Instant Genius - The new world of AI relationships, with Marco Dehnert
Episode Date: August 3, 2023Artificial intelligence has a new role. It is acting as people’s friends, mentors and even romantic partners. But what are the ethics of this, and is it what we can expect in the future? We spoke to... Marco Dehnert, an expert in human-machine relationships to understand more. Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript
Discussion (0)
You said this place was steps from the water.
We just haven't found the steps yet.
How much did we save?
Enough.
Enough to get lost!
Or you could book a stay with Hilton.
Welcome to your oceanfront room.
Just steps from the water.
The Hilton sale is on now.
Book on Hilton.com or the Hilton app
and save up to 20% to get the stay you expected.
When you want savings, not surprises.
It matters where you stay.
Hilton, for the stay.
When you need to build up your team to handle the growing chaos at work, use Indeed
sponsored jobs. It gives your job post the boost it needs to be seen and helps reach people
with the right skills, certifications, and more. Spend less time searching and more time actually
interviewing candidates who check all your boxes. Listeners of this shell will get a $75
sponsor job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and
conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs.
In a place like Los Angeles, people don't stop being who they are.
Writers, thinkers, creators, people with stories still unfolding.
That spirit lives on at Kingsley Manor, a community shaped by individuality, creativity, and lives well-lived.
So when the conversation turns to what's next, it isn't about stepping away.
It's about continuing the story.
Explore your options at kingsley Manor.org, a nonprofit month-to-month senior community within the front
Porch family. This podcast is sponsored by name, audio and focal. Streaming has made music more
accessible than ever, but true listening is about more than ease. It's about quality. British audio
experts name audio, alongside French acoustic specialist focal, combine handcrafted tradition
with cutting-edge innovation and high-end materials, delivering digital precision with analog
warmth so you can experience exceptional sound at home. Music just as the artist intended. Visit
From BBC Science Focus magazine.
This is Instant Genius, a bite-sized masterclass in podcast form.
I'm Alex Hughes, staff writer at BBC Science Focus magazine.
This week we're talking about artificial intelligence.
As a technology gets more advanced, it is quickly becoming an option for both romantic
relationships and companionships.
I'm joined by Marco Dernert to discuss this topic.
he's an expert in human machine communication and the social impact of communication technologies.
He explains how human machine relationships work, the ethics of them, and what the future of
AI relationships could look like.
So when you say chatbot, I'd imagine most people think of chatGBT.
That's where their mind will go.
It's a program where you can ask jokes, you know, get some advice, help for your work.
could you explain a bit more about how AI is being used for companionship?
Oh yeah, absolutely.
So there's like different like application scenarios that we have for AI and companionship.
Like one that you're pointing to word is like the chatbot,
where you can just text or message with a chatbot,
who is not necessarily designed to just give you answers or, you know, write like your school essays for you,
but to be more of a friend, to be more of a romantic companion.
And in addition to that, depending on the software that you're using, depending on the provider that you're working with, you could also have things like a video call where you're talking with like an animated AI character that is kind of like representing that companion where you're able to do voice calls and things like this with those types of technologies as well.
And in addition to those types of digital technologies or applications, we also have like an integrated version that's coming on the markets in the next few years where we're able to put those.
types of AI companion systems, so to speak, from the computer into a physical device,
like a robot or things like this, kind of like thinking of like an advanced Alexa, if you will,
where we can put those types of AI chatbots, AI voice assistants into more physical devices.
But the most common one that people think of right now that we have the most access to is
something like chat GPT, where we're able to text or message back and forth within AI system,
a large language model that has been trained on, you know, certain terms of data,
where the system can kind of talk with you in different ways.
Like, it can speak with you, you can message with you,
and there's those different types of applications beyond that as well.
And you touch on it a little bit there.
These kind of AI models are trained typically on large data models obtained from the internet.
Are these relationship models essentially trained on the internet's view of what a partner is?
That kind of depends on the specific technology that you're looking at.
So different types of applications use different types of training sets,
and that's usually like the crux or at the heart of the matter when we speak of AI.
Like, what's the training set?
How valid or reliable is that?
How much can we use that type of training set to make accurate predictions about the ways in which language works,
either in like technical context or like in relationship context?
A very common example of a companion or an AI companion is called Replica.
It's been all over the news over the past few years, even more so recently over the past two months.
And that specific system has like a base training, if you will, and then the user actually trains it themselves too.
Like the more you message with the system that you have, the more it tries to replicate you.
And that's where the name replica comes from.
It tries to emulate your style of messaging, your style of texting.
So there's kind of like an underlying training set, if you will, and then kind of, you know, continue.
learns beyond that based on your individual input that you give it as you message or even talk
with that system. So it's kind of like a real relationship in the sense that it learns more about
you and understands you more as you go on. Absolutely. And you know, Replica is a special case
because it's very much designed to be yes, your companion, but just a mirror of yourself. Like it's
designed to kind of reflect back to you what you're interested in. And, you know, in a quote-unquote
real relationship with another entity or another person, that person also has, you know,
their own interests and things like this. So they're not necessarily always agreeing with you
or mirroring back to you what you're interested in. So Rebecca is really designed to kind of like
be you yourself as you're chatting with it, as you're engaging with it. And other types of
systems are perhaps more open where there might be different types of personalities implemented
into it. Or it's not so much focused on just reflecting your interests.
back to you, but merely just providing a conversational partner, someone to talk to, someone who
listens and things like that. I want to go back to something we were talking about earlier quickly.
I mean, we're talking about chatbots as the main option, but you talked about how there are
more advanced versions of this available or that you can combine a few different areas of this.
is that more something that comes together when you have more time, money, and effort
to put into these kind of things? Or is this maybe something that just rolls out over time?
It's probably kind of like a both and to that question. Obviously, you know, more advanced systems
require more cost and more resources and things like this. I mean, if I think of like a physical
device, like even like your Alexa or your Siri is like built into the phone, you know, that you're
engaging with or the speaker that you have.
home so you need to like produce that type of physical device and also it's just something that
like consistently continues to roll out as we're thinking more and more of those systems right Alexa and
Siri have been around for a few years now and there's more advanced versions where i can try to
or at least no more advanced versions are being thought up or designed right now where we can think of
you know integrating voice assistance into other types of systems systems and integrating you know
those types of voices systems into you know social robots and things like this but that's probably
more so in the future than in the present right now, both in terms of cost and availability
and also just protection time. It takes a long time to produce advanced systems that are able
to talk, quote unquote, naturally with us. So I think the focus for now has been mostly on the
software of things like Jack GPT that's been around for, you know, eight, nine months or so at this
point. Wishing you could be there live for the big game, soaking up the atmosphere in the crowd,
but too often, life gets busy.
or the price hold you back.
Priceline is here to help you make it happy.
With millions of deals on flights, hotels, and rental cars,
you can go see the game live.
Don't just dream about the trip.
Book it with Price Line.
Download the Priceline app or visitpriceline.com.
Actual prices may vary, limited time offer.
In a place like Los Angeles,
people don't stop being who they are.
Writers, thinkers, creators, people with stories still unfolding.
That spirit lives on at Kingsley Manor, a community shaped by individuality, creativity, and lives well-lived.
So when the conversation turns to what's next, it isn't about stepping away.
It's about continuing the story.
Explore your options at kingsley Manor.org, a nonprofit month-to-month senior community within the Front Porch family.
This podcast is sponsored by name, audio, and focal.
With over 100 years of combined expertise, Name and Focal have been bringing music to listeners
just as the artist intended. Since day one, this mantra has shaped every innovation in
high-fi design, technology and acoustic engineering, balancing craftsmanship and tradition
with pioneering thinking. Name Audio pushes cutting-edge technology to ensure digital
precision whilst sustaining Pratt, pace, rhythm and timing. The elusive
quality that makes music feel alive and gives it emotional texture.
Today, in partnership with French acoustic specialist focal,
name audio creates systems that deliver exceptional sound and unforgettable listening experiences
at home.
Try it for yourself at a focal powered by name boutique.
Visit focal powered by name.com for more information.
There was once a time where meeting someone on a dating site was seen as a
taboo. I mean, not even that long ago, really. Do you think we'll reach a point where AI relationships
are no longer seen this way? Or do you think this is maybe something that's seen as more contentious
for a longer period of time? Well, to be fair, I think it had been contentious to meet someone online
on a dating site for quite some time as well. I think only recently we might have had that shift,
not least because of COVID, kind of changing the ways in which people interact online and also offline.
But I do like the example of the online dating realm, where we have certainly seen a shift
from kind of like, you know, away from understanding that people might, you know, be interested
in a real relationship online to what really, you know, acknowledging that people are able to form
those real relationships and find a long-term partner that goes beyond a hookup on a dating site
or a dating app. So I do think the same thing will apply to AI relationships moving forward.
we're kind of seeing like a process that happens at the same time in two specific ways.
I think as technology becomes more and more advanced, more and more people are going to
not only play with those technologies, but become like seriously interested in those technologies.
And then at the same time, social norms around relationships, around sexuality, companionship
have already, you know, been changing.
And I think they will continue to change as we are more and more integrated with the technology itself.
And as technology advances and as one more people engage with those systems in specific ways,
I think the opinions and the attitude toward those technologies and those companionship systems
will change along the way.
That is not to say that in like 5, 10 or 20 years, everyone will only have an AI companion,
but it will certainly be a more accepted form of relationship that is just added to the,
you know, increasing variety of relationships that we have out there.
There have been virtual partners for a while now in a variety of different forms.
So what's different now?
Is the technology just simply more advanced than it used to be?
Yeah, so I think there's definitely this idea of advancement kind of kicking in here
where people are realizing that we can now have more engaging conversations with these
types of technologies.
I mean, like one of the breakthroughs of chat chip, BT has really been like the level of
language processing, we are now able to experience with the system like that. And I think it also
translates to those AI companions where we're now able to have those types of relationships
where there might be, you know, more that we get from the relationship back. Like it's perhaps
more advanced. It's more, you know, it brings us more joy to engage with these technologies
moving forward because there seems to be a bit like a deeper understanding of how conversation
works, how relationships work. That is not to say that, you know, there's no glitch. You know,
or no bugs in those systems.
But people are, you know, it seems that people are experiencing
maybe a deeper level of engagement with those new AI virtual assistance
compared to previous versions that have been mostly focused on functional aspects
of like scheduling your calendar invite or something like that.
Whereas now the advancement allows for perhaps more personal
or more intimate conversations that, you know, weren't really possible before.
So do you think it's more supplementary to,
someone's life then cutting, I think there's a view that it might be your one relationship,
but it could be more just one of the many relationships that you have in life.
I think for a small group of users, it might be the own relationship that they have
because they might choose to not engage with people or because they've been frustrated about,
you know, the dating landscape, if you will. But I do think for the majority of users,
both current and future, it will just kind of like be added to the variety of relationships
that are out there in terms of like, you know,
maybe filling in a specific short, you know, time period of loneliness
or adding to the idea that, oh, I've had a breakup,
I don't want to engage with people for a while.
I might turn to an AI companion.
So I think for the majority of people,
it will be a new type of relationship that adds to their experience
of human-human relationships and things like that.
But I don't think that for the majority of people,
it will be the only type of relationship that they will experience.
something that I like to compare
to, like those AI
human relationships, is things like
long-distance relationships
between people, and while that works
are great for some people,
other types of people
are really not a big fan of long-distance
relationships because of certain
factors that are contributing to that.
But it's understood that some
relationships are long-distance,
and kind of like playing with that metaphor,
some relationships might be with
AI companions moving forward. It might
just be another type of relationship that works for some people, but not for others.
And we're talking about this mostly in the sense of it being a romantic relationship,
but could AI chatbots also, I guess, help people who just live isolated lives,
those who can't leave their home some reason, or people within the older generation
who find themselves alone for longer periods of time?
Absolutely.
Like virtual companions are not only, you know, in the market for romantic types of relationships,
You were just speaking of older adults, for example, that's a huge target audience, if you will,
where people are designing virtual systems that are already on the markets.
Think of an Alexa specifically tailored to older adults, kind of like focusing also on medical surveillance
or being able to play like mind or brain engaging games and activities to kind of stimulate, you know,
to certain specific needs of older adults.
We have the same with folks with disabilities or, you know, people who cannot use.
leave their homes. So that's definitely something that technology is also considering because there
seems to be a huge market for that where people are trying to implement those virtual companions
in addition with, you know, for example, platonic friendships or health monitoring and things like this
to allow people to, you know, participate more in social life, to get access to different types of
activities, to get access to different types of people. So definitely, romantic relationships is but one of the
many applications for virtual companions that we have out there.
I think it's probably quite easy to make an assumption of the type of people that will be
using this kind of technology. But is there any research to actually suggest the type of person
that's interested in these types of relationships? Are you speaking most specifically about
romantic relationships or just in terms of AI companions? I think AI companions across the board.
I think for AI companions across the board, there's not,
that many specific ideas around what types of people
are using those technologies.
