Learning and AI - Is ChatGPT our friend or foe?
Simon Dewar
Group CEO | Transforming learning for the future of work | Director | Austrade Delegate
Reflecting on AI Trends in Learning
You’d have to be living under a rock if you haven’t heard about ChatGPT and AI-generated content. The world (and by extension, the learning profession and many others) has been abuzz with doomsday conversations about the next revolution, bots replacing jobs, and the efficacy, legitimacy and ethics of using AI in our everyday lives.
As a team creating digital content, we spent some time reviewing the tool early this year, and thinking about its impacts (positive ones, as well as threats) as part of our first piece of team research for 2023. We’d previously looked at other AI tools such as DALL-e, and Midjourney for images, and a range of other AI based tools, but we saw immediately the potential of ChatGPT to drastically change our workflow.
So, we’d like to share the outcomes of our investigating various AI trends, including the ubiquitous ChatGPT, to discern which ones can, should, and just as importantly, should not be incorporated into the adult learning design and development work that we do. The field is changing quickly, with both the machine learning of AI and new versions of the tool in development, but here is our current thinking as of March 2023.
What does it do well?
It has incredible power to structure early learning designs:
- Can understand methods, tools and learning models
- Can provide base or High Level Design-type frameworks for building learning plans
- Can easily structure and write common, public domain topics (Think Health, Safety and Wellbeing)
What have we used it for?
- Write high-level plans
- Summarise content and help formulate outcomes
- Write and tweak scenarios
- Change the tone or format of learning content (eg. formal to more casual, or dot points to a story, or to translate text to other languages)
- Make verbose content more concise
- Write short scripts
- Brainstorm creative concepts and approaches
- Understand complex topics in simple terms
Where does it struggle?
- Currently its datasets are based on American content, therefore local content and information is lacking
- Referencing its sources is a challenge, as it aggregates data and will preference more commonly occurring information rather than correct information
- Has limitations with biases, and needs to be carefully reviewed and validated particularly as it speaks with a high level of authority (even if it is wrong - it might make up the answer - these are known as "hallucinations")
- It can incorporate learning theories at a high level, but clearly does not fully understand them therefore building out a high level design using a range of adult learning principles to sequence and determine modalities is currently beyond its capabilities.
- Creativity and uniqueness of ideas/content and information coming out of the tool is currently a little lacking without human input to hone results effectively.
Used right, just like a smarter version of Siri, or Alexa, ChatGