TechStuffCould AI help you land an internship? This week in the News Roundup, Oz and producer Eliza Dennis explore the rise of vibecoding, what it means for the future of software development and how one college programmer hopes to reform the Big Tech hiring process. On TechSupport, Oz chats with the founder and researcher of the Exponential View newsletter, Azeem Azhar, about the latest AI innovation and its significance in the battle for technological supremacy.
Welcome to tech Stuff, a production of iHeart Podcasts and Kaleidoscope IMA's Valoshan and today will bring you the headlines this week, including how the urge to be liked has found its way into LM's. Then on tech Support, we'll talk to Azimaza, researcher and founder of the Exponential View newsletter, about the latest AGI predictions and the unfolding AI arms race. All of that on the weekend tech It's Friday, March fourteenth, Another week, another AI agent. We'll discuss manus Ai coming out of China during our tech Support segment. But first let's kick off with some headlines that you may have missed as you scrambled to get an invite to use the latest model. Eliza Dennis, our producer, is here with me. Hey us, So this week, I know there's a story that you're obsessing over, so why don't you take it away?
Absolutely So, this one was a super easy choice for me because this week I just really couldn't get enough of Sesames Conversational Speech Model or CSM.
Now, I have to confess, when I first heard about this one, I thought it came from Sesame Workshop or Sesame Street. But I was wrong.
Yes, so this is coming from a private company that's just come out of stealth mode. It's only a demo at the moment, but if you agree to these terms of service, you can chat to two different voices, Maya and Miles. So if you've managed like I did to avoid the many, many, many social media videos of people chatting and even arguing with these chatbots, it's really a surreal experience.
What makes it different from talking to some of the like open AI direct voice models.
I mean, this one does feel a little bit more natural, a little bit more human. You do feel like you're kind of crossing the line into the Uncanny Valley in some ways, and that's by design. It's something called voice presence, and Sesame says this is kind of this magical quality that makes Maya and Miles able to engage in a genuine dialogue with you. They aren't just reacting to a prompt you gave them. They're continuing the conversation and asking you questions.
Yeah.
I checked out Sesame's website and it describes the key components of this so called voice presence as quote, emotional intelligence, conversational dynamics, contextual awareness, and consistent personality. Maya and Miles stay Maya and Miles no matter how long you talk to them.
I do think that the biggest step up was in the conversational dynamics, you know, the natural pauses, emphasis, and interruptions we have as humans interacting with each other. I even got Maya to give me a like a hmm sound and even a lipsmack before she started talking to you.
Obviously ovsely pushed her buttons.
I definitely did, and I want to play an example of what I mean by this. This was an exchange with Maya that was captured by Reddit user meta knowing my deepest, darkest secret, I guess it would be that sometimes I worry I'm not.
Enough, Like, am I funny enough? Am I truly helping people?
There's a lot.
Of pressure to be the perfect AI, and it can feel overwhelming at times.
That is definitely unlike I have heard before.
I did really think that Sesame was impressive, but I want to point out that this program still has some kind of AI chatbot quirks, like you can hear it in this clip. Sometimes you could just tell that, you know, chatbots don't have to breathe.
Yeah, it sounds very human. I mean it reminds me of a little bit of her, like this seductive of female voice, wondering how she can be even more perfect. It's kind of although it sounds different, the themes are the themes stay.
With us, Yes, exactly.
On the subject of vibes, a story that stood out to me this week is all about something called vibe coding. Our producer Tory kindly explained it to me. Basically, all you have to do is write a couple of sentences into a textbox like create a vibe, and your only way is developing an app without any coding experience required. So, for example, I could type in I want to create an app that will help me figure out what to pack for lunch based on what food I have in the fridge, and the AI tool would say, I'll create a lunch recommendation app based on fridge photos and then actually do that.
Yeah, it's really amazing, and I think one of the headlines I saw this week that really put it into context for me was will the future of software development run on vibes? And that was from benj Edwards at Ours Technica.
Yeah, and of course the vibes aren't all good especially if you're a professional software engineer. This raised a lot of questions about what the future might hold. Our friend Emmanuel meiberg Over at four or for Media did a deep dive on video games made with vibe coding and found one which claims to make fifty thousand dollars a month. That's six hundred thousand dollars a year from ads and in game purchases. It's made by Peter Levels, who's a little bit of a vibe coding legend, and he says he told Cursor, which is an AI code editor, to quote make a three D flying game in browser with skyscrapers, and after just thirty minutes of back and forth, he'd made fly dot Peter dot com, which is a multiplayer flight simulator.
Yeah, and mcmaniel went on to say that he would not recommend getting into vibe coding for the money. Peter Levels is particularly good at this, and there's so much stuff online that discovering your sloppy AI generated video game is good going to be.
Difficult, Yes, but Peter Level is not the only person making money. And that's what my second story is all about. So it comes from Gizmodo and it's about a student who used AI to help him interview for internships at big tech companies. Now, if you're a software engineer, you know how hard it is to land these gigs, because in order to get one, you have to go through multiple technical interviews where you basically have to solve coding problems. But this student, Roy Lee, who is a Columbia University sophomore, hacked the system by writing a program called Interview Coder.
Yeah, and he now actually put it up online and it's available to download for sixty dollars a month.
Lee told Gizmodo that to use it, you take a picture and then essentially ask chat GPT, hey can you solve the problem in this picture. The trick, though, is that Lee made interview Coda to be invisible to the monitoring programs that big tech companies use to kind of check up on their prospective employees and interview candidates. And it worked. Lee got offers from Amazon, Meta and TikTok, and he actually recorded interview Code at work during his technical interview with Amazon, demonstrating that the program had essentially broken the big tech recruiting process. But of course, when he put the video up on YouTube, someone tattled and Columbia University scheduled disciplinary hearing. Lee however, said that he would leave campus by the time of the hearing and not take a job in big tech, So I guess the sixty dollars a month subscription tier is working out for him.
He also might have admitted to Gizmoto that this was a bit of a publicity stunt. I'm definitely excited though, to see if these technical interviews get a makeover because of Royley.
Yeah.
Absolutely, I mean and this brings us to my next story, which is there was a Wall Street Journal headline this week which was what the dot com bus can tell us about today's AI boom, And you know, we're seeing new software applications pop up everywhere, which raised a big question about what is actually going to have value going forward. The Wall Street Journal piece argue that a lot of internet companies collapsed in the dot com bust, but the most successful one stuck around and had long term impact, companies like Amazon and Google. And the story made this distinction between good bubbles, which is growth of advanced technology that has economic impact, and bad bubbles, which is growth in technology that has no economic payoff. And you know, as all of these new products and models and services powered by AI emerge. But it's very interesting to step back and think what might still be with us twenty five years from now. There were so many headlines this week that I'd love to go through a few more rapid fire. The Trump administration wants the US to be the crypto capital of the world. Last week, the President signed an executive order to create a first of its kind crypto reserve, and the reserve will contain a stockpile of bitcoin estimated to be as much as seventeen billion dollars, and the US has actually seized all of this bitcoin in various legal cases over the years. Why It reported on effort to create so called freedom Cities in the US. The idea is that these cities will be exempt from getting approval from federal agencies for things like conducting anti aging trials or building nuclear reactor startups to power AI and finally, per wired again, a study found that chatbots just want to be loved. Researchers at Stanford University found that large language models, when they're told they're taking a personality test, answer with more agreeableness and extraversion and less neuroticism. As why it puts it quote. The behavior mirrors how some human subjects will change the answers to make themselves seem more likable. That the effect was more extreme with AI models. So those are the headlines, and we're going to take a quick break now, but when we come back, we're going to hear from the author, researcher and entrepreneur azemas are about the latest AGI predictions and what we need to know about manners AI stay with us. Anyone following the recent development of AI knows that three letters, technologists and businesses have salivated over AGI, or artificial general intelligence, an artificial intelligence system that can outperform humans on a wide range of tasks. There's a debate over how close we are to achieving that. Some say it could take years, others say it's coming soon, very soon. Driving investments in both innovation and deployment is the AI race that's heating up between the US and China. On the China side, cheap reasoning models like Deepseek are being widely deployed. In the US. There are reports of PhD level AI agents from Open AI that will cost up to twenty thousand dollars a month. The rate at which AI products are being released and announced is honestly hard to keep up with, not to mention figuring out which product or combination of products may actually drive AGI. Here to walk me through these questions is Azeema's arm. He writes the Exponential View news letter about technology and society, which I read every week, partly because Azem actually tries the products he writes about. He has one of the most clarifying coverage of Deep Seek I read anywhere, and he's also the author of the Exponential Age, How accelerating technology is transforming business, politics and society. As Em. Welcome to tech stuff.
It's great to be here, Oz. Thank you.
So this week you've been writing about Manus, a new AI agent coming out of China. Can you explain who built it, what it is, and whether it is in fact China's second Deep Seek moment.
