When asked to think of instagram influencers, you probably picture gorgeous people wearing virtually nothing or perfectly-polished teens dancing to the latest TikTok trends. It’s unlikely that you’ll think of virtual ‘life-forms’ that are near-perfect as a result of some clever programming.
The future is closer than you may think, and we might just live to see the likes of SkyNet become a reality if things go much further (Terminator reference, for my less nerdy readers). Influencing seems to have taken a step into the future with the rise of Artificial Intelligence (AI) capable of replicating human social behaviour and placing itself into the real-world. Machine learning has reached a point whereby the tasks they perform are no longer boring and repetitive; instead, AI is learning to possess unique quirks and realistic biases, making these virtual characters increasingly passable as individuals who act on their own free will¹.
It seems that, in qualifying as a virtual influencer, the influencer must first be computer-generated (obviously), but they must also sit within that uncanny-valley of emulating the ‘look’ of something human. Lastly, they must be able to do what all good influencers do — real or otherwise — and control/affect people’s behaviour and decision making¹. Those three things combined are a recipe for success, with the ultimate determination of an influencer being how well they can ‘sell’ themselves or a product. Some virtual influencers have gotten so good at emulating human behaviour that they now have online cliques, feuds and relationships, just like your favourite human influencers. If we were to continue at this rate, we may not have the need for living influencers anymore. But how does it all work, where does the money they earn go and is the use of AI influencers sustainable?
In 2018, Miquela Souza (@lilmiquela) was named as one of the top 25 most influential people on the internet by TIME’s magazine. What’s crazy is that Miquela isn’t even a person — she’s a first-of-her-kind AI influencer² created by Trevor McFedries, Sara DeCou, and the team at Brud, a transmedia studio³. The character has been programmed in such a way that she recognises her programming and the fact that she is AI, but continues to assume an otherwise human identity. In the years since her conception, Miquela has had considerable success endorsing a number of real-world brands like Prada, Calvin Klein, and Samsung, paving the way for a host of other AI characters to enter into our world.
Whilst still a new field of study, one paper found that there was no difference between human and AI influencers when it came to an individual’s desire to follow or how personable they found them to be; this goes to show that consumers found just as many relatable qualities in an AI emulation of human life as they did in the real thing⁴. In some cases, especially in consumers who held greater value in uniqueness, AI influencers were shown to have greater persuasive power than humans when it came to influencing the purchasing of products⁴. Another study found the considerable influence that AI personalities had over human behaviour to, again, be evident in their ability to commit transgressions that led to the degradation of the brands endorsed, just like in real-world influencers who also make mistakes. If an AI personality does something disagreeable, it reflects negatively on the brands they represent and this can often result in a reduction in sales. Except in this instance, a transgression in the behaviour of one AI influencer appears to tarnish the view that people have on ALL AI influencers, showing that we may not be as accepting of AI personalities as first thought, with many people still perceiving them to be less individual than their human counterparts (as we wouldn’t start blaming all celebrities for the wrongdoings of just one)⁵.
So, despite the notion that AI influencers lack individuality in the eyes of some, they do show promise. It’s quite a frightening prospect, but these uncanny emulations of human life may hold a place in our society one day — they may already have a greater influence than you think, with audiences worldwide already exposed to other AI simulations of human life in TV and gaming. Surely it won’t be that much of a step toward a completely controlled, artificial approach to all media, especially when you consider that AI influencers are somewhat of a commodity; they never age, can be in countless places at once, are fully controlled by their programmer, and prove to be significantly cheaper in terms of hiring cost. Perhaps we’ll see them putting their less efficient, more expensive human counterparts out of business one day?
An interesting thought…
The idea of AI influencers is cool, but it raises a perplexing environmental idea. Is the carbon fingerprint (digital carbon footprint) associated with these influencers more or less than that of their human counterpart? Factor in mobile phone use and the travel that countless influencers make to shoot content and we have ourselves an interesting question. Of course, there is insufficient research to irrefutably answer this question as AI influencers are somewhat of a new phenomenon, but let’s be logical about this; these AI internet personalities can literally be ANYWHERE in the world at the push of a button, they don’t have to travel anywhere, as long as the creators have all the digital content they need to overlay their characters onto a real-world background. They also don’t use mobile phones, since technically, the algorithm controlling these AI influencers doesn’t need one to access the internet; it is already on the internet because it’s part of the internet! The two things are mutually exclusive, one would not exist without the other and this would cut out the need for end-user devices and the carbon emission associated with their creation and subsequent charging. That being said, there is a HUGE carbon fingerprint associated with the programming of Artificial intelligence, especially those trained in Natural-language processing (NLP), which is a prerequisite for any AI influencer. Emissions associated with this computer learning process are likened to that of the oil industry, emitting more than 283,000kg of CO2e, five times greater than the emission created over the lifetime of a standard American car⁶. So how does the combined emission associated with the teaching of all AI influencers compare to real-world influencers and their use of the internet, technology and travel? I guess that’s a question for the future!
So what do you think? Are AI influencers the way forward or are they a bit too unreal to wrap your head around?
: Jalan, A. (2022) What Are Virtual Influencers and How Do They Work? Available at: https://www.makeuseof.com/what-are-virtual-influencers/ Accessed: 24/05/2022.
: Robinson, B. (2020) Towards an Ontology and Ethics of Virtual Influencers. Australian Journal of Information Systems, 24. https://doi.org/10.3127/ajis.v24i0.2807
: Blakiston, L. (2021) Lil Miquela and the rise of robot influencer. Available at: https://thespinoff.co.nz/society/15-07-2021/lil-miquela-and-the-rise-of-the-robot-influencer Accessed: 24/05/2022.
: Sands, S., Campbell, C. L., Plangger, K. and Ferraro, C. (2022) Unreal Influence; Leveraging AI in influencer marketing. European Journal of Marketing, ahead of print. https://doi.org/10.1108/EJM-12-2019-0949
: Thomas, V. L. and Fowler, K. (2020) Close Encounters of the AI Kind: Use of AI Influencers As Brand Endorsers. Journal of Advertising, 50(1). https://doi.org/10.1080/00913367.2020.1810595
: Strubell, E., Ganesh, A. and McCallum, A. (2019) Energy and Policy Considerations for Deep Learning in NLP. 57th Annual Meeting of the Association for Computing Linguistics, Florence, Italy. https://doi.org/10.48550/arXiv.1906.02243