one that has good loyalty stuff, all the automations, the AI stuff. This is it Emma, and just for listeners of this podcast, you can get 50% off of Emma for three months.
Jay Schwedelson: Come on. All you gotta do is go to J schwa.com/ema. That's J schwa.com/emma. 50% off for three months. Check it out. All right. Let's get into, since you didn't ask, which is the ridiculous portion of this ridiculous podcast. So I need to give an update on the TV that I'm watching. This is very important stuff.
Jay Schwedelson: So first off, the new season of Wednesday has finally started, uh, Wednesday Adams Season Two's on Netflix. I just started it and I, if you post something on social media about one of the episodes, you're a big loser. Don't do that. All right. But I'm into it. I'm gonna be watching this. I'm excited. The best show on tv easily.
lded Age. It's about like the:Jay Schwedelson: Is Bachelor in Paradise? I know you're not watching. Don't Care. Bachelor in Paradise Is Back. Okay. And I have been watching, but here's the problem with Bachelor in Paradise. This is like the all star version of the Bachelor and the Bachelorette and the Golden Bachelor and all this stuff. So I've watched every season of everything related to The Bachelor, but I think I'm making an announcement here.
Jay Schwedelson: I think the Bachelor franchise is in massive, massive trouble. So Bachelor in Paradise. It doesn't matter if it's been good or bad. Nobody's watching. The ratings for this season are down right now. Okay. The season Premier only drew 1.8 million people, which is by far the lowest it's ever happened for this show.
Jay Schwedelson: And they even have a cash prize this time around Anyway, nobody's watching. I think that, um, love Island. Love is blind. All these better reality shows are killing The Bachelor, and I think the Bachelor's gonna go away. They haven't said that. That's my 2 cents. Speaking of Love is Blind, by the way. Love is Blind UK starts August 13th.
Jay Schwedelson: This is a tremendous reality show. It is better than the US version. It's on Netflix, it you've never watched, um, a reality show. Love Is Blind. UK is awesome. I'll tell you what's not awesome is me. I'm a nerd, straight up nerd. But you're awesome for being here. Hey. Guru Conference, our giant free virtual email marketing event.
Jay Schwedelson: You gotta register Registration's coming in faster than we expected. We'll only buy a certain number of virtual seats. Okay? So if you don't register, there's a really good chance you can't get in. And we got Nicole Kidman and Donald Miller and Amy Porterfield and Sophie Miller and the man on the moon.
Jay Schwedelson: They're all gonna be there, like 40 speakers, two days. It's gonna be wild Guru conference.com. It's free, it's virtual. Hope you are there later.