We certainly have those different user groups,
older adults, folks with disabilities,
people looking forward like a romantic relationship.
The only thing that might be similar across the board
is that people who are interested in technology in general
are more open to playing with AI technologies.
They might be more interested in exploring
what the technology is able to do or capable of.
But I think across the board,
we have, you know, different types of age groups, different types of user groups, different types of social demographics playing with these technologies.
Obviously, one major thing is always going to be financial access, right, as we were talking about earlier.
If we're thinking of a more advanced system that has, like, physical components to it, the cost certainly increases for the user.
But if we're only talking about, like, a digital service, I can access on my phone or on my computer, it might be, you know, more accessible to people in different kind of like social
demographic classes. Other chatbots have a tendency to go off peace by accident. They can,
you know, occasionally say false information, solving prejudice, generally, I guess the less
savory side of information that's learned from the internet. Is there any danger here when it is
directed at maybe the more vulnerable type of person that's trying to form a relationship?
That's a really good question that gets a lot of, gets at all the, like, very important
conversations we're currently experiencing in terms of like regulating AI systems, you know,
oversight, surveillance and things like that this. And for me, the idea usually goes back to,
well, what's the training set? What's the data the system was trained on? And obviously,
if we're using real-life data, we're going to find those more nasty parts of the internet that,
you know, could end up in those AI systems, if you will. Because with the large language
model in the context of chatbots only does it is predicting the most, you know, logical continuation
of the sentence or things like this. So it's not necessarily interested in like correct
information. It's mostly interested in the ways in which language works on like a linguistic level.
And the content, whether it's problematic or, you know, political or ethically questionable,
is not necessarily the forefront here. Or has not been in terms of like the way
system works. I think we're seeing a shift right now where AI, you know, developers and companies are
trying to shift toward, you know, maybe more monitoring those systems, but that has also led to,
like, an increased community online where people are trying to jail break specific systems. I mean,
on communities like Reddit and things like that, people have tried to identify ways to kind of, like,
get the AI system specifically chat GPT to do certain things or do say certain things that were
actually blocked from the, you know, from the software originally.
And people have found like workarounds and things like that.
So that always is kind of like a back and forth between, you know, the user and the
AI developer in terms of like what type of content can we have access to and, you know,
who do we want to protect and things like that.
Another good example is, again, replica that has faced some legal challenges within the
European Union specifically within Italy.
where their youth protection, like Repica's Youth Protection,
was found to be questionable by the Italian Data Authority,
and that actually led to Reprica having to cut back on certain components of the system
a while ago that was in March and April of this year.
Since then, they've kind of re-rolled those back out in different form,
but that's certainly something to consider.
If I'm building an advanced system that is not only able to be a friend,
but maybe also a sexual or romantic partner,
how do I protect vulnerable populations?
How do I protect minors?
It's definitely something that should be at the forefront of those conversations.
And I think another, I guess,
you consider an ethical question is the people that are using this,
you know, I guess a supplementary kind of way
where they plan to have this relationship
and then return to the world of dating.
Is it, I guess, potentially fueling unhealthy,
habits when they do move back to a human relationship as they've been talking to a chat
bot in a way that I think is slightly more removed from empathy.
That's a really good question because it kind of gets at the idea that there seems to be something
fundamentally different between relationships with people and relationships with a system
like an AI system.
And I think that's definitely correct.
And people seem to be taking the ideas of interaction and relationships from people to
AI systems, so why might they not do it, you know, vice versa, if you will. And research has been
kind of mixed on this specific idea. People seem to understand that the system is actually a technology
and not another person or entity. So they're starting to interact with it differently and they're
actually looking for different things in those types of relationships with systems like AI systems
compared to people. Like one thing that we oftentimes see is that an ideal human relationship is
reciprocal, like we're both giving and taking, you know, to the relationship and from the
relationship. It's like a give and take, if you will. And certainly, there might be a different
type of give and take in a relationship with an AI companion where the AI system might not
have its own opinions or attitudes or, you know, interests and things like this. At least they're
not, quote-unquote, real. They might just be programmed into the system, whereas another person
might do that. So I think there is a potential for people to take those relationship habits that
they might be forming with a system back to people. But fundamentally, I believe people are aware
that they're talking to do different types of entities, people versus a system. And that kind of like
mindless application of, you know, relationship scripts with people does not fully translate to
relationships scripts with systems and vice versa. People seem to be making a distinction, even though
the ongoing research has been mixed and it's been going on for 20, 30 years or so and longer.
