Some Thoughts on Generative AI
By Nigel Paine
Demis Hassabis the CEO and Founder of Deep Mind, the Google owned AI company based in the London Tech Hub Kings Cross, reminded us in a recent pod cast on the BBC, that we should not see AI as something invented last November but an area of intellectual exploration for at least the last 60 years.
In his view generative AI is a staging post in a long, complex journey to machine intelligence. He claimed that even the founders of Chat GPT were astonished by its success as they could, like all the experts, only really see its limitations. (The Bottom Line, BBC Radio 4 released on 12th October 2023) But somehow Generative AI caught the public imagination and foregrounded the power and apparent nous of large language models that had been quietly digesting huge amounts of information at an industrial scale in the background that could be put to use helping co-pilot (to use a Microsoft phrase) work and life.
The fact that everybody had access to what appeared to be a remarkable new technology, and could revel in its astonishing capability, took this from specialist interest to suddenly becoming headline news, and the headlines were not all positive. The British tabloid newspaper The Daily Star (think National Enquirer and go down market ) claimed on the 4th October this year in a surprised tone that: “Psycho scumbag chatbots aren’t ALL bad” because these “Big-brained wrong ‘uns will extend our lifespan…and allow us to work a 3-day week” unless, and here is the punch line, “they don’t murder us first”.
The invisible power of LLMs is therefore, at once, a brilliant aid to thought, organisation and creativity as well as a threat to our humanity! Gen AI was picked up by everyone, but fear aside, it was used by all sectors of industry and use cases began to spill out. Learning & Development was no exception, where their main focus became content and course development.
Each profession looks at the most obvious usage first, rather than engaging, profoundly, with what this kind of technology could offer across the entire workflow. We tend to grasp at the obvious, and ignore the profound.
L&D revealed its obvious preoccupation when it grabbed at an opportunity to speed up content development. Much in the same way, software developers demonstrated the ability of Generative AI to generate code or at least a first draft of code very quickly.
It is interesting that Hassabis, when talking about AI in that podcast, decided to focus on its ability to aid and support humans. He even referred to generative AI as a "co – pilot”at one point. This takes us straight to Microsoft, whose implementation of large language models has been generically branded as ‘Co-pilot’ and the company has brilliantly embedded the ChatGPT foundation model into Microsoft products across the board. Whether you are writing an email or building a presentation, Co-pilot is there to help you do the task better and faster,
What this means is that Microsoft generative AI is designated as your assistant across the workflow. Co-pilot sits quietly on the desktop, and can be brought to life by voice or keyboard to support whatever application you happen to be working in. It can, for example, check your emails for tone, summarise your Teams meeting or pull out the action points. It can provide some substantial evidence to back up an argument or enhance a paper you happened to be writing.
It is always there, like a spell checker on steroids, and ready to be invoked whenever needed. Very soon, it will help arrange your calendar, remind you about key tasks or meetings. It will be, as if every team member were able to access an executive assistant whose role was to make your life smoother and easier.
At this point the real issue is, not does this work, but how can we make it work as effectively as possible in our organisation. If this isn't a primary task for the L and D team, I am not sure what is!
Just as an huge industry developed to integrate IT into the workflow, and get up to speed with basic IT applications, there is now a new opportunity to deploy the skills necessary to establish generative AI inside workplaces. The really big opportunity for L&D is not content development, but the ability to empower the entire organization to transform work, and the workplace.
Harvard business school recently published a fascinating article that explored in detail how one company (Boston Consulting Group) deployed AI tools for its consultants and discovered that there were astonishing productivity gains to be made amongst knowledge workers (up to 22% in some instances), but it was not a universal benefit. Indeed, in some instances, AI was positively misleading and sent consultants down rabbit holes that were distracting and time wasting. This is why they used the metaphor of a jagged frontier to describe their results.( “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects if AI on Knowledge Worker productivity and Quality” Harvard Business School Working Paper 24-013 October 2023). The progress and advantages were unevenly distributed.
The obvious question is who will help organizations navigate this jagged frontier? Who will support the integration of AI into the workflow? If the role of L&D is not close to the tip of your tongue at this point, it has pretty much made itself redundant. There is a need to bring new knowledge in, and share that information, whilst working across the organization on collective projects and programs.
If this can be made to work, a true revolution in organizational knowledge management will emerge. Just as the current operational structures emerged in the early part of the 20th century as a way of creating consistency and structure in the work place and to build competence, and efficiency in the workforce, this time we need a new workforce to be productive in a different kind of workplace. This will be completely transformative and revolutionary. Organizations remake themselves for new contexts all the time. This time is no different. There is an opportunity to be seized.