Loading summary
Transcript47 lines
- [00:00]
A
Foreign. You're listening to Perpetual Traffic.
- [00:11]
B
Hello and welcome to the Perpetual Traffic podcast. This is your host, Ralph Burns and the founder and CEO of Tier 11 alongside my amazing co host, Lauren E.
- [00:21]
A
Petrulo, the founder of Mongoose Media, who.
- [00:24]
B
Has dressed very casually today but still has the good hair which you got to check out over at perpetualtraffic.com forward/YouTube. Today we're talking about deal Deep Seek. If you don't know what Deep Seek is, well, maybe you should read the news every now and then, scroll through your newsfeed because this stuff is blowing up as of really just this past week and I think there's a lot of confusion as to what it is. And today's episode is really is on Deep Seek explained what you need to know as a marketer and how relevant is it to you at this current moment in time if you have not been using either chat GPT or Gemini? I tend to use Gemini a lot. I'm not sure why, but I find it actually is a little bit more helpful for me in the tasks that I do than chat GPT. Then this is sort of the next evolution. It's really interesting because if you're a stock market watcher like I am, I think a lot of our listeners are. I know Lauren now is one of the best stocks of 2024 was Nvidia and they make the GPUs which is basically the chips that power AI and they're super duper expensive and their profits have just been insane. It's been, I think it was the top stock of 2024 and it got cut this past week. So there was billions, several hundred billions of dollars in market cap in literally a day because of this Deep Seek news, which has been coming for a little bit of time. And if you don't know what it is, it's a Chinese AI firm that seemingly kind of came out of nowhere. It wasn't really even a startup specifically for AI. They kind of did this as a. That's the story we're getting right now. And it could potentially rewrite the future of AI and power AI and large language models that we've talked about many times here on the show at like 1/10 of the cost and maybe even 1/100 of the cost. So it's a game changer. I know you've done a lot of reading on it this week. I'm sure you've had a lot of questions from your staff. What's your take on it at this point in time? From what you, you know about it.
- [02:31]
A
Honestly, and forgive my voice, I've got the hot tea. So if you're watching a little scratchy. But for me, honestly, I'm super excited about it. I know that some members on my pack were nervous. They thought it was super disruptive and there's a little bit of a panic component. However, I was like, finally, there's going to be a larger step in innovation. Because like you said, the expense behind that gpu, it was so expensive that they used to have armored cars transport those chips. It was inaccessible. So the idea of a product that was developed at a fraction of the cost and the investment cost now allows everyone else to start from a different benchmark. I think of it the same as where like a lot of countries in Africa never had to deal with dial up. They went straight to fiber. Or how like they get to have those jumps. They didn't have to suffer through the. That AOL dial up tone. Oh my gosh. And if someone answered the phone while you're in the Internet, it just cut everything off. I think this is super empowering and like the best course of action because of the accessibility it finds. It's like, you know when people run. No one could run a mile under four minutes until someone did. And then all of a sudden five more people did. So it's that forced innovation. I worked at the Walt Disney Company. I saw it firsthand. Walt Disney World dominated the tourist market in central Florida. And then Universal Studios launched Harry Potter World. And their first year of opening, Harry Potter World eclipsed the last seven years of revenue. The totality of one year was greater than the last seven years combined.
- [04:12]
B
That's insane.
- [04:13]
A
Which lit a fire under the mouse's butt to add more to their parks. So what happened? Central Florida then got new Fantasyland. Universal responded back with another Harry Potter Land, which meant that then Disney had Avatar Land and Star Wars Galaxy Edge. Like, it's just forced innovation. And the Central Florida economy grew because of that competition. So I think deep seat coming in, having a more affordable and potentially better AI solution will mean that more small businesses will have access to the AI technology that they were limited to. Venture capitalists are going to be able to start investing in alternative businesses because all the money is going into AI solutions. And now it can start to stabilize innovation in marketing, in my opinion. I'm super excited.
- [05:00]
B
Yeah. I mean, I think first there was a lot of paranoia to start. Like, I know this is what I know. Like it's a widespread issue is like when my wife is texting me about it. And she's like the least techie person in the world. She's like, should I be afraid of Deep Seek? Should I download it? I'm like, don't download it quite yet. You really don't need it. It handles different types of tasks than I think a lot of marketers handle. And there's some videos that we'll leave. Our good friend Matt Wolf, who was on the show about a year ago, did a great video on this which explains it in super techie terms. So we're not going to get into the super techie terms here, because these are marketers that we're talking to here. Lauren E. Petrulo so we're not going to go into that depth of it.
