Science of Success: AI Takes a Deep Dive Into Podcasting
15:33 Share

With artificial intelligence, creating an ethical foundation isn't just the right thing to do, is crucial to success. Join IBM of the break to hear why from federal binet eris IBM consult into global leader for trustworthy ai.

I have to tell you about the new hit podcast. It's about any topic you can possibly imagine. The hosts seem to know just about everything. Their banter is warm and engaging. Frankly, they sound like people I love to hang out with, but I can't because they're not real a twist.

The whole city, even human, yeah.

it's very well is happening here.

The duo u just heard host, deep dive and experimental audio feature released by google this september. And the host, their voices and everything they're saying .

is AI generated to create a deep dive episode. The user drags a file, drops the link, or just dumps text into a free tool called notebook L M. They can input up to fifty P, D, F web pages, were dogs, or really any piece of information.

Notebook A M then takes that information and makes IT into an entertaining, accessible conversation. Those conversations are fairly accurate, but they're not without the hallucinations. And my interpretations many of us have come to expect from generative A I tls google calls the resulting clip of around ten minutes in audio overview.

You might just call IT a podcast and I can't stop listening. I'm not the only one from students in need of a study aid to AI tightens of silicon valley, the platform is gaining popularity. Andrea carpathia, a deep learning expert who cofounded OpenAI and let text less team working on computer vision for autopilot, wrote on x deep dive, is now my favorite podd cast.

The more I listen, the more I feel like i'm becoming friends with the host. And I think this is the first time i've actually visually liked in A I two A I. So could IT soon be your favorite podcast aside from this one, of course.

From the wall street journal, this is the science of success. I'll look at how today's successes could lead to tomorrow's innovations. I'm been coin. I write a call lumen for the journal about how people, ideas and teams work and when they thrive. This week were diving into google deep dive how IT works and when IT doesn't.

The exchange you ve heard at the top of the show came from dropping my recent column on this new google audio feature into notebook L. M. But you can make podcast out of pretty much anything, wikipedia pages, youtube clipsed, random PDF, your college thesis, your notes from that business building last month, your grand muslims on your recipe, your resume, your credit bill. I listen to an entire podcast about my foo 1k as one of the deep dive hosts put IT。

It's pretty amazing what people are doing .

with IT karpov, the deep learning expert who sang the tax praises recently, has used IT to make deep dives on mars, gold or shorts and egg and pala grands. He even generated an entire podcast series called histories of mysteries, ten episodes based on wikipedia pages ranging from atlantis to the bronze age. But there are other reasons you might want to turn information into an audio conversation.

Maybe you're an auditory learner and prefer listening to reading. Maybe you find podcast more engaging than pitch deck. To dig into how IT works, the potential pitfalls of those convincing A I voices and why you still can't trust everything they say. I spoke with deep pa. Thon, who covers A I for the journal that's after the break.

How do you start to lay the foundation for responsible A I in your organization? Here's fator. Bonnier is IBM consultants global leader for trust or the ai IT .

starts with asking the question, what is the kind of relationship that we ultimately wanna with A I? The purpose of A I is not meant to display human beings. That is meant to all meant human intelligence. Soon, as you have a glimmer in your eye about how you're thinking, you might want to use A I then asking the questions like what would be required in order to earn people's trust in such a model.

How does no book I M take whatever you feed IT and turn out an entertaining ten minute chacho to answer this question and look to the future of deep dive AI voices, I spoke with wall street journal A I reporter, deep ron. Hi deepa.

Hey, ben, how are you?

I'm okay. I feel like we are now the host of deep dialog as we sounds like, okay, so let's get into how this works. I spoke with some people over at google labs, which is a division within the tech giant that incubates and build these AI power products like notebook.

They explain to me that when you upload your source material, which is just like a link or PDF or a document or anything you wanted to be, notebook instantly digested and spits out a basic written summary. What does digests mean in this context? Like what is the tech actually doing?

Large language models are basically prediction machines. So some people call them fancy auto complete, where they run tons and tons of calculations to predict what the next most likely word is. So if you type in bed into google, obviously the most likely next one would be coin.

Yeah, totally. I mean, that's obvious. That's how IT produces these long sentences. And these huge blocks of text that sound like humans wrote IT because they've analyzed a whole lot of human text and figured out what is most likely come in certain context.

So then you could see how that same technical skill can be used for things like summarizing. And so there's a lot of different types of summarizing. There is extractive summarizing, where IT identifies specific sentences and phrases from the original text that seem really important. And then abstracting summarizing where IT generates ideas and a phrases and sentences that capture the ideas that are presented there. But either way, all of this is these models underneath the these products taking the text summarizing into stealing IT and making IT shorter and more snapp er and also more digestible for deep that specifically .

the google labs people told me that under the hood there is this model that is writing and constantly editing a script for the conversation based on the goal of being entertaining and bringing out insights or not just summarizing but specifically pinpointing the stuff that is most interesting and most surprising. How does that work like is is bringing out new thoughts here.