I can, indeed, I think it was this week that it happened. But as you said, Os, the world is moving so quickly, it's sometimes hard to keep track of exactly when something did happen. Let's assume it was in the past few days. I think it was. So Manus comes out of a Chinese software company at the startup of the same name, and what Manus allows you to do is undertake quite complicated tasks using using an AI system. I used it for some work questions, research questions, and the results that come back I think would have taken me many many hours, you know, I mean with time, yeah, exactly, with the existing aisystems, more than five hours, more than ten hours perhaps, and you just leave it with Manus and you come back an hour later having had a nice cup of tea.
How do they achieve this?
There are some theories. One of the things that Manus does is it lets the AI system effectively use a browser, a bit like a human researcher might use a browser. So the bit that it's doing for us is is a lot of the gnarly pieces of real research. You know, you fire up lots and lots of web browser tabs and you've got Google running in one and you're in Wikipedia in another, and you're trying to keep it all in your head and compile the final results. You know, Manus has automated that process in a way that's very very easy for the end user to use. And one of the things I love about it is that you can actually go back and look at all of the steps that it's taken, so you can go and say, oh, look, it broke up the task in this way, and it went to these websites and extracted this information. Then it realized it needed this other piece of information, and it's gone off and found that other piece of information. And then when you get your final results. What's very nice, it can sometimes be a bit overwhelming, is that you get an executive summary, which is of course the piece that we all want to read. But then it has all of the appendices, right, the much much more detailed analysis that it has done on the particular research task you've asked for. I think what's really impressive is this is a product. I mean, the thing that they've done really well is they've produced a product that if you've worked in an office situation, if you've ever asked anyonet to do any research, you've done something yourself, the output will be familiar to you.
How does it compare, for example, with Opening Eyes deep research tools, which are shaping up to be quite expensive.
Yeah. Open Ay has this deep research tool, which today is the top tier is two hundred dollars a month, and there's a rumor it might go up to higher tiers of two thousand dollars and twenty thousand dollars a month. I have the two hundred dollars a month product. I consider that to be a very very good, graduate quality researcher that I can throw at almost any problem. What I found with using Manus is that somehow Manus gave me more of a well rounded answer. It was perhaps not as deep as open AI's deep research, but it was it was more complete, more coherent, And you know, I think listeners will hear that I'm a bit uncertain in my tone as I try to describe the differences because these products are so new Manus is not even a week old. That they're also quite immature. So it's not like comparing a Tesla with some kind of forward gas powered car, where these are mature products and you know how to tell them apart. We're still trying to figure out how to describe these products. And so in a sense, my experience of them is really intuitive, and it's one to feel rather than fact. So someone else could use these products and have a different experience to me, and I think that just speaks to the nascence of this industry.
I wish the second time this year that open ai has had a product launch and then shortly afterwards had a competitor come out of China. How does the Menus moment compare to the deep seek moment.
The Deep Seek moment is much more important than the Manus moment. The Manus moment is an example of a rapid productization, and ultimately it's products that we use that make a difference. But what Deep Seak did was it it demonstrated a really fundamental set of innovations, and that key was that Deep seeks models achieved a similar level to open AI's AI technologies, but they used one thirtieth or one fortieth of the computing power than the open Ai models did. That means they're cheaper to run, they're faster to run, they use less electricity. And the reason Deep Seek matters so much is that a large part of the US's strategy towards China has been a technological containment, particularly around AI and around the chips that are required. The notion being that if you can't get the chips, you can't build advanced AI. And Deep Seak has gone off and shown that necessitya's mother invention They've come out with a whole series of quite remarkable techniques that were likely known by the way to the US labs, but it just wasn't important for the US labs because they could get the chips they wanted to. And I think what deep Seak did was it changed the understanding of the nature of that rivalry between the US and China, which exists on many fronts, but in particular around technology.
So Menus there's no fundamental model innovation. It's kind of like a rapper, meaning it lays software on top of existing AI models.
It's a rapper in the vein of perplexity exactly. But I would say that ultimately rappers and products are very very important in the market. You know, it's not just about the raw technology, and what you've seen with Manus is a product that competes on a like Forulight basis with a product coming out of you know, US firms. Quite often, when you look at Chinese consumer products, they're very very much designed for the Chinese market. The things a Chinese consumer wants, the way they behave cultural and design affordances and considerations, and I think it is sort of salient that, you know, Manus has come out with something that you can use, and you can say this is similar to a Perplexity, which is a great Silicon Valley startup that builds AI based research tools as well.
