Good Disruption: Generative AI
Generative AI offers benefits for content creation, but it also raises concerns about misinformation, automation, and control of online content.
00:00 / 42:00
Deepfakes are becoming more pervasive on social media, while deep-learning algorithms like GPT-3 offer benefits for content generation. The episode explores whether the benefits outweigh the risks and how policymakers can protect people from misinformation.
Transcript
The Disruption, a lively discussion between UVA Darden School of Business professors Yael Grushka Cocaine and Mike Lennox on cutting edge technologies and practices that are challenging the status quo.
Good morning, Yael.
How are you doing today?
I'm great, Mike.
Good morning.
It's exciting to be back with our podcast.
Um, what's our topic for today?
Oh, our topic is so timely, and I am very excited to hear your take on it.
Our topic today is deep fakes or generative AI.
Basically, when the machine or the AI. creates content that could be new images, new text, new stories, new poems, uh, really anything right now.
And it's hot, hot, hot in the world of tech today.
You know, we're in the holiday season, so I was thinking about, you know, this topic in the holiday season, and, and I have a, I have a hot take for you here.
Um, that we're in the midst of maybe the greatest deep fake of all history here.
I, I'm talking about one that uses numerous video assets to confirm this deep fake, that uses numerous, you know, stories that have been generated here.
It's actually a conspiracy of millions of people working together to advance this narrative, um.
And and this is a spoiler alert.
So anybody who has children, please turn off the podcast right now.
To all of our young, young listeners, please turn this off.
But I'm talking about Santa, right, the ultimate deep fake, right?
So, uh, what's, what's different about what we're talking about than historical deep fakes like Santa?
Yeah, is it fiction?
Is it imagination?
Where does deep fake become, uh, problematic?
Well, This is really taking maybe context where we thought, um, took for granted that they were truths, they were facts, they were created by people that we can attribute pictures that were actually taken by a camera, by an individual of actual individuals or items in the world, stories that were crafted by authors.
Um, so these are in the context is really things that maybe even as we grow older and less naive as kids, we still believe that they are real.
But you're right, at the heart of it. you know, maybe Santa is real.
I just saw Elf last night, and you know, you know, in some hypothetical.
Exactly, exactly.
You never know what we're going to conclude, but it is testing reality as we as we know it and as we see it, especially in terms of content that we consume and are willing to treat as maybe somewhat objective.
Now, in many ways, this is kind of our second episode on the broader topic of generative AI.
So we have an episode where we talked about creativity and AI, especially around music.
And art and I think at the end of that we kind of concluded that this was a co-creation process, uh, and especially our guests kind of pushed the narrative of this enhances, um, the opportunities for creatives.
I suspect we're talking about something different here like what again, what puts the fake in deep fake?
Yeah, it's a great question.
Is this a kind of a joint process between creators and the machine or creators and AI, or is this something that is going to basically replace the creators, um, in, in the long run.
Um, anything deepfake is really creating something uncontrollably or to, to us human, it seems uncontrollably, but it's also about the speed and the processing.
So it's connecting the dots, it's creating kind of connective tissue where normally somebody would have to do that manually or gradually, and now the AI is helping you do it for you.
So part of the deep fake is really creating, you know, combining images, faces and bodies of people that don't really exist.
Voices and faces and words out of the mouth of somebody who never said it.
So that's the deep, that's the fake part of deep fake, OK?
The deep comes from the sophistication of the algorithm.
The fake comes from creating something out of nothing, something that didn't exist before.
But generative AI right now is even broader than that.
It's about searches that are faster and more sophisticated than ever.
It's about pulling together pieces of information and distilling them and synthesizing them in a new way.
Yeah, to push on the AI part of this deep fake conversation.
Um, arguably, you know, there's been, you know, at least since the history of video, but even going further back, um, opportunities for people to try to pass themselves off as someone else, um, you know, making a literary reference like Cyrano de Bergerac and, you know, stories of that type.
Um, what difference does AI make in this?
I think it's mostly the volume.
So you're spot on in terms this is nothing new, faking content by some, you know, some kind of ghostwriter and creating a new entity.
It's not necessarily new in terms of the idea, but it's just the quality of it, the speed in which it gets generated, the volume.
So we're talking about thousands of, you know, instances where usually it would take you.
