In L&D, we’ve seen some examples – applied with varying degrees of success. AI can now track your learning activity and recommend future development options; AI powered chatbots can coach individuals. In some cases, AI powered tools are supporting learning designers to develop new modules and build learning interventions.
But the big buzz has been about Generative AI – tools like ChatGPT.
The difference between earlier models of Machine Learning and AI is that Generative AI seems to talk to us. It can use unstructured data – text, images, video – spot patterns and predict associations. It will report back on what it has found in ways which resemble the output of a real person.
This represents Big News - according to McKinsey some $12bn has been invested in the development of generative AI in the first five months of 2023 alone.
So what does this mean to organisations and the people they employ? As learning and development teams, how do we help prepare for and engage with this new technology landscape? The first point to make is that ChatGPT and it's like is not the finished product. These are marketing tools, produced to showcase what can be achieved. They demonstrate new capabilities to get people thinking about the future uses of Generative AI.
When Generative AI really gets to work will be when it has been trained on in-company data sets in order to automate specific activities which currently require human interaction. Traditionally we have thought about automation replacing routine jobs undertaken by relatively low skilled – and low paid – workers. This trend has been reversed for some time before the rise of Generative AI.
When a Japanese insurance firm automated its claim handling over 10-years ago, they invested the equivalent of £1.4 million in the software. They gained a return on that investment in less than 12 months by reducing the staff in the claims division. To make these sums add up, the average wage of the people who were displaced was in excess of £47,000 per year.
It will be the skilled, moderately or very highly paid jobs which will be automated if this kind of technology investment is to make sense in the short to medium term.
The first thing for L&D teams to consider therefore, is which jobs, or parts of jobs, could, should or might be automated? The second consideration, is ‘if these activities are automated, what will the people who used to do them, do instead?’
I don’t have a crystal ball. The degree to which AI - and specifically Generative AI – will impact work is as yet unknown. However, predicted capability of Generative AI may impact many millions of jobs, especially those involving working with knowledge.
Despite the uncertainty, I think there are four things we can do now.
I would start by helping people to understand what Generative AI is. While some will have a very good idea of what it is and what it can do, others will not have moved beyond the social media hype and the apocalyptic news reports. They need to see it in action with informed guides.
This awareness raising is the forerunner to a more informed discussion of what might come next. Those creative knowledge workers in our organisations – whose jobs may change the most - are probably as well placed as anyone to define the kinds of routine, repetitive or resource intensive jobs they do now which they would (perhaps happily) have some assistance in making quicker and/or more efficient. Identifying where automation delivers joint benefit – freeing up staff to do more interesting and rewarding work as well as enabling more output with similar inputs – seems like a good place to start. It also seems to be one of the areas of decision making which is best not left to IT consultants with ideas of what could be done, but limited insights about what should be done.
Prepare people to work alongside Generative AI. Some of this will be about mitigating the negative effects of AI. There have been well evidenced examples of Generative AI providing answers which sound correct but which are based on inference. As OpenAI’s Chief Technical Officer has declared, ChatGPT and other AI tools generate answers which sound reasonable but may be factually inaccurate. What’s more, depending on the data set which the AI has been trained on, there is the potential for inaccuracies to be amplified. There will be a requirement for checking and quality control of AI generated outputs. This may reduce over time. After all, machine learning is continuous and one would hope it will, therefore, get more accurate the more it is put to work.
EU Commissioner Margrethe Vestager has also warned about the potential for Generative AI to entrench, and even amplify, existing societal discrimination. This is especially a cause for concern when it is used to make decisions based on precedent. We have already seen AI embed existing biases in the sentencing of those in court, of misapplying facial recognition and determining whether hand driers recognise black skin. Decisions made in future should be subject to human scrutiny, rather than an unquestioning acceptance of existing norms which may be unfit for purpose.
‘Computer says No’ could move from a mildly amusing commentary about inflexible decision making to a real world downside of derogating responsibility to the data sets on which these tools have been trained.
Effectively, we should prepare people to use Generative AI as a tool, to work alongside the judgment and experience of real people. Generative AI is more than a sophisticated search engine. Its capability to enter into a discussion, to answer follow up questions (such as ‘Are you sure that’s right?’) is one of the break through features of the technology. People need skills to question, check, and manage the outputs it provides.
Above all, in the aftermath of automation there will be work activities which Generative AI cannot do – or at least tasks it will not be trusted to perform by employees, customers and service users. These will be the skills which will be most urgently required during and beyond any transformative introduction of automation. Organisations will still require communication skills at a level beyond AI; we will need to invest in the skills to perform where the human touch is still valued and may well become more valued.
Handled well, Generative AI has the potential to free creativity and make collaboration and idea sharing boundaryless and more democratic. If we are to deliver on the promise of being freed from routine, mundane and repetitive work, we will need teams who can come up with creative sparks of genius which go beyond a repetition of what has gone before.
In the past, we have been seduced by the promise of automation to deliver increased efficiency, greater productivity and more leisure time. Instead, we have experienced increased inequality and - for those who retain their jobs - a need to work longer, harder and for less reward. We have another chance to get this stuff right. Lets not blow it in a machine inspired race for productivity which happens to us, not with us.
This article was first published in Training Zone