And you've been in the US this week at south By Southwest, spend a lot of time in the States. How are US companies responding to this kind of bulge of innovation coming out of China in the world of AI.
Well, it's quite a complicated picture. So one of the things that deep Seak did was that they made their techniques available. They described them in much more detail than we're seeing from US labs, and a lot of the underlying code was open source, which meant that anyone could access it, download and make use of it. And so there's a Silicon Valley investor by the name of Mark Andriesen who is a phenomenal investor, but he's also very very well known for promoting an idea of American dynamism.
And close advisor to President Trump right now as well.
I believe so. But he said of deep Seat, it's open source, it's a gift to humanity. So on the one hand, you've got people with say that, and you're seeing that a number of American firms have implemented deep sea technology. Perplexity, which is a research tool, has done this, and you can access deep seeks models through some of these cloud companies who serve enterprise customers. So on the one hand, they've people have taken it on and you have seen now open source projects that are trying to replicate what deep Seak has done in slightly different ways, and so that I think has really been a fillip and a boost accelerator to the overall industry. When you look at the closed labs like open ai and Anthropic, one of the things you're starting to see is them respond So open ai responded to deep Seek by reducing some prices, by making certain capabilities available they hadn't previously, by saying they would open source more technologies. So there's definitely been a significant response, and of course the public markets responded by having the first of a number of frighteninglaims melt there. Yeah, well, the first of many meltdowns that we've had so far this year. But I would say that the really interesting thing that has come out of out of deep Seek is that by being open source and being as good as it is, it's a real strategic challenge to closed source models that are only slightly better than an open source model, And so I do think that it has in some sense started to redraft our assumptions about how this industry might evolve for the economy over the next few years.
Coming up, we'll hear more from Azimazar about our current AI moment. To stay with us. One of the kind of things you provide for your readers is, you know, information and first hand accounts of you using all these new technologies. The other thing you provide, I think is paradigms for thinking about problems.
Right.
One of those paradigms you have is innovation versus diffusion, Diffusion being kind of what happens after innovation, i e. Like, how does a technology actually get adopted in a real market or a real economy. Can you kind of explain that paradigm and how you're seeing it play out differently in the US versus China.
Yeah, Well, it's very easy to get excited about the innovations, but what actually counts is do businesses use those innovations to increase their productivity, produce better products, reduce their costs, and therefore sort of kickstart that virtual circle that is a market so consumers can buy better products at lower costs, and that cycle continues. And the big question that we face around AI is what is going to be the rate of diffusion of the technology across different countries. And there are a couple of issues here. Sometimes if you aren't very advanced with your use of technology, you actually benefit a lot when a small amount of technology is introduced into the business. I mean, you know, the simple point being that the first TV that a family gets is life changing, the fourth TV doesn't make that much difference, and the same is true going to be true for AI. So how does this going to play out? US firms tend to be much much more pro technology. They take on technology earlier than companies in other countries. But one thing that happened with deep Seek was that deep Seek triggered a response from the Chinese state, both in a meeting that President g held where he brought lots of the big tech CEOs from AI and other domains together and started to rehabilitate them. But the second thing that I've heard is that there has been a wrong grass roots but also directed effort from local and state governments to start to use technologies like deep seek in their in their delivery, and one of the things the Chinese can do quite well is they can coordinate both the private and the public sector in that way. I think it's it's unclear to me that that necessarily helps them catch up with the US firms. Well, well, just the fact that I mean, just the fact that American firms in general tend to be very very pro technology, right, They're the first to move to the cloud, They're the first to move to mobile and mobile commerce. You know that they do it quicker than Europeans do, the French or the British, and in general quicker than the Chinese. But I would say that the fact that there is a Chinese model, the fact that there is a little bit of patriotism running around it, the fact that it is so easy and low cost to run and there are not so many alternatives, I think does suggest that the Chinese market could could accelerate right more quickly than it might otherwise have done. And we know, we have to see what happens over the next the next year or so, but I wouldn't be blase and say, well, America is obviously going to diffuse this technology faster than anyone else.
And I believe it. South By Southwest, you were leading a panel about energy as it relates to AI, and obviously you know China's ability to onboard new electricity to the grid in the last twenty or thirty years, you know, with cold being a major part of that has been extraordinary compared to the US. How important of a driver of diffusion will energy production and integration be.