It's kind of I mean, that sort of what this abstraction summize ation ideas about is looking at the text in identifying what are things that humans typically find very interesting about this text, and then bringing that out, and then using math again to figure out the best way to actually take that text and summers ize IT and think about IT. And I use thinking air quotes, which you can see on a podcast, but i'm doing that because it's just layers of math being used to figure out what the most likely and the most liked version of the sentence and these sort of summaries would be. And sometimes it's using the language that you would see in the text itself, and sometimes it's using in other languages that you d never appeared.

yeah. But IT can also make mistake. We've all heard, and maybe these fluctuations that LLM s have managed to produce and IT is capable of bias and draw connections and influences that don't work or maybe don't exist or don't make sense. This particular product seems to be Better at that than other viral AI products, in part because it's based on the source material alone. But why is IT still making mistakes?

It's gonna be hard to completely avoid mistakes in any large language model based product. It's never gonna a zero. And that's partly because at its base, these large language models will produce what he thinks is the most probable answer, not the correct answer.

So an answer that sounds right, but maybe isn't right. And there are things you can do on top. It's grounded in a specific tax that can definitely cut hu cino.

But I never I mean, research has shown this a time and time again. IT never actually gets a rid of IT one hundred percent and IT could still happen because it's still Operating in that principle. It's not really trained to understand what the right thing is. It's training to think about the .

most probable thing. Okay, let's talk voices. They are not perfect, but they're really close, like really close. What can make an AI voice sound? Human.

human voices. I mean, really studying human voices again and again. And the volume game, a lot of these also are training like gigg antic purposes.

It's it's just listening to human voices over time. IT starts to sound a little bit more human. You you mention it's not perfect. It's definitely not. There are certain things that humans do like pause at awkward moments or you like have just and say a lot of arms and start and then restart.

Sentence is basically the way i'm speaking this entire time is a very human way of speaking because it's sort of imperfect and you don't see that so much with these voices. They don't say sort to have as much. They don't say like as much. They start as a thought and end IT all at once, which I don't think a lot of humans do.

right totally. So the google labs people told me that they actually program all those arms and posses and likes into the conversation from deep dive yeah, because they understand that like the human year has evolved over the years to expect those things right.

If the voices spoke like computers in these perfect sentences, nobody wants to listen to that, right? And so in order to make IT more listenable, they also had to increase the noise, as it's called, right? And so it's actually worked for this.

They say it's called diflucan ces. You need more of these diflucan ces in these conversations to actually make them sound like conversation. Like in order to make IT Better, you have to make IT .

slightly worse. It's interesting you mention noise to there's something about the frequency which an AI voice speaks that's different from the way human speaks. So there are certain techniques you can use to even if you and I don't hear IT software, can hear IT, and they can tell the difference between human and and A I, but even those kinds of detection mechanisms are getting harder and harder to different.

Eight, I fed notebook, my own peace about deep dive. And there's this part where they are talking about IT and saying, the google has managed to make the A I voices sound human and that parking is like this.

They really figured out how to make A I sound human.

And that makes the whole experience so much more enjoyable. I mean, you can get drawn into the conversation like you're hanging out with friends or something. But let's be real. This technology can be perfect.

Wow, head.

so there's something super uncanny here that makes me a little but uncomfortable, but also like really fascinated. What is your reaction?

There's something to that. There's a lot of A I S to talk about this and one of them mean principles of all these AI products is supposed to be, you know, you're talking to an A I system like it's not supposed to cosplay. It's not supposed to be human or pretend to be human.

And so that little moment is interesting because IT, it's both weird, right? So you kind of sounds not human but also it's um it's pretending not to be human and that I think struggles that line between disclosure and hiding behind something. How would you want this system to handle this interaction in the future?

All of this is pretty mind blowing. You are in this every day. I come in every now and then and have my mind blown by a.

But what does this mean? Is this a toy? Is that a peek into the future? Is that somewhere in between? How are you thinking about?

sure. I think you'd be both. I think you would be a toy and a peek into the future and a warning sign. Before I covered A, I covered social media for eight years. And I think that when you look at the lessons of social media, which are its incredible, is a great way to connect with people that you don't Normally talk to. There's a lot of good if I can come from that.

And then there's also quite a bit of bad, right? We've also seen that globally in terms of disinformation and all friends of things that make you kind of wonder to what extent is this broader project worth IT. And I think A I has a really similar potential in the sense that IT is really useful and like we're enjoying IT, but we're also going to have to navigate the future where we're developing relationships, what basically math like software systems, and we're going to have to figure out what the guidelines are for those kinds of relationships. IT just creates a whole flurry of fascinating questions that I do think are very much a part of the future. But yes, IT is also toy that's fun to play.

Deepa IT was delighted to have a conversation with you and actual human being. Thank you for making .

a smarter about this. Thank you for having me.

And that the science of success this episode was produced by charlik garden burg, Michael level and jsc fenton wrote our theme music we had helped from Daniel Lewis. I'm actually then coin, be sure to check out my column on W S. J. Dog column and if you like the show, tell your friends and leave us a five star review on your favorite platform. Thanks for listening.

Earlier, we discuss what responsible AI looks like in practice. Here's fator point energy from IBM consulting again on why that begins with data.

My favorite definition of the word date up is an architecture, the human experience. A I is like a mirror that reflects our biases back towards us.

But we have to be brave enough and introspective enough to look into the mirror and decide, does this reflection actually line to my organization values? If IT allies betrays parent about, why did you pick the data that you did? If IT doesn't a line that when you know you needed to change your entire approach.

learn more about IBM artificial intelligence consulting services and IBM dot com flash consulting.