💼🧑🤝🧑 Working with an army of clones: The future of productivity with AI
The first episode of the second season of Black Mirror is called “Be Right Back.”
It tells the story of Martha, a young artist whose husband Ash tragically dies in a car accident. While she is in mourning, a friend who has been in a similar situation tells her about a new technology that promises to “clone” her husband using artificial intelligence.
If someone approaches you at the worst moment of your life to tell you that you can talk to your loved one again, I think no one would say no.
All she has to do is send all the texts, videos, and audios she has of her husband to the company. Completely devastated by Ash's death, Martha decides to try it.
A few days later, she receives an email.
The email leads her to a chat with Ash.
There he is. Your husband (?) who just died. With the same jokes, the same typos, the same personality. Is it your husband?
I won’t spoil the rest of the episode because it’s worth watching. I’ll just mention that the thread of the episode revolves around the authenticity of these types of “people.”
This episode aired in February 2013, 10 years before we all went crazy with ChatGPT.
What seemed like science fiction is now (almost) a reality.
Tools like Seance AI allow you to provide key information about a deceased relative (such as their personality and writing style) to generate a chat “from beyond.”
I say “almost a reality” because the tool doesn’t fool anyone.
It doesn’t look much like me.
But it’s close to being able to create my “digital clone.” I think the only thing it lacks is more information about me (I could only provide 200 words about how I write).
Creating a digital clone
If we wanted to get serious about this project, we would have to consider four aspects: content, voice, appearance, and movement.
Regarding content, I want my clone to express ideas similar to mine. To know my story. And to respond articulating words as I would.
The elegant way to do this is through a complicated process of “fine-tuning,” where you retrain ChatGPT with your information.
But there are simpler ways to achieve this. Tools like Botsonic allow you to upload documents, PDFs, presentations, and links to create a custom Chatbot without needing to know how to program.
Here I’m passing links to my blog to train the bot to write like me.
Let’s assume we manage to train the model. Now my clone needs to sound like me.
This is easier than it seems.
There are already several tools that specialize in “cloning” voices. ElevenLabs is one of the most well-known. You provide a short sample of your voice and then write what you want it to say, and it does it in your voice.
This week, Meta surprised everyone with a similar model, but better and faster.
You provide a 3-second recording of your voice, and it can clone it. On
there are several examples worth listening to.
Good. We now have a clone that speaks and sounds like me. But it’s still not “visible.”
A Google tool called Dreambooth allows you to train a model to generate images of you.
Imagine photos of you instead of that cute dog.
And the final touch: movement.
Yes. There are already tools for that, like
,
, or companies dedicated to this, as shown in
.
This leads us to the final question: why would anyone want to create a digital clone?
For me, this is one of the most fascinating applications of this technology, but it also raises the most ethical challenges.
Let’s start with the good.
Leverage
Entrepreneur, angel investor, and Indo-American philosopher Naval Ravikant says that to make a big impact with your work, you need leverage.
It’s impossible to become a millionaire or change the world if you don’t seek forces that multiply your work.
Leverage, graphic description.
According to Naval, there are four ways to leverage our work:
Labor: the oldest form of leverage. If you can get more people to work for you, then you can achieve bigger things. This can happen through coercion (a pharaoh leveraging slaves to build pyramids), leadership (Martin Luther King leveraging supporters to change the racial history of the USA), or paying wages (a business owner leveraging their employees to sell products/services).
2.
Capital:
money can transform into leverage opportunities. The most obvious way is through hiring people to help you do more (as we saw in the previous point), but capital can also transform into investments (for example, a machine that allows you to multiply your output) or shares in a company (which then generate much higher returns than your investment).
Was anyone visionary enough to have bought NVDA a while ago? I wasn’t :(
Media: here we can go beyond traditional media (which is undoubtedly a source of leverage to reach many people). Today you can write an article like this one and, if all goes well, make your ideas reach thousands or even millions of people, continuing to generate value long after they were created.
The famous Mr. Beast video replicating Squid Game has been viewed nearly 500 million times. Record once, distribute to millions of people.
Code: something similar happens with software. You can write a program, create a website, or develop a game, and replicate it practically infinitely at no additional cost, transforming code into a very powerful source of leverage.
The game Among Us, created by Marcus Bromander, has been downloaded more than 500 million times on the Android Play Store.
In general, leverage through labor and capital is more cumbersome and complex. The former requires leadership and coordination skills, while the latter has associated risks.
Leverage through media, particularly social networks, and code, are the most powerful today. It’s no coincidence that the new under-20 millionaires are content creators (YouTubers/TikTokers) and genius programmers (like the creator of Among Us).
But thanks to generative artificial intelligence, a new form of leverage is emerging: digital clones.
You could have an army of robots scattered across the internet leveraging your work:
An email or Slack assistant that does a first filter, responding to low-priority messages as if you were you and organizing the important ones for you.
A work team to carry out your projects: while one clone conducts market research, another is drafting the first version of a report, while a third is cleaning a database for analysis.
If you are a very busy person, you could have several copies of yourself available to hold virtual meetings with people who want to ask you questions.
But opening the door to digital clones also brings several challenges.
Consent and Ownership
If I am a sufficiently public person, anyone could create a digital clone of me.
All it takes is to take the articles from my blog, train a model on them, and you can write a completely new one with your clone of Matías.
Or take the videos of HolaSoyGermán, clone his voice, appearance, and movement, and start making new videos.
His videos are public. And I created his digital clone. And the new videos I generate are mine. But it doesn’t feel legal, does it?
The music industry has already shown us that we are not prepared for digital clones when a song by Drake and The Weeknd completely created by AI appeared, causing annoyance among artists and an ongoing debate in the industry.
Do we want clones of anyone to be possible? Or do individuals have the right to decide what can and cannot be done with their digital personalities?
Scams
I don’t know about you, but I’m already used to receiving at least one scam call a month. I answer, and someone tells me that a relative is in trouble and that I have to pay to fix it. Generally, I hang up within the first 5 seconds and go on with my day.
But it would be very different if I answered and heard my mom’s or brother’s voice.
That happened to this grandmother in Canada.
If there is any video of you on social media where your voice is heard, then there are scammers who can use it to clone your voice with the tools we discussed earlier.
Some scammers have gone a step further and have also cloned their victims' appearance and have conducted their scams via video calls to be more convincing. And with good results: a Chinese businessman was scammed out of USD 600,000.
Misuse
Having an army of clones at your disposal allows you to do extraordinary things, both for good and for evil.
This is a video of the calls I received on a particular day last week.
All are calls from companies that want to sell me something. And this is without AI. Imagine the harassment power that companies will have when they start using digital salespeople.
Other examples of misuse:
Misinformation and fake news: imitating the voice and style of trusted individuals to make false content seem credible.
The first episode of the second season of Black Mirror is called “Be Right Back.”
It tells the story of Martha, a young artist whose husband Ash tragically dies in a car accident. While she is in mourning, a friend who has been in a similar situation tells her about a new technology that promises to “clone” her husband using artificial intelligence.
If someone approaches you at the worst moment of your life to tell you that you can talk to your loved one again, I think no one would say no.
All she has to do is send all the texts, videos, and audios she has of her husband to the company. Completely devastated by Ash's death, Martha decides to try it.
A few days later, she receives an email.
The email leads her to a chat with Ash.
There he is. Your husband (?) who just died. With the same jokes, the same typos, the same personality. Is it your husband?
I won’t spoil the rest of the episode because it’s worth watching. I’ll just mention that the thread of the episode revolves around the authenticity of these types of “people.”
This episode aired in February 2013, 10 years before we all went crazy with ChatGPT.
What seemed like science fiction is now (almost) a reality.
Tools like Seance AI allow you to provide key information about a deceased relative (such as their personality and writing style) to generate a chat “from beyond.”
I say “almost a reality” because the tool doesn’t fool anyone.
It doesn’t look much like me.
But it’s close to being able to create my “digital clone.” I think the only thing it lacks is more information about me (I could only provide 200 words about how I write).
Creating a digital clone
If we wanted to get serious about this project, we would have to consider four aspects: content, voice, appearance, and movement.
Regarding content, I want my clone to express ideas similar to mine. To know my story. And to respond articulating words as I would.
The elegant way to do this is through a complicated process of “fine-tuning,” where you retrain ChatGPT with your information.
But there are simpler ways to achieve this. Tools like Botsonic allow you to upload documents, PDFs, presentations, and links to create a custom Chatbot without needing to know how to program.
Here I’m passing links to my blog to train the bot to write like me.
Let’s assume we manage to train the model. Now my clone needs to sound like me.
This is easier than it seems.
There are already several tools that specialize in “cloning” voices. ElevenLabs is one of the most well-known. You provide a short sample of your voice and then write what you want it to say, and it does it in your voice.
This week, Meta surprised everyone with a similar model, but better and faster.
You provide a 3-second recording of your voice, and it can clone it. On
there are several examples worth listening to.
