Whilst other forms of AI will still drive the bulk of the business impact, generative AI will remain the hotly debated topic. All the big players have now put their first cards on the table, though new challengers keep popping up left, right, and centre, such as recently the French start up Mistral. This Cambrian explosion of foundation models will continue, but as a result it becomes crowded at the top of the leaderboards, with differences between each getting smaller and smaller. The paradox is that this increased competitive pressure leads to commoditisation.
What is the lesson for enterprises? The more models that come out, the less you need to worry which one to pick. Ignore Large Language Model Fear Of Missing Out or anxiety and start building use cases, run pilots and scale out using whatever generative AI services are good and adequate for now. Companies are busy experimenting with the use of generative AI to create marketing content, summarize customer service calls, unlock domain-specific knowledge via chatbots and generate workflows and apps. As you long as you have a centre-out architecture with your own generative API, you can swap out any time to the model or service that is the generative flavour of the moment.
Generative AI is Gaining Agency
But can we also grant generative technology more autonomy, so that it can make plans and take actions on its own, in a safe manner? It won’t be long before we can link generative AI-driven agents to all forms of internal and external data and knowledge sources. This will enable them to formulate plans, gather data and take actions from start to finish. Human intervention remains essential to monitor and manage the AI, approve plans, as well as to establish clearly defined parameters regarding what AI can and cannot do. But this is the key next step all major labs are working on.
Generation AI is entering the workforce
So if generative AI is maturing in terms of agency, who is going to take it out for a walk? To understand how transitions work, we should not just look to generative AI researchers, but perhaps also listen to philosophers of science, such as Thomas Kuhn. He argued in the 1960s that science doesn’t evolve incrementally, but through major paradigm shifts, and these revolutions have less to do with technical inventions and more with the influx of new generations of researchers.
The same could happen in enterprise AI. Generation AI will be entering the workforce – the generation that wrestled themselves through the final years of study or their first-time consulting gigs using generative AI tools, and spent their free time filling up their socials with AI generated content. Their frame of reference is ChatGPT, MidJourney or TikTok – not business applications like Siebel or Cobol. They bring expertise, an open mind set and high expectations about AI use within companies based on their personal experiences. And as a purpose-oriented generation, they oppose irresponsible and unethical use of AI, instead focussing on creating value for both the customer and the business, without getting lost in technology. Instead of worrying about AI taking over their jobs, perhaps people should wonder if their work will be taken over by a new starter who knows how to use AI.
Happy generative new year
So what can we expect for 2024? Yes, even more generative AI model releases, but the focus will shift to specific, generative AI driven use cases, providing value for both the customer and the business. Development and scaling will be conducted using central governance, and a low code, iterative approach to building these features quickly and learn from them. Further innovation will come from giving generative AI more agency, and a fresh generation of AI-first employees will be ready to co-create and adopt generative AI solutions.
Peter van der Putten is Director Pega AI Lab and Assistant Professor at Leiden University