Uh, days or hours, you know, days or weeks to perfect 11 photo so it looks real.
Now you can create these real looking images, um, in a matter of seconds.
So it's just about like a lot of parts of, you know, AI and digital technology.
It's all about the pace and the quality with which it gets produced.
So talk a little bit about some of the, the, you know, things coming down the pipeline here, some of the examples we're seeing.
Yeah, well, this week is an exciting, I guess, week.
The last month.
I guess since November it's been pretty exciting because there's been a new generative AI, uh, released to the public.
So the chat, uh, GPT has been given to, given to the world as a gift from OpenAI.
We've talked about them in the past when we talked about, uh, Dali, but, um, they really, it's a company that basically shares publicly all of its technology, and they've developed arguably the world's best language model so it can.
You can prompt it with a question, you can talk to it like you would talk to a, a, a colleague or a student or a professor, and it will within seconds, craft a very sophisticated answer.
And you can go back and forth and continue to challenge this, this chatbot, and you can get very intense and, again, sophisticated and detailed responses that to the naked eye doesn't look like a computer generated them.
And the way that it gets done is using traditional, you know, AI technologies, so neural networks, you know, the deep version of those, so sophisticated models that learn on tons and tons of data, data that is out there in the public domain.
And it really allows, it will basically save people time.
So students are of course very excited to see new, new ways to study and new ways to get answers to their essays.
Yeah, and I'm glad you raised that.
So in in higher ed or really education broadly here, some, some concerns here that suddenly students will be using this to write their essays, uh, maybe for a course or maybe for college admissions.
Uh, so I am married to one of the leading college admissions experts in the country, and she's also.
Expert in particular on essay writing.
So when this chatbot came about, I decided, let's give it a try.
I, gave it a prompt that you typically see on the common app, which is, what is your favorite word and why?
And I was shocked.
Yeah, within seconds it had given me a full essay on solitude as its favorite word.
Uh, I tried again, put a different, um, you know, the same exact prompt, got a different word.
I do find it ironic that an AI machine chose solid.
So I'll maybe, maybe we're moving into general AI it's getting consciousness here, um, and you know, my, my wife's, uh, assessment was it was OK, right?
It wasn't the best essay she's ever written by any means, but it also wasn't unreasonable response there.
So what, what's your reaction to this idea again that, you know, people are going to be passing on AI work as if it's their own.
So that's the biggest kind of downside, if you will.
So, you know, when Going to talk about good, bad, and, and potential with this technology.
I think the notion that somebody might use it to, to, in addition to their own, uh, or to substitute their own writing on their own capability or to hide the fact that this is not them who wrote it, that's a bit of a problem there.
But in general, um, the technology is not necessarily all that new.
It's been used around us in various forms, so news outlets have been using it for, you know, more mundane.
Types of tasks like reporting on the weather or sports analysis or all kinds of reporting about financial outcomes in very you know dedicated, specific, very focused areas.
It's been used before.
The power of this new chatbot, if you will, is that it's, it's open-ended.
It's general.
Anybody can use it for their purposes, and the purposes are multiple.
You can write poems and songs and limericks and essays and just get answers that are comprehensive for understanding.
In a broad array of topics, and it passes like an average individual, right?
So the quality is high.
And because there's a reinforcement learning model behind the scenes, it will also continue to get better and better because people will keep on giving feedback back to the AI, back to the machine, and help it reflect and refine its responses.
So if, you know, your wife has some critique, and if that kind of feedback gets fed back in, you can imagine that it will be able to improve longer.
Determined to kind of close the gap and what we think is excellent responses, not just mediocre.
Wow, so we could even have even better responses that are gonna make it even harder to, uh, you know, determine if it's a fake or not.
You know, one of the interesting things about open AI is this notion that when we have a powerful tool like this, it needs to be broadly in the hands of the population because otherwise this might be, um, a source of power or or the like.
Do you agree with that?
I'm thinking, you know, about what's happening with Twitter and Elon Musk, uh, ownership of it now.
Um, is democratization of the technology, you know, the, the best path forward, or is it maybe the scariest path forward because now it's in the hands of any nefarious individual who wants to misuse this technology?
Well, I think, um, the recent history of the past 2030 years has definitely shown us that the growth and the opportunity lies in the, in the general domain.