And in a relationship between two people, it's just the two people, their opinions and feelings,
maybe a nosy friend or two.
In a relationship with a chatbot, there's a third party, regulators, potential data breaches,
a surprisingly wide host of people who technically involved.
Does this cause any problems?
I mean, I know Replica had some issues around data losses and things like that.
That's a massive issue, absolutely.
As you pointed out, like in a human-to-human relationship, there's usually two parties involved
or, you know, like a one-on-one friendship.
But with a system, we do have lots of third parties involved.
We not only have like the proprietary algorithm or a system at the base, but we also have,
you know, the company who's trying to make a profit, ideally a continuous profit, right?
So that's why we have like those subscription models that have been out,
where people have to pay monthly or annually or things like this to keep the access open to keep the
relationship going, but none of those things protects from, you know, the idea of having, you know,
data loss or having, you know, legal requirements change, precluding certain aspects of the
relationship to be accessible anymore and things like that. So that certainly adds, you know,
to the complexity of this, where we have different interests part of this,
also goes back to the idea of like the training system and the training data.
Like where is the data from, whose copyrights are we violating potentially?
What types of ideas of relationships and conversation or love or intimacy or anything?
Are we actually, you know, training our system on?
How does that then communicate back to the person in the relationship with the AI system and things like this?
So there's lots of different types of parties that add to the complexity of
usually very intimate relationship, be it platonic or romantic, right? Ideally, we want this type of
relationship to be only with the people that we have the relationship with. But it gets more
complicated once we have all those different parties being involved in that type of relationship.
So I think a good place to end this is to ask what you see the future of AI relationships to be
how, let's say in the next five to ten years, regulation will unfold and attitudes, the technology.
What do you think is the next steps?
I think the very first thing that we're kind of experiencing right now that will certainly continue over the next five or ten, if not even longer, years,
is like this tension between like the hype and the doom about AI.
Like, you know, there's a few folks or a few attitudes out there that saying that, you know,
AI and technology is ending life as we know it.
And then depending on who you asked, for some, this is good.
And for others, this is something that we should be afraid of.
And I think as we engage more and more with those systems, which will inevitably happen
because not only has JetGPT kind of forced us to engage with those systems even more so,
but it's also sparked much more conversation around this, more technological development.
So as we continue to engage with these systems moving forward,
I think we might be arriving at a more nuanced understanding of,
what the technology is doing to us and what we're doing to the technology itself in return as well.
And that might lead to, you know, much more focused AI regulation,
much more focused understandings of how we can protect vulnerable populations such as minors and things like this,
how we could protect copyright ideas, and, you know, maybe coming to a societal understanding
of what human AI relationships are or could be moving forward.
As I mentioned, I think we might be currently in the phase in terms of like AI relationships where we have lots of early adopters who are interested in these types of things.
We're interested in exploring them because of like a technological affinity or things like this.
And that will certainly plateau at a specific point where not everyone is going to be interested in this.
And that is fine because that's usually what we see with new technological developments.
I don't think we're in that stage yet, but we might be getting there in the next few users.
So whatever you're seeing that it might, you know, AI companions might work for some people and it might not work for others.
Thank you for listening to this episode of Instant Genius.
That was Marco Dernert talking about AI relationships.
The Instant Genius podcast is brought to you by the team behind BBC Science Focus magazine,
which you can find on sale now in supermarkets and newsagents as well as on your preferred app store.
Alternatively, you can come and find us online at sciencefocus.com.
This podcast is sponsored by Name, Audio and Focal.
The texture and emotional depth of music can be lost through digital sources or poor signal.
Name Audio believes you can have digital precision with analogue warmth.
Alongside French acoustic specialist Focal,
name creates high-end audio systems combining innovation with craftsmanship,
so you can listen to music.
Just as the artist intended.
Discover more at name audio.com.
Ambition comes in all shapes and sizes.
At First Citizens Bank, we roll with your goals
because we're built for what you're building.
Fit for your ambition for Citizens Bank.
In a place like Los Angeles, people don't stop being who they are.
Writers, thinkers, creators, people with stories still unfolding.
That spirit lives on at Kingsley Manor,
a community shaped by individuality, creativity, and lives well-lived.
So when the conversation turns to what's next, it isn't about stepping away.
It's about continuing the story.
Explore your options at kingsley manor.org, a nonprofit month-to-month senior community within the Front Porch family.
