Opinion AI is now the influencer, not humans

It’s like a river permanently changing course.

Information flows in two ways: Direct flows, from sender to receiver (like a phone call), and diffused flows, that spread organically (such as viral memes). This is because information contains interpretive weight. The person who receives it can decide what it means, change it and redirect it.

According to the Two Step Theory, gatekeepers, those typically with greater influence, visibility, money and power, set this weight, filter it and direct it. As per the Diffusion Innovations Theory, information flows in an S-shaped curve, such as through social media networks. This also runs on a hierarchy, (let’s say a key reporter posts breaking news but only those whom he follows can DM him for an update). In Network Theory, information flows across individuals and relationships. According to Agenda Theory, platforms decide which agendas rise (as with X), and Actor Network Theory (ANT) constructs meaning dynamically between humans and non-humans (documents, facts). These information flows often fit the flow of society in the way that a river fits the landscape. Think of the Gutenberg press that took power from the few, such as the Church, and redistributed it, enabling the neighbourhood pastor to print his sermons. Thus, a changed information flow has the capacity to change the social landscape.

While Two Step and Diffusion Innovations have been the typical gatekeepers, since the social media disruption, information flow has been shifting into Network and ANT. AI takes that a step further by decentralising the nodes from a purely human (and thus subjective) agenda to force a human to non-human interaction. Which makes AI the influencer.

An example of this is the Studio Ghibli image trend that recently went viral with OpenAI (and subsequently other LLM agents) enabling user capacity for it. This was more than a spontaneous burst of joy. It was a disruption of the traditional power centres of information flow. Hence, some of the anger surrounding it, couched as a disregard of copyright, and disrespect of the original creator. This isn’t strictly true. You wouldn’t hesitate to put Totoro on a birthday cake, a Halloween costume, or on a school bag. And Japan invented cosplay virality as a tribute to the creators of anime. In other domains, singers have been using autotune for a while, digital art is its own domain, and computer-aided animation is effectively covering up for a lot of shortcomings. Exotic filters for cameras exist (including those that make the frame anime-like). So, why did users transforming beloved family photographs using the Ghibli Effect matter so much? One difference is the amplification of scale. The second is that tech companies seem to be profiting off such development (though OpenAI opened the innovation to all free users as well). No one who used it was claiming to be as talented as Miyazaki — most were probably just gaining pleasure from replicating and personalising his style. The disruption here is that algorithms and bots interacting with the mass of the public makes an established hierarchy fluid, if not redundant. This rearranges the social trust factor. The power balance is shifting.

Are you more likely to believe an LLM, or less? Your answer would likely have changed from when you first interacted with an LLM agent a few months ago. As LLMs are constantly being improved, people are able to interact with them, ask questions, clarify, feeling less judged, less influenced by bias and opinion, and more open to correction. Top-down models are being rendered obsolete. This is the power of such interactive network flows. Even if deepfakes are created, the democratisation of flows permits counteraction from other models.

Typically, two factors permit breaking social hierarchy: Marriage and education. AI allows those who haven’t had access to these to become upwardly mobile. It is becoming clear, as with the Lively-Baldoni case, that it is not some mythical merit or skill that helps one prevail, but rather access and influence. The pushback is now by AI as an actor within the networks, co-shaping the nodes, controlling algorithms, and capable of toppling influential narratives. This is why it matters that the philosophical framework of AI is leaning towards equity. AI is the influencer, and it turns out that those with pedigreed credentials are not necessarily the ones most talented at wielding it. Hierarchy has now shifted to society at large.

The results of these disruptions are multiple fold and still unravelling. You can no longer claim to be trustworthy on the basis of having spent a certain number of years working in a domain, or earning a number of degrees from expensive universities, or your network of influence, friendships, family and pedigree.

A recent innovation by the MIT lab ran 500 neural networks to create real-world audio notifications for laughter on wearable spectacles for the hearing impaired, as well as for hearing a horn, a doorbell, or a pressure cooker go off. As such technology becomes commonplace, the deaf can participate in everyday sociability. Hopefully, the disadvantages of disability will be evened out to some extent.

For such things, you don’t even need technological expertise. It is another example of the survival of the fittest, a comfort with change and an ease with the new social order. The adaptable will inherit the earth.

Das is an author, therapist, AI ethicist and co-founder of a global AI philosophy focus group