So the more we share these tools, the more exposure they get, the more people get to play around with them, the better the quality is across the board.
And so if the technology is to improve, if the quality of the machine is going to improve, and even if the general conscientious usage or or ethical usage is going to be vetted, it's only done when it's in the public domain.
Think about Google and search, right?
If they were to charge, I remember.
At the beginning, there were some debates whether search should be free or we should buy the tool in order to, to, you know, to pay in order to get good search results.
It would have made all the world, we would probably be stuck, you know, definitely 1015 years behind in terms of the quality and the speed of the search that we have today if it wasn't open to the public in the way that it is today.
Um, so I believe that the only way forward is to open it up and for everybody to sink their teeth into exactly what the technology can and cannot do.
Where are some edge cases and what do we as a society have to deal with, and then solutions will emerge, whatever those are.
Do you think at a minimum we need to have, I don't know if it would be a regulatory framework or otherwise, to at least assert like, hey, this was generated by an AI.
That's like to be very, very clear that this was not.
A journalist or a human writing this, this is actually coming from an AI.
Yeah, that's one of the proposals is indeed some kind of disclaimer, if you will, or some kind of alert that everybody can get so they know exactly where things were created, generated, and by who.
And that goes for images, videos, articles that you can see.
When you read an article, you, you look for the author, you look for who created that content.
Um, and if it's generated by this, you know, by any one of these tools out there, there is an idea that that would go a long way to reduce some of the anxiety that we're hearing around us.
And what about the training of these, and in particular I'm thinking about the concerns that as they're training on a data set, uh, things can go off the rails pretty quickly here.
What I'm thinking is we've been talking about the open AI one which is really attracting a lot of attention because it, it is so powerful.
Seems to work so well, but there was another recent example coming out of, uh, meta, uh, on a, on a science bot, if you will, uh, that they had to close down within like 48 hours.
Can you help us with, with what happened there?
Well, like anything else, and we've talked a lot in in our various episodes about biases that exist on data with once you use data to train models, there's going to be biases.
And, um, when the results are too biased.
To misleading in one way or another, at some point you have to go back to the drawing board and fix your model and retrain.
And that's basically what happened to Meta, is that their models were just not ready to cope with more general questions and that the data that they were using and that the the way that their models were fit were problematic.
Um, but this is typical.
We just don't see it as much in the headlines because companies don't always.
Expose their algorithms to the public, but when you have so many people, you know, thousands and thousands of people testing out your technology, you're going to find edge cases, and it's just about how you deal with them.
Uh, but actually, we have an expert online that might be able to give us some more answers, uh, just because, uh, we want to get some, some more clarity on the technical components.
So I'm going to call in or bring in from England, all the way from England.
Uh, Doctor Tom Chatfield.
Tom, are you with us?
I am indeed.
Hello.
Wonderful.
Well, Tom, I'll just introduce you briefly, so I'll tell our listeners exactly who you are, um, in a, in a brief introduction.
So Tom is a British author and a tech philosopher.
I love that definition, a tech philosopher.
Mike, you should have that on your CV.
Um, Tom is, Tom is interested in improving.
Our experiences and understanding of technology, and he's written both textbooks and, um, novels related to how to think and how to really, how do we kind of advance ourselves in, in the digital age.
How do we thrive in the digital age.
Tom has worked with numerous universities around the UK and the world.
He has his doctorate from Oxford University.
And interestingly, during his daytime, he also spends a lot of time thinking and working with organizations that deal with copyrights for various authors.
And so Tom is today's expert.
Tom, thank you so much for joining us.
Oh, it's my very great pleasure.
Thanks for having me.
Uh, Tom, you have thought long and hard and have talked in TED Talks and the like about the ethics of AI and everything that AI and generative AI is doing in In the, in the creative spaces.
So what, what keeps you up at night?
I'm going to start there and then we'll open it up for Mike to see, see more, more questions that he has.
Well, I'm, I'm glad you mentioned that I'm a philosopher, um, because what keeps me up at night is not the kind of details of the tech.
If you like, it's the mismatch between the speed at which human morality and norms and practices evolves and the speed at which technology evolves and develops.
And I think the mismatch between those, um, is a kind of, is a very dangerous gap in which a lot of bad things can happen.