- [05:43]
A
But we're not going to go deep on Deep Seak.
- [05:45]
B
We're not going to go deep on Deep Seq or Deep Fake, either one. So just to give you an idea, like the H100 chip from Nvidia, which is their most powerful chip, costs about $40,000, 40, 45,000, which is insanely expensive because I mean chips right now. My son actually works for Applied Materials, which makes the tools that makes the chips. And any chip that you have in your car, like it's a fraction of the cuts. It's like pennies on the dollar just by comparison. So these are super, super expensive. So there was an opportunity in the market that this company called High Flyer and the CEO Lang Wen Feng realized there was a hole in the market. They started quietly sort of accumulating not the H100 chips in the past few years, but because of Chinese trade restrictions or exportation restrictions, they were actually importing the less restricted H800 GPUs chips. So those were the ones that they used, the less expensive ones, slightly less powerful, to then tie in and create this Deep Seek AI platform, which is as fast, if not faster, at 1/10 to maybe even 1/100 of the potential cost, which changed the entire market here.
- [07:08]
A
And that's at the OpenAI's current paid model. That's the comparison that we're making it against ChatGPT's paid version versus this current free R1 version, one potentially surpassing the ability of a paid solution that's more widely used at the moment.
- [07:26]
B
Right.
- [07:26]
A
However, Ralph, there's like one big piece of this conversation that I think is worth noting where obviously with the conversations going on politically, we're based in the US there was a potential TikTok ban still up in the books. The terms and conditions of Deep Seq expressly state that they will use any inputs you provide it to train its model and that all data is stored on Chinese servers inside the People's Republic of China. So again, that's just a small, large footnote. If you're large footnotes, like the size of a billion people's physical living space, that has incited a lot more hysteria and panic to this as well, which is going to be interesting. And like the current political climate, is this the new way that we have, like, it's not a nuclear war, but an intellectual war that's going to happen. So like from the sci fi version, I'm like, oh, I can't wait to see what TV shows and books are about to be written this year because of this type of stuff. But that's an important aspect to the conversation.
- [08:31]
B
Now here's the staggering thing, is that if you look at as of this, the publication of, as of this recording of this podcast, rather not necessarily a publication because we're not exactly sure what day it's going to come out, is that the Deep Seek app is the number one downloaded app like Leapfrog ChatGPT on app downloads as of Tuesday of this past week. So even with that potential threat and you know, the scaredy cat text from my wife, God love her, but this is when she's like, I don't even know anything about AI, but this is really bad, isn't it? I'm like, well, I wouldn't download it right now until we know a little bit more. But the fact that in the terms and conditions and this is a big issue, it is actually housed on Chinese servers. And if you know much about the Chinese government, we do know that they have very tight relationships with a lot of their companies, which is the reason why our friends over at TikTok in New York City by way of Singapore, by way of Beijing are in a bit of hot water right now as to whether or not they're going to be a viable platform in the coming year. The question really is, is what is being done with the data that's being gathered by all these Americans that are downloading thing and entering in some pretty personal questions into chat GPT. Like I ask the reason why I do it on Gemini. I realized this the other day is that we have a group chat GPT for tier 11. And so I don't want, oh no, I don't want anybody. Like I'm so cheap, I don't want to get the $20 version. So I was like, yeah, I'll just use Gemini. And the thing that I'm realizing is that because it's tied into my Google environment, which is my entire world is on Google, I get better results that sound more like me and that like they understand who I. Because once again, going back to the memo that Qasem intercepted two, three years ago, Google does have 72 plus million data points, demographic and psychographic factors on every human on the planet that has access to the Internet. I should say that as a caveat. So they do know a lot about me personally. So I do feel that it actually does come through in Gemini, whereas I know that's an American company. But I think we're all out there doing as best as we can right now to balance privacy, our own privacy concerns with using new technology. And for Deep Seek, I'm going to hold off on downloading it, I think for a time. That's just me though.