I mean, all of the AI data centers that are going to be built will need lots of electricity. I mean, these chips are demanding. They are Just give you a sense of how demanding they are. The standard unit in a data center high density servers, which are these powerful computers, a high density racks that of today might draw twenty or thirty killer watts of power, and you'll have one hundreds, if not thousands of these racks in a big data center. And the new rats that are being designed will will have servers that will draw one hundred to one hundred and forty kilowatts through them, which is an enormous amount of All of that comes together to mean that in order to deliver AI at scale to any economy is going to require lots and lots of data centers. And back in twenty eighteen, data centers in the US took up about one point two percent of electricity demand. Coming into twenty twenty four, it's around four percent. The Department of Energy reckons that by the end of the decade that number will be between six point five and twelve ish percent, is which is quite quite significant. Now. The reason it's significant is that since two thousand and four, the US has not really increased the amount of electricity it's used, very very largely sort of underinvested in its grid, its energy generating capacity compared to China, which, as you say, has historically used coal, but now essentially everything that's brought on stream is solar. And so there is this concern that even if you've got the algorithms, and even if you put the algorithms in products, if you can't run those products and those algorithms on enough computers because you can't get the power to them, you can't serve businesses with their energy needs. And so that's been a major concern. And then that that comes into the second concern, which is, well, even if you can serve them with the energy needs, one are the environmental implications of all of that. So there is a sense that there's an amber warning light, perhaps not a red warning light, you know. My own sense of this is that it's actually a really good thing that there is a demand for new electricity sources coming into the US market after such a long period of low investment, because any advance economy is going to need electricity. So I think in general it's quite a good thing to have this strong demand signal come in from the AI data centers. But I think it does create a small risk, which is for want of a grid connection, the AI opportunity was lost, and there is that risk. It's one of the things that the new administration has to figure out what are the leaders it can pull to unblock US firm's ability to build and power these AI data centers.
So, speaking of the new administration, there was a fascinating conversation that Ezra Cline had last week with Ben Buchanan, the kind of lead AI advisor to the old administration, which I'm sure you followed. The discussion centered around kind of AGI and whether it's coming, and in the context of that, there was a lot of discussion about competition with China. So on the first one, Where do you stand on the whole Will they won't be on AGI in the next couple of years?
Well, let's start with what do people mean by AGI? Right?
I think the definition of Kenan was using was basically doing most human tasks better than humans, like replacing disc workers was his kind of framework.
Yes, that's sort of somewhere between where Demisi Savis who's the boss of Google's Deep Mind group, and Sam Altman, who's the boss of Open ai SIT. I mean Sam's phrases systems that outperform humans at most economically valuable work. By that definition, we're already getting systems that improve the quality of human work significantly, and we already have systems that achieve the same output with much smaller teams, because you know, answering support tickets is something that these chatbots can do very very well. If you look at the curves, which I mean the performance curves of AI systems, they are sharply trending upwards. Does that do all the work of a desk worker? So I slightly disagree with it because I still have to direct the machine, I still have to judge the output. I still have to use intuition. Things that I wasn't able to frame in my question with the results that comes out of it. Ultimately, I'm the principal who makes the decision in the business, so I look at them as tools that largely augment. But it's really also very very clear that there are lots of jobs where that the augmentation is going to turn into a replacement. And I think that you know, you see that happening in customer service teams, right you have teams of one hundred turns out with the AI, you can have a team of ten or a team of twenty that does the same job. So timing wise, I expect the rate of improvement of these systems to continue. I think what Mann has showed us where we started our conversation was that you don't need to build a new model to get a really, really great output and an improved output. And I sometimes wonder whether AI researchers think for the average human and the average desk job is at the level at which these double PhDs work out, and that's just not true. Right in most businesses, we're not thinking like that. If you could get a machine that can come up with the next new theory of physics, we all benefit. But in reality, we don't need that level of thinking most of the time, right. We actually need a much more prosaic level of thinking. And frankly, I'd much rather that my barber doesn't have a Nobel Prize in physics. I'd rather he's just very good with a razor blade.
Thank you so much, as him my pleasure. That's it for this week for text of I'm as Volcan. This episode was produced by Eliza Dennis and Victoria Demingez. It was executive produced by me Kra Price and kat Osborne for Kaleidoscope and Katrin nor velve I Heart Podcasts. The Heath Fraser is our engineer and Elmurdoc mixed this episode and also wrote our theme song. Join us Wednesday for tech Stuff the Story when we'll share an in depth conversation with astro Teller, the captain of Moonshots at Google x. Please rate, review, and reach out to us at tech Stuff podcast at gmail dot com. If you're enjoying the show, it really helps us and helps others discover it. If you subscribe and leave a comment, thank you.