At the same time it's very exciting because, you know, we get to play around with things when they're still molten, when they haven't yet kind of set, um.
This period in time right now, these weeks even, we're, we're seeing the release of a cascade of astonishing technologies.
How are we gonna look back on this moment in 1020, 30, 40 years' time?
There's a lot of stuff that's up for grabs.
The, the future is, is being determined right, right now.
And so I think it's hugely important and terrifying and exciting and amazing.
That we get to chat, you know, it's like you get to sit down in the morning and talk about flight and then in the afternoon refrigeration and then television and radio, and sort of what are these going to mean?
So it's a, it's heady, heady times at the moment to say the least.
Whenever I talk about technology, digital technology in particular, I talk about the exponential growth we've seen in the kind of core underlying technologies including uh processing power, bandwidth and storage and the like.
And though what I always emphasize is if we continue to see that exponential growth over the next decade, the world will be radically different 10 years from now, uh, than it is today.
Tom, how do you, how as a philosopher do you think about that change when you like you were saying 2040 years, but even 10 years out.
It's really hard for us to kind of forecast and foresee what this, what this world of, of generative AI is going to be like.
We've all, I think Yaya and I, as somebody who follow this, have been shocked at how much progress has been made even in the last, you know, 48 months, uh, let alone thinking out 1020 years.
It, it's shocking and impressive, and I guess the two things I'd say are that first of all, as, as I'm sure you know, when you talk to people who work in AI and computer science, they're not as surprised as the rest of us.
Now that might seem really obvious, but it's partly because what we're experiencing now is an intensification of stuff that was, that was done quite a long time ago, you know, deep convolutional networks, deep learning.
A lot of this stuff is decades old or at least a decade old, and what we're seeing is really the way that a difference of, of scale in terms of speed. and complexity um yields a difference of degree in results, so, you know, GPT chat is absolutely remarkable to interact with.
But what it's doing is simply dialing up to 11 stuff that's been done for the last decade, but suddenly there's this breakthrough psychologically where really the Turing test is being kind of passed a million times a day.
Now you interact with it and it really is as good as a human in terms of its convincingness.
But the other thing I'd say which I think is really important is that if you're able to have a fundamental conversation about kind of how these things are done, what's going on underneath the hood, then you can look for these kind of larger patterns and then you can have a conversation.
About the, the ways in which humans and machines are kind of are importantly different.
The ways in which these results are being achieved by machine systems is nothing like the way human minds work.
So I get a little bit obsessed with this future and well what.
Are we going to want humans to do in order to both kind of bring value to the world, but also to interact in a kind of undeceived, rich and successful way with and through advanced machine systems.
So I write and think a lot about skills like critical thinking, about creativity, and about kind of governance and ethics because.
Actually, the values we build into a lot of systems now, the assumptions we bake into them, they're gonna be with us for the long term.
They're gonna be very hard to remove uh in in 20 years' time if we've done things wrong now.
So I guess focusing on this, this underlying stuff is for me a hopeful and important thing to do.
I, I find it interesting that, um, let's call it the average person on the street, which might be like a family member of mine is how maybe I, I measure that is, um, there's a general, I think, negative reaction to this whole notion, right, of generative AI.
And yet, if you point out that you're probably already experiencing it, maybe it's that news article on your favorite sports team or it's something about a company's, you know, annual report release.
Um, they, they quickly seem to be OK with it, um, and, and how, how do you get maybe the general populace to think about these kind of deeper issues that are obviously at the core of this, um, because again, especially in our democratic societies, that's gonna what's gonna ultimately motivate any type of political action that needs to be taking place, uh, to regulate or to monitor these technologies.
I mean I think to some degree, as you say, there's this enormous friction between in theory, how people react to a story, a story about, oh, you know, bad AI is coming for your jobs or good AI is gonna save you, and how we'll react to that storytelling has nothing whatsoever to do with how we deal with technology in our everyday lives.
And I guess the most basic pedagogic point here is that you have to talk in terms of people's.
Of immediate lived experience.
So in terms of someone's work, you're much more likely, if you like, to get the point across to someone about the advantages and hazards of machines.
If you say, well, look, what's the most boring part of your job?
What do you hate doing?
Answering loads of email, picking up the phone, not having information at your fingertips.
What's really frustrating in your life, and people will talk and talk about that.
