Supporting human and artificial intelligence co-working

This piece is naturally speculative, but speculating about a far closer potential future than we recently thought likely. A future where human workers extensively use artificial intelligence (AI) tools to support them in nearly all tasks and where a number of tasks – currently performed by humans – are no longer exclusively human provided or even majority human provided. Whilst this future is by no means certain we must now acknowledge that human realistic outputs are easily generated in a growing number of areas and that recent AI advances increase the likelihood of future disruption.

Where are we now?

Artificial intelligence came to widespread public attention with the initial release of ChatGPT in November 2022. ChatGPT is a chatbot based on a large language model (LLM) and its development is widely seen as accelerating the current AI boom. It is certainly the fastest growing consumer ‘app’ in history, gaining 100milion users in its first two months. The LLM technology behind the chatbot has facilitated a plethora of rivals with Google, Anthropic and Meta releasing similar chatbots.

At its core, a large language model is a statistical tool which predicts the next word in a sentence. Often, it does this incredibly well, sometimes it does not. This ability allows it to generate human like language, code, images and more. The Turing Institute define generative AI as: (generative) create new content, (AI) automatically with a computer program.

Whilst this content is not perfect – or not necessarily even very good – we need to query what the ability to quickly and cheaply produce good enough human-like content means for our economy, society and skills development.

Even now – if nothing else changes – everyone in the economy has a simple choice, do we pay for human excellence or are we happy with free artificially generated content. Of course this has always been the mechanism, we traditionally pay more for higher quality but now we must choose in many more circumstances.

Does a manager decide that a marketing campaign doesn’t need to be perfect, but an AI generated one will do? Does a business owner decide that a free customer service chatbot – which occasionally makes mistakes – is preferable to a costly customer service team.

To be honest I doubt it, at least right now, the technology is still a way a way from being consistently good enough, and it is not necessarily bespoke enough to slot neatly into the variety of business models which make up our economy. However, both examples above are happening and undoubtedly readers of this article have experienced customer service chatbots and AI generated advertisement content.

What is happening now?

Rather than wholesale replacement of roles and teams, we are currently seeing AI as an efficiency increasing and time saving tool. LLMs may well not live up to their original hype but they are certainly having an effect. Organisations appear to be shifting from freelancers to AI for one-off or none core projects, with both wages and job postings falling on freelance platforms. A recent survey noted that 70% of surveyed US managers considered that AI could do the jobs of interns, 69% said AI can do the work of a recent college graduate and 78% said they planned to lay off or not recruit because of AI.

In his book, Co-Intelligence: Living and Working with AI, Ethan Mollick suggests that a key danger is AI use replacing the talent pipeline in organisations and in the wider the economy. We are seeing hints of this in the above. The immediate and long-term implications could be significant. Our talent pipelines are long term investments, difficult to change course and need an end-point. Simply for an individual to commit to a course of education and training, taking several years and costing significant amounts, there needs to be a high likelihood of a positive outcome for them.

On another note, LLMs as we noted earlier are not perfect. We need experts to check their outputs, as time goes by the number of experts decreases due to AI impacting the long term talent pipeline.

Navigating this new world?

The AI environment is evolving rapidly, whatever system you are using right now is likely to be the worst one you will ever use. As AI evolves, its impact on the economy will change and it is very difficult to plan for this. The Resolution Foundation describes the impact of AI on the world of work as radically uncertain, saying that this future is unknowable in a way that rules out even envisioning some of the possible outcomes. In short, we cannot even imagine all the impacts of AI, let alone predict what is most likely.

However, skilled humans will likely be needed, many studies so far suggest that a skilled worker supported by AI performs better than an AI alone or all but the most highly skilled workers. The benefits of AI will likely come from expert humans using an AI system, rather than an AI replacing a human worker – or indeed AI taking over routine and mundane tasks to free up human creativity – let’s hope that this future gain is not sacrificed for short term efficiencies or settling for just good enough.

By Dr Richard Whittle, University Fellow at the University of Salford’s Business School.