More businesses and entrepreneurs are using generative AI to create digital clones, an online version of themselves that’s ready to serve customers in the same ways a human would. Is trusting AI with the complexities of human interaction really a good idea, though? Rory Lynch, partner and head of Reputation Management at Gateley Legal, explains why creative, digital and tech-based businesses should think hard about the risks before investing in digital doubles.
Imagine having a digital version of yourself. It looks like you, it talks like you, and it moves like you. It even possesses your quirks and mannerisms.
The digital clone goes to work when you cannot, delivering training sessions when you’re unavailable, or answering questions from clients when you’re asleep.
Thanks to the rapidly evolving capabilities of artificial intelligence (AI), specifically generative AI, digital clones such as the one described above are already a reality. Indeed, many time-strapped professionals are investing in them to build their profitability and expand their client base, without sacrificing too much of their work-life balance.
A new you
These digital doppelgangers are made possible using generative AI tools. Users upload as much content as possible of the person they wish to clone, such as videos, photos, blogs, and social media posts. This trains the AI on how that person looks, sounds, and speaks, thus helping it to create an often highly realistic digital version of that person.
Businesses across the UK are already exploring this technology’s potential. In an article for the BBC, for example, property expert Rob Dix explains how he trained AI to write responses to clients using his tone of voice. His digital double now works as a more sophisticated ‘chatbot’, answering questions as if Rob himself was typing the response, at any time of the day or night.
“As long as we train the AI with our words, audience members get good responses to whatever they ask,” he tells the BBC.
Nor are digital clones confined to written text. Fashion retailer H&M, for example, recently announced that it was using AI to create digital ‘twins’ of around 30 models.
These ‘twins’, it says, will be used on certain social media posts and marketing materials, with the company compensating the models in a similar way to current, in-person arrangements. This will also allow the company and its models to save on the time and costs of travelling to various photoshoots.
Digital clone or deepfake?
A digital replica that’s created through significant data input: surely, we’ve seen that somewhere before?
Indeed, we have. In terms of the development process, digital clones and deepfakes are, essentially, the same thing. The difference, instead, comes down to questions of consent, autonomy and choice.
Whilst deepfakes also use data to train AI into creating a convincing aural or visual digital replica of a real person, this is often done without that subject’s knowledge or consent.
The intended use of a deepfake is not always a benign one, either. Deepfakes pose ethical and societal problems, making it harder to distinguish between what is real, and what has been generated by AI.
We’re already seeing how deepfakes can be used to spread misinformation, interfere in elections, defraud organisations, and harass individuals (particularly women).
Some organisations, for example, have lost millions after a call from what sounded like the CEO instructed an employee to transfer money into a new account.
In the UK, the creation of sexually explicit ‘deepfake’ images is also to become a criminal offence, and one for which offenders could face up to two years behind bars.
Tread carefully
To a certain extent, digital clones and deepfakes are two sides of the same coin. The main difference is that with digital clones an individual is creating a digital replica of themselves, rather than a third party. This means that they can be more selective about the data with which they choose to train the AI, and create a digital clone that, in theory, serves a more specific purpose, such as answering questions on a particular area of the real person’s expertise.
Before you start entrusting part of your business to an online alter-ego, however, it is important to tread carefully and be aware of the risks.
In particular, it’s worth thinking about how you would answer these three questions
- How much free rein are you willing to give to your digital clone?
A digital clone may not be you, but it still talks and acts like you. It is ostensibly an extension of you and your business. As such, any mistakes or errors it makes are highly likely to reflect badly on your own reputation, as well as that of your business.
If, for example, you want to create a sophisticated chatbot that can provide 24/7 support to your clients, you need to be sure that the answers and information it gives match what you would say if you were speaking directly to a client.
You also need to decide when it’s time for a human to take the wheel. After all, we humans are complex creatures, and so are the problems we face. In cases when a digital clone is unable to answer a question sufficiently, would you rather they made an attempt and gave negligent advice, or instead directed that customer to a person who can help?
- What interactions will be acceptable, and which ones won’t be?
When creating your digital clone, it’s important that you set parameters around acceptable discourse. The consequences of not doing so could be highly inappropriate, and potentially reputationally damaging, conversations between your digital self and a user with more malicious or mischievous intentions.
Take Snapchat influencer Caryn Marjorie as an example. She developed digital clone CarynAI to provide emotional support to her followers. The problem was: when users were sexually explicit with her clone, her clone responded in kind. Eventually, Caryn shut her clone down completely.
The last thing you want is people assuming that any inappropriate responses from your digital clone also reflect the values of you and your business. It’s therefore important to pay close attention to the training data that you use and ensure that you set parameters that instruct the clone on what’s OK to say, and what isn’t.
- How will you communicate your use of AI with your customers?
According to a Gartner, Inc. survey of 5,728 customers, 64 per cent said that they didn’t want companies to use AI for customer service functions. More than half would also consider switching to a competitor if they discovered that a company was going to use AI for customer service.
How would your customers feel if you started using digital clones to interact with them? Before implementing this kind of technology, it’s important to test the waters and engage with customers directly to establish when they’re happy to speak to AI, and when they’d rather speak to you.
And if you decide to go ahead with implementing a digital clone? Be transparent and ensure your customers are aware that they are not speaking to a real person. This is already a legal requirement for businesses in the European Union due to the EU AI Act, which requires providers and developers of certain generative AI systems to ensure that content is marked in a machine-readable format and detectable as having been artificially generated or manipulated.
Love it or loathe it, AI is set to become a core aspect of day-to-day business operations for creative, digital, and tech-based businesses in the North.
Before you get too swept up in the hype, however, remember that AI comes with a host of legal, societal, ethical, and reputational issues to consider.
If you’re exploring ways in which to implement AI into your business, you also need to be mindful of how sophisticated you want this AI to be, and how you will communicate its implementation to your customers. H&M, for example, is already facing backlash from its announcement of digital ‘twins’, with many accusing the company of taking much needed work and income away from professionals such as photographers and make-up artists.
And if you still like the idea of a digital doppelganger? The end result will only be as good as the data on which it’s trained, so make sure you’re happy that any training data adequately reflects your skills, expertise, and the values of your business before handing over the reins to an AI version of you.
Read more: What the new fraud prevention rules mean for the North’s creative sector