There will be stuff they love, stuff they hate, and then you can say, well, look, you know.
Broadly speaking, for most jobs, a lot of the stuff people hate is due to tech and systems that are designed with um a lack of alignment for the, you know, the benefit of the people using them, so people working in the gig economy can feel absolutely um oppressed by systems that are trying to maximize their productivity without respecting their rights, by surveillance, by all kinds of negatives at the.
At the same time, people can feel incredibly excited by their ability to connect with others around the world, to find global markets, to work with international teams, to collaborate, to find research, to source parts, to source recipes, and so on.
So in some ways you can start boiling stuff down to the, the human experiences and then talking about the ways in.
Which technologies are implicated in these.
And for me, the big lesson here is not that any technology is kind of bad or good in and of itself.
It's that it has certain kind of values and assumptions baked into its deployment, it's instantiation in, you know, the economy and your life and so on.
So you say to people, you know, don't go out and say oh these are bad, I don't want these or these are good, I want all of these.
Say I want you, my government, my employer or whatever, to try and align the great powers and gifts of technology with, with what I see as, you know, my model of, of human thriving, of a, of a life in which I have agency and respect and meaningful work.
So we don't need to look to the far future to, to see the good and the bad.
We just need to look a little bit more carefully at the present moment.
I love that and.
You know, you mentioned a few things in both questions that you responded to Mike.
First, around the fact that the technology in and of itself is not necessarily all that new.
It's the use cases that we're seeing and the quality of those implementations, which, you know, uh, and your second part about the present.
So we both Mike and I are professors at a business school and we represent the impact on business or we'd like to think about how technology impacts the business world and the business leaders that we train.
And so the need to constantly revisit our values and to think about the ethics of business models and business decisions.
On the one hand, Mike, both you and I are going to encourage our students to come up with business models and new business ideas that leverage some of this technology, but at the same time, our colleagues in the ethics department will have to continue to challenge our students to interpret what ethical dilemmas are today, given the technology today, not given technology that was maybe existed and used or deployed 20 years ago.
But exactly in these kinds of circumstances, how do we come up with new businesses that cut out the mundane, make people still very proud of their place of work, uh, to Tom's point, but also are challenged to consider the ethical implications.
So I think that that's a great, a great kind of way to think about it.
And so Tom, from your perspective, you've worked with authors and copyrights and the whole business model around authorship and creation of, of, uh, you know, literature.
How do you think about the impact on that specific business?
Well, one of my starting points is just that in the year 2022, leaving aside AI, the business of trying to, to make your living or make a predominance of your living through authorship is tough.
It's really tough.
One of the organizations I'm a director at, the um Authors Licensing and Collecting Society in the UK, we, we do an.
Annual kind of survey looking at authors' earnings and they have dropped, um, you know, kind of year on year over the last decade.
You know, authors are making considerably, professional authors on average in the UK are making considerably less than $20,000 a year.
Um, this is the ones who class themselves as professional, and the reason for that is not technology as.
Such, AI isn't writing everything yet.
It's partly economics, it's partly the amount of content that's out there for free.
Um, I think for authors when they look at this, then it's another thing, another way in which the authorship of words seems less likely to bring people money and I don't have any kind of magical or easy solutions to this, but, and it's a really big but.
I think we do need to remember that no matter how brilliant and staggeringly powerful the systems are, at the moment, the understanding that they appear to have is entirely parasitic upon human creativity and content and words.
And so, you know, in a.
Practical way, kind of copyright before everything just gets mined by machine learning systems.
We really need to think quite hard about the authors and creators and indeed citizens who've created the words and the images that can then be mined in their millions and billions for patterns.
Um, and what I want to do as a kind of educator if you like, is just tell people that.
Large language models, these, these models, they are astonishing, but what they are not doing is answering your questions or responding to you in anything like the way a human would.
What they are doing is nothing more or less than coming up with statistically plausible extrapolations from or continuations from the pattern of words that you put into them.
It's a really important distinction.
Um, anyway, I think I'll stop there.
I'm gonna go off on one at this point fascinating.
Mike's, I can see Mike's eyes going like our engines are, are just, well, what comes to mind here is I think, you know, there's been a trend, um, probably going back to kind of the connectivity part of the internet of putting my strategy hat on, you know, lowering barriers to entry in the creative arts, uh, let's say writing as an example.
So now what used to be the, the local newspaper writer who had a geographically constrained audience is now writing articles that can instantly go around the world, and others have observed going back, you know, 2030 years that this is leading to very much a boom bust dynamic, right?
That now if you are one of the elite writers, let's say, your market is huge, your potential is, is almost unlimited, but.
It gets very hard for those kind of further down the pecking order.
And, and this conversation makes me think, is that even gonna be accelerated further to a point that do we even have the opportunity for the, the cream to rise to the top, thinking about like, it used to be those other jobs that might be automated now where the way those writers got their first opportunities before they hit it big.
If those are removed from the economy, do they even become a writer in the first place?
And as a result, we never get to discover their excellence, um, uh, in the world.
Yeah, you can go down some pretty dark paths pretty rapidly here.
And I think unfortunately there's a lot of problems.
First of all, the kind of person who can afford to have a career in the arts, in the performing arts, in writing, and so on is more and more going to be the kind of person who has various privileges in terms of financial security, in terms of connections, in terms of social cachet, and so on.
And then again you're absolutely right, you have this, you know, global connectivity which is a wonderful and a beautiful thing in so many ways, but it also means that a person who already has, if you like, fame or reputation or an audience can, you know, very easily, relatively speaking, become a huge kind of global node and find that audience.
And so when I look at the book lists, you know, what's what's coming out for Christmas, there's lots of celebrities and actors.
Famous people, now they've led interesting lives, great, but of course in a lot of cases, what's happened is a publisher wants to make some money.
They pick someone who's got 2 million followers on Instagram and they say, well, I'm gonna give them a book deal because they're probably gonna be able to convert 5% of their followers into pre-sales.
So we are guaranteed to make money on this.
I'll pay someone else to write their book for them, it doesn't matter if they can write or not.
Now this famous person may be amazing, they, they may be a brilliant writer.
But often they've done great stuff in one area, they're great at football, they can't write, but they've got an interesting story that people want to read, so it goes out there.
And, and the trouble with this exactly is the kind of death of what we call the mid lists of the person who, who has got a tremendous amount to offer but doesn't have the fame for the pre-sales.
And so you are outcompeted wherever you go as an author.
And I guess the last thing I'd say is, you know, at the moment.
I find it wonderful with my with my phone, with my kids, I can kind of make stop motion movies, we can do filmmaking, you know, we can make movies.
I mean when I was a child, not that long ago, that was the province of professionals.
But now, you know, with text to speech and speech to script and script to video, we're getting to a stage where my six year old can say through a speech recognition engine, please GPT chat, write me a.
Script for a movie in which a cartoon unicorn flies to the moon, it can write the script and then stable diffusion or its sequels can animate it into an HD movie.
And and there's literally no human creative professionals involved at all.
That's both democratization and also behind the scenes monopolization of the most extraordinary kind because of course, and this is an elephant in a room, open AI there.
Releasing stuff to the public, yes, but running and owning and training these large models is the province of a tiny number of million or billion dollar organizations, and, you know, a lot of it goes on obviously behind closed doors, so there's a very, very small number of people who own and operate the infrastructure of this apparently highly connected, accessible modernity.
And I'm glad you, you. that because we've danced around a little bit, especially like deep fakes and videos, uh, and this idea again, especially let's say in the political arena, um, that these technologies could be used in so many nefarious ways.
I brought to mind the recent case where there was a fake, deep fake of the video of the Ukrainian president basically, um, you know, conceding the war to Russia, obviously fake.
Luckily it wasn't well done, so it was pretty quickly um.
Uh, you know, revealed to be a fake, but you can think of so many nefarious ways this could be used.
Uh, and again, if it's just out there and anybody can get their hands on it, um, what do we, what do we do?
Yeah, yeah, there's so many TV shows that depict like, uh, you know, a future world.
They're not that.
Long distance future in which these technologies are used to basically change the entire even legal system and and conviction system, yeah.
Yeah, so we're, uh, you know, we're rapidly outpacing science fiction here.
I'm waving my hands around unseen.
And I think just the most important thing, you know, putting my philosopher hat on.
Is that once this kind of thing is possible, it can debase the currency of truth.
You don't even have to be faking stuff.
So let's imagine that someone catches on video a person saying something very objectionable or inflammatory or whatever, but if this technology's out there, they can simply put their hands in the air and say it's all fake.
We've heard this, you think about, you know, US politics, fake news, fake news, it's all fake.
I mean, what does this mean?
In a sense what it claims is that we're living in a world in which because almost anything can be kind of claimed or made up, therefore you can never hold people to account.
So it, you don't even have to be using this tech to be able to use it to sort of make a claim to your followers or acolytes or whoever.
And the second thing I'd say interestingly is that in some ways it puts everybody into a kind of infosec information warfare mode.
So I'm, I'm really interested in that.
I've written a novel, uh, you know, that features a kind of dark web and that sort of thing and.
I'm fascinated by the strategies that people come up with.
So for exa tiny example of how this gets into kind of everyday life, corporate email leaks.
You see that in order to catch people, say at Twitter under under Mr.
Musk leaking potential emails, you send a slightly different email text generated by computer to 1000 different people.
So the meeting room and the floor for your meeting are different for all of them, and that way the person who leaks it is revealed because every email had a kind of unique signature.
So we've already heard people working at places or working around um large language models.
Saying well you might be able to embed a signature in them that only a machine can tell if it's machine generated, a certain sequence of words or letters.
And and what this gestures towards, of course, is exactly this kind of information warfare, mutual escalation thing whereby the provenance and trust and truth is this is this extraordinary kind of battleground and coming up with ways of verifying and authenticating and trusting.
Um, is, is enormously kind of fraught but enormously important, um, and I do think actually this is going to be a really interesting area for, for innovation, um, around some kind of mark that, that shows whether something came from a model or a mind.
Well, I want to take us in a slightly different direction because at some point, Mike's going to ask me, you know, good disruption or bad disruption, and I want to kind of shed maybe some positives, if I will.
I mean, I feel like we've been, uh, skeptical or, or maybe we've been doubting some, uh, or, you know, drawing a very bleak future, but, you know, um, I was reminded of, uh, metaphysics.
I don't know if you get America's Got Talent, uh, over there in the UK.
It's kind of the American version of British Bri Britain's Got Talent.
And in that, um, in the last season, there was a company that basically had singers on the stage, but the technology, the deep fake technology had other faces doing the singing.
So you had Simon Cowell singing, uh, in amazing voices, but it looked like it was him because they basically had his face kind of imposed on somebody else's body.
Um, but then they also brought back Elvis Presley, and that reminded me of all the work that we spoke about, and you've heard Mike, my enthusiasm around.
The Rembrandt pictures or um maybe somebody's going to write the next Shakespeare play, um, bringing back folks that we've lost, bringing back, I mean, I think of my, my dear father, right?
If I could create a conversation, if I could do the whole hallucination, uh, uh, kind of, uh, idea live with somebody who's deceased who's passed and kind of experience that again.
Again, maybe there's some real hope there in terms of humankind and how we think about the future.
It's, it's funny you go there because this is immediately we're going into Black Mirror territory.
So they had a whole Black Mirror episode about someone bringing back her deceased husband who started as a chatbot, yeah.
I don't know if I get hopeful vibes from kind of reincarnating my, my parents as ghosts in Microsoft's basement, um.
But I, I would like to wear a happy hat here.
I mean I think we can't turn back the clock on this stuff.
Um, I would like to be optimistic about, you know, our ability to work with these things, as you say, with extraordinary kind of creativity.
Um, and also not to just take it as inevitable, um, you know, that, that progress in this domain means progress in every domain.
It challenges us in a good way to define what is unique and valuable about human lives and human minds and human thriving and so on.
Um, artificial intelligence keeps on teaching.
Us more and more about human intelligence by showing us how human achievements can be replicated and showing us what can't be replicated so far.
And the most important thing for me is simply, if you like, the incredibly biological nature of our consciousness, the fact that our consciousness is emergent from this squishy gray stuff in our head.
Heads that evolved, that's linked to the rest of the animal kingdom.
It works nothing like a computer, absolutely no resemblance to a computer at all.
We can replicate many of our achievements with silicon, but our wants and desires, our love, our passions, um, our ability to kind of do kind of massively parallel, um, you know, sort of small data processing.