Good. We now have a clone that speaks and sounds like me. But it’s still not “visible.”
A Google tool called Dreambooth allows you to train a model to generate images of you.
Imagine photos of you instead of that cute dog.
And the final touch: movement.
Yes. There are already tools for that, like
,
, or companies dedicated to this, as shown in
.
This leads us to the final question: why would anyone want to create a digital clone?
For me, this is one of the most fascinating applications of this technology, but it also raises the most ethical challenges.
Let’s start with the good.
Leverage
Entrepreneur, angel investor, and Indo-American philosopher Naval Ravikant says that to make a big impact with your work, you need leverage.
It’s impossible to become a millionaire or change the world if you don’t seek forces that multiply your work.
Leverage, graphic description.
According to Naval, there are four ways to leverage our work:
Labor: the oldest form of leverage. If you can get more people to work for you, then you can achieve bigger things. This can happen through coercion (a pharaoh leveraging slaves to build pyramids), leadership (Martin Luther King leveraging supporters to change the racial history of the USA), or paying wages (a business owner leveraging their employees to sell products/services).
2.
Capital:
money can transform into leverage opportunities. The most obvious way is through hiring people to help you do more (as we saw in the previous point), but capital can also transform into investments (for example, a machine that allows you to multiply your output) or shares in a company (which then generate much higher returns than your investment).
Was anyone visionary enough to have bought NVDA a while ago? I wasn’t :(
Media: here we can go beyond traditional media (which is undoubtedly a source of leverage to reach many people). Today you can write an article like this one and, if all goes well, make your ideas reach thousands or even millions of people, continuing to generate value long after they were created.
The famous Mr. Beast video replicating Squid Game has been viewed nearly 500 million times. Record once, distribute to millions of people.
Code: something similar happens with software. You can write a program, create a website, or develop a game, and replicate it practically infinitely at no additional cost, transforming code into a very powerful source of leverage.
The game Among Us, created by Marcus Bromander, has been downloaded more than 500 million times on the Android Play Store.
In general, leverage through labor and capital is more cumbersome and complex. The former requires leadership and coordination skills, while the latter has associated risks.
Leverage through media, particularly social networks, and code, are the most powerful today. It’s no coincidence that the new under-20 millionaires are content creators (YouTubers/TikTokers) and genius programmers (like the creator of Among Us).
But thanks to generative artificial intelligence, a new form of leverage is emerging: digital clones.
You could have an army of robots scattered across the internet leveraging your work:
An email or Slack assistant that does a first filter, responding to low-priority messages as if you were you and organizing the important ones for you.
A work team to carry out your projects: while one clone conducts market research, another is drafting the first version of a report, while a third is cleaning a database for analysis.
If you are a very busy person, you could have several copies of yourself available to hold virtual meetings with people who want to ask you questions.
But opening the door to digital clones also brings several challenges.
Consent and Ownership
If I am a sufficiently public person, anyone could create a digital clone of me.
All it takes is to take the articles from my blog, train a model on them, and you can write a completely new one with your clone of Matías.
Or take the videos of HolaSoyGermán, clone his voice, appearance, and movement, and start making new videos.
His videos are public. And I created his digital clone. And the new videos I generate are mine. But it doesn’t feel legal, does it?
The music industry has already shown us that we are not prepared for digital clones when a song by Drake and The Weeknd completely created by AI appeared, causing annoyance among artists and an ongoing debate in the industry.
Do we want clones of anyone to be possible? Or do individuals have the right to decide what can and cannot be done with their digital personalities?
Scams
I don’t know about you, but I’m already used to receiving at least one scam call a month. I answer, and someone tells me that a relative is in trouble and that I have to pay to fix it. Generally, I hang up within the first 5 seconds and go on with my day.
But it would be very different if I answered and heard my mom’s or brother’s voice.
That happened to this grandmother in Canada.
If there is any video of you on social media where your voice is heard, then there are scammers who can use it to clone your voice with the tools we discussed earlier.
Some scammers have gone a step further and have also cloned their victims' appearance and have conducted their scams via video calls to be more convincing. And with good results: a Chinese businessman was scammed out of USD 600,000.
Misuse
Having an army of clones at your disposal allows you to do extraordinary things, both for good and for evil.
This is a video of the calls I received on a particular day last week.
All are calls from companies that want to sell me something. And this is without AI. Imagine the harassment power that companies will have when they start using digital salespeople.
Other examples of misuse:
Misinformation and fake news: imitating the voice and style of trusted individuals to make false content seem credible.
