Where are all the social use cases for AI?
In the race to build AI applications that help ‘me’ instead of ‘us’, we are overlooking what might be its most valuable potential.
The arrival of sophisticated large language models has ignited a race for companies seeking to find the most valuable applications of their current and emerging capabilities. From homework help to software coding, companies are building – and discovering – new ways to help users accomplish a vast range of things. One of the most popular use cases over the past few years has been companion apps and AI friends of various kinds which some studies suggest might help us feel less lonely and thus help solve the growing loneliness crisis in society. However, solving loneliness with AI friends is a bit like trying to solve hunger with fast food. It might fill some stomachs but relying on empty calories isn’t going to be healthy in the long run. And in some cases, over-reliance on AI friends has been alleged to be downright disastrous.
The problem with these and so many other use cases, is that the value of AI is framed in terms of its potential to imitate and stand in for human beings. This overlooks the much bigger potential AI has to address the loneliness crisis and add value to people and businesses; not by replacing human contact, but by helping us to connect more meaningfully with each other. Modern AI systems have a set of critical and often overlooked capabilities that make them uniquely suited to augment the natural social powers that humans already have.
There is a tonne of untapped potential here for both tech companies and society more broadly, but many stakeholders seem to misunderstand AI’s social superpowers by being overly focussed on solutions built for “me”, rather than “us”. AI is supposed to be your personal assistant, your intern, your coach, your co-pilot, and now increasingly also your friend and your confidante. We see this individualistic AI focus across many sectors of the economy, from the latest investments in consumer gaming all the way to the focus on individual workflow automation in enterprise software. This is perhaps rooted in a much larger implicit individualism in Western societies in particular, which historically has influenced the product development mindset to be about “jobs to be done” rather than wider, systemic concerns. But just as we went from the solitary early days of Microsoft Word to the now collaborative, easy fluidity of Google docs, we need to do the same with AI.
At ReD, instead of artificial intelligence we often talk about collective intelligence. We believe this is the most significant emergent capability of current LLMs to date – the ability to draw on humanity’s collective wisdom to give you the most ‘normal’ or ‘average’ response possible, without the eccentricities, performances and judgemental gaze of a human being. The AI wave has proven that when it comes to humanity, “normal” is often great. And this is where AI’s superpower lies: it can explain or instantly communicate what “normal” looks like in a given context.
Solving loneliness with AI friends is a bit like trying to solve hunger with fast food. It might fill some stomachs but relying on empty calories isn’t going to be healthy in the long run.
If normal has a slightly uninspiring ring to it, that’s kind of the point. By drawing on humanity’s collective wisdom, AI has the power to strip away individual biases (even if it may reinforce our collective ones) and arrive at the most common sense approach. Here, we mean common sense literally as in the sense which is most common in its training data, not the everyday common sense that humans have which lets us navigate novel situations. The ability to have a computer at your fingertips that lets you understand how ‘most people’ would navigate a situation, perceive an action you’re contemplating, or answer a question is an enormous superpower in a world where we are increasingly connected to each other and dependent on the people around us.
While sociologists have been studying collective intelligence for decades, the proliferation of AI tools and platforms now affords individuals easy access to tap into that collective intelligence. At ReD, we have spent years studying how people relate to AI, as a third entity, distinct from their relationships with other humans or classic technology. On the one hand, we have observed how many people feel the LLMs of today frequently fall short when it comes to helping individuals navigate complex social dynamics. While on the other hand, some of the most meaningful experiences people have enjoyed with AI are not just about time-saving or finding quick answers, but rather when it augmented their abilities, like developing new skills and enriching their working processes.
The collective intelligence of modern AI systems means it can, if shaped right, help us better understand and connect with each other across language, culture and context barriers (e.g. what is the best way of getting my point across to this kind of person or group?). It can help groups of people reach consensus or find common ground (e.g. what is the fairest resolution across the group?). It can open up new forms of collaboration (bridging what is normal in your group with what is normal best practice). It can even potentially play roles of convener, instigator, space maker and encourage serendipitous and meaningful human connections in ways that humans never could.
If normal has a slightly uninspiring ring to it, that’s kind of the point. By drawing on humanity’s collective wisdom, AI has the power to strip away individual biases… and arrive at the most common sense approach.
It is time for AI to truly go multiplayer. We are already seeing some movement in this direction, as more and more new products and models are built not just for “me” but for “us”. For example, in June of this year, Anthropic, the company behind the Claude AI assistant, announced an update to its platform adding new features to improve team collaboration and productivity while OpenAI, also in June, acquired the videoconferencing startup, Multi, and last month launched Canvas, a way to write and code collaboratively and track changes in ChatGPT.
In the short term, there is tremendous untapped opportunity in multiplayer AI. In the long term, building AI that leverages collective intelligence for social connection also helps us imagine how we might harness “Artificial General Intelligence” or “Strong AI”, what many tech companies from Microsoft to Meta have set as their north star. To date, much of the optimism about AGI has centered on the potential breakthroughs it can drive in the hard sciences – see for example, CEO of Anthropic Dario Amodei’s, recent blog post. But just as the true power of AI lies in its social use cases, we see AGI’s transformative potential also belonging to the social domain, in its ability to not just improve the lives of people, but improve relations among them.
Ultimately, if we can relate to AI not as a replacement for other humans, but as a tool to tap into collective intelligence and better connect and collaborate with other people, the potential benefits on society can be truly transformative.