I, when I'm feeling optimistic, I think that we are creating things that remind us how close to nature we are, how close we are to other life on Earth, that we, you know, can't get our ethical problems solved by machines.
They will parrot back to us things, they will help us see more clearly, but why we live, why we want a meaningful work, why we want to.
To watch art to make Simon Cowell sound like he can sing, which is an astonishing trick.
We're thrown back on these questions of why, we're thrown back on these questions of what is human life about.
And of course AI can help us answer them, but its answers will only be as good as the questions we come up with.
Yeah, and I love that, you know, I try to encourage my kids to go out and play, right?
They're teenagers.
I want them to go out and And hang with their friends.
Like, in reality, you know, if you, if we take this conversation, nothing that they interact with online can be as real or as truthful or as passionate as the people that they can actually physically be near.
And so maybe the hope is that we go back to our roots and don't seek all the answers online in this virtual, you know. world of, of the broad wide world of the web, but actually we go back to the roots and find our communities and our people and look people in the eye and give a hug and, you know, do a handshake and kind of build our relationships around that.
So I think it's a good time to, to ask you, Yael, we're at that point in the podcast, in the podcast, good, bad, good disruption, bad disruption, or no disruption?
Definitely disruption.
Um, I believe it's a huge disruption actually.
I think it's, it's going to be as, you know, as instrumental as any search online, as the web itself.
I think this whole area is just going to continue to grow and really shape the way that we learn, live.
And experience.
Um, but I do think that it's going to be a good disruption.
I think it's going to be saving time and focusing our attention on the more important nuances.
It's going to push us all to be smarter and faster and draw much higher level connectivity and come up with new ideas in a way that we haven't been able to do before.
So that's my hope.
Um.
Yeah, I, uh, I'm gonna go a different direction on this.
I can't wait, and, and I'll explain why.
So I'm gonna lean on bad disruption, and, uh, you know, you know me, I'm, I'm usually very pro-technology.
I'm usually very, uh, an optimistic person in general.
Um, I agree with everything that's been said, you know, technology, I always like to say is, is amoral.
It's not immoral, it's amoral.
Um, humans have morality and can put that within.
The technologies we use.
I, I think what I worry about the most is what Tom started out with, is the pace of change and our ability to keep ahead of it in terms of the use.
So I, I love the vision both of you just have laid out of the better nature of, of humans and to be able to embrace technology and evolve it in a way that advances us.
And clearly, you know, we have a long history of being able to do that.
This one worries me that it's moving too quickly and, and that it's going, it has so many again nefarious uses as I keep saying.
Um, that the, the worst angels of our nature, uh, will, will leverage and, and use this technology.
Um, I don't know what to do about it, to be honest with you, because I do agree as well.
It's, it's hard to put this back in Pandora's box at this point.
It is only going to improve and evolve.
Uh, it has geopolitical implications.
Uh, this isn't just like a US based technology, clearly, um, you know, as mentioned, there's many large commercial players who are advancing it, um, so I see it continuing, um, whether.
That ultimately has a good impact on society.
I, I hope you're right, um, but I'm gonna, I'm gonna side on bad just because we're going to end this one with a controversy.
Yes.
Well, my hope, my hope is with the younger generation, um, our students who are even not necessarily the younger generation, our kids, um, our kids' kids, that this generation that is growing up digitally native, aware of some of these pitfalls, understanding the limitations, you know, our generation arguably is all about the hope and the capability.
Oh, we can.
And then the next generation will be how or should we, and they'll be, I think they'll be the ones that I trust will lead us to a much better place.
Oh, in bringing it back to where it began, they're, they're going to actually make Santa real.
Correct.
There you go.
There you bring this, uh, bring this full circle.
Exactly.
Allow us to keep our naivete all the way.
Tom, let me just, uh, extend my, you know, gratitude and thanks for you for being with us.
These are tough issues and I Appreciate your, your perspective, uh, and it's always good to have a philosopher on the podcast here.
Uh, as always, let me thank our producer Gary, our chief, uh, researcher Becky here, and as always, a shout out to Mark Wahlberg.
We're still waiting on that, uh, recording of Good Disruption for us, but I'm sure it will arrive in the new year.
Hey Mike, now we can actually have a bot just record it for us.
That's right.
We don't even need his permission anymore.
Good Disruption is a podcast from the University of Virginia Darden School of Business.