- [10:57]
A
That's fair, that's fair. I'm sure a lot of people listening to it are like, oh my gosh, do I need to delete this right away? I mean, there's the one thought of, it's already on your phone. There's the other thought of like when I got into marketing and when I was working with university, the amount of information that I knew was so easily available about consumers that early on, I'm not gonna say how long ago, but I was like, all the information is already out there. So I'm not the person that's like, oh, I'm super protective of my data. I know I should be should being the word. I just at the end of the day, there's only so many things I can focus on at a time and I think where my alignment with you not having it connect as a team. Like, I actually really love meta AI. So I'll go into WhatsApp and I'll talk to Meta A about emotional stuff, like how I'm physically feeling, how my mental spaces at various different times. If I'm triggered by something, my immediate journaling is to meditate.
- [11:49]
B
No kidding. I didn't realize that. That's super interesting.
- [11:52]
A
Oh, it's like free therapy. Oh my goodness. And if my team had access to that.
- [11:57]
B
Oh God, no you don't. Yeah, you don't want to put that into a group chat GPT account.
- [12:03]
A
No, no.
- [12:04]
B
So interesting.
- [12:04]
A
But I like, I tell Meta AI I give nicknames to everyone, so I don't use any real name just in case, dear God, it comes out.
- [12:12]
B
So then it knows even more like, oh yeah, they know you.
- [12:17]
A
I don't want. Well, I got tired of seeing ads for preschools when I don't have any kids. Like I was like, give me relevant ads. But the thing is, you have to know that because I do so much with advertising and because I'm going to be like overseeing different ad accounts and I'm checking on stuff, the engagement that I have on ads, if it's a good ad, I'm going to be super into it. So I've gotten advertised from dating apps to divorce and everything in between because I just send all the wrong signals.
- [12:44]
B
They're totally confused as to who you are from that business standpoint. But however, the meta AI personally, yeah.
- [12:52]
A
It knows my deepest fears and anxieties.
- [12:55]
B
Super good. So this is not necessarily a marketing thing, but I think it's important to understand like what deepseek does best. And like I said, we'll leave a video in the show notes for Matt Wolf, who he sort of goes through this, which is fascinating. And what does it do best? It seems like it's good at reasoning and problem solving. Like this deep seq R1 model excels at these tasks that require sort of logic and reasoning, deduction, sort of step by step analysis. So if you're doing that kind of work, you know, if you're listening to the show, chances are maybe you're not. I mean, these are coding challenges, scientific research, strategic planning, high level things. There is definitely that part to it. What's really interesting is when you put in a deductive reasoning question, you can actually see it, thinking through it. It's fascinating and it talks to you as you're doing it. Like I said, we'll leave links in the show notes for that video because it really does show exactly how you do it. The other part to it is like it seems to reduce bias. And this is one of the things that Matt talks about is that claims to have like tackled this issue of bias in AI head on. It curates training data and avoids these fine tuning techniques and it aims to be sort of more neutral and objective as an AI mechanism. So we'll see how that plays out. That's a sort of a big opportunity, especially when you're really looking for factual accuracy. You know, you're a scientist, you're a developer, you're a logistics expert, even an educator. Like this is super important. So there is some nuances to it based upon how powerful the platform is. And it's definitely, it's accessible and it's open source, which is another part to it that I think is absolutely Fascinating, because that is not the case. I'm not a programmer, obviously, but I do know that OpenAI is not open source. So the way that this thing is going is it's like, all right, it's going to be collaborative here. It's almost like Linux to Microsoft or even to Apple platforms, like way back, if we're looking 10, 15 years ago. So there's some things that are very, very interesting here. So like for scientists, like I said, developers, there's a developer application for it if you're in a logistics, like you're doing project planning, like I said, and even like professors and educators. So there might be some really cool applications with a disclaimer of what you said that all this data is on.
- [15:23]
A
Servers, living on a server in China.
- [15:26]
B
China that the government you do not.
- [15:29]
A
Own, it may have access to.
- [15:30]
B
So there is that.
- [15:32]
A
So, but you don't own. Let's just be super clear though. Any software you're using, their data lives somewhere else. Every single software that you interact with, we live in the space that we work on the cloud. Where does that cloud physically reside? Maybe it's servers in Amsterdam, maybe it's in Ireland. There is that aspect to every app that you're using, you're borrowing to every.
- [15:52]
B
Or you're sharing it with the mothership, whatever that is.
- [15:55]
A
Yes.
- [15:56]
B
The old argument is that the reason why I like Google and Gemini is because it knows so much about me. Like, I have all my tax returns for the last 20 years on Google Drive. I have pictures of my kids. We're going through a thing with my mom right now. Like, they know that I am the power of attorney from my mother who is dealing with like some advanced, like long term care issues, like intimate stuff, you know, they know.
- [16:19]
A
Or I drive in bed with Google.
- [16:21]
B
I'm in bed with Google, they know. Like when I drive and go to the liquor store, you know what I mean? Like, and how many times that happens? Not much in January, unfortunately, because it's been dry January. But anyway, the point is, like, they know all this stuff, but the thing.
- [16:35]
A
Is they've known this too. Like, I want to make sure that people understand that this isn't novel. It might be novel to you and you, the listener, like, you are a savvy marketer. I know that this is not applicable to like 99% of those listening, but for your clients, for your colleagues, for your family members, for them, the access to their data is a relatively new start of the conversation. If you go back to 2015, I don't know if you remember this, but meta released when they had meta dating or Facebook dating based off of how many times you interacted with each other's profiles, how many pictures you had tagged together, how quickly and how soon you had mutual friends. Meta could get to a near prophecy, like accuracy of when you two would break up.
- [17:20]
B
Crazy. It's the story of the woman. It's apocryphal now, but the woman who was told by Google or was informed she was pregnant before she even knew she was pregnant. So we'll leave a link to that story in the show notes too, which is a super interesting one. But it's like all of these systems, this is the price that you pay for convenience. This is the price that you pay for efficiency. I was looking up the other day on Yelp, we were trying to go to, like a good Chinese food restaurant. Google now knows I search, like Chinese food near me. They know I like Chinese food when I'm in Natick because I'm with my mother at her, you know, rehab facility. That kind of stuff. Like all these things that get pieced together, I'm comfortable with that. Do they know my Social Security number? Do they know my checking account numbers? They probably do to a certain degree.
- [18:10]
A
They probably.
- [18:10]
B
They probably do. So this is the price you pay. And I think if you're dealing with clients, if you're dealing with people individually in your office and they have questions about this, hopefully you can sort of formulate a narrative based upon today's show. There are risks inherent with everything that we do online. And yes, the privacy policy for Deep SEQ specifically states that user data is stored on secure servers in China. Whether or not that is a threat to you or not, that is a personal decision.
- [18:40]
A
Yes. Our decisions are there's enough data already out there, at least for me, where we will be using this and looking into it as a company, but we will not be doing this by touching any account or anything. There's a lot more research phase because we develop a lot of AI assistants and our AI assistants leverage ChatGPT and a few other language learning models. So while it's still super new, it's something that we have to be mindful of. Is this going to become a new standard for AI enabled? What becomes the ownership of it? There are previous AI tools we used to use that we discontinued our relationship with because when we had a lawyer look at the terms and services, we lost all ownership of anything we created with it. All ownership. So we had to do a mass exodus from one AI software we were using and it became like a big internal discussion. But for me at least just being transparent. What makes me interested and intrigued by R1 is what will that mean for AI solution providers? Will that become a new standard? Will that become something that our clients absolutely want nothing to do with? Does that become part of like our AI assistants? Will people start engaging with our virtual assistants, asking where is the data stored? So that's just the perspective, while it's still super new, that we're evaluating whether or not this will become a solution we'll introduce for our company and introduce for our clients. Right now, obviously we're not making any decisions because there's so much to be known and understood about it. But from a personal level, everyone already has my data. I'm not super nervous.
- [20:13]
B
Me neither. So, well, hopefully this has been helpful for you all here today. I mean, this is obviously this is a developing story with a lot of consequences, potentially like far reaching consequences. And we'll continue to update you as we get information here on the show so that you can make informed decisions as to which way you're going to go with regard to Deep Seek and AI just in general, because it's one of the biggest hot topics in 2025 and beyond. And I would say more beyond because it's just going to become more front page headlines here. So this was a huge story that broke this week.
- [20:49]
A
So if you're listening, tell us on Spotify.
- [20:51]
B
Tell us, tell us on Spotify what you think. That's right.
- [20:54]
A
Absolutely.
- [20:55]
B
We are monitoring those comments and questions. So make sure that wherever you do listen to podcasts, be it Spotify, Apple Podcasts or other, you leave a rating and review or leave some comments in there. We'd love to know what think. And of course, all the resources that we mentioned here on today's show are over@perpetual traffic.com so on behalf of my awesome co host, Lauren E. Petrulo, ciao till next show. See ya.
- [21:25]
A
You've been listening to Perpetual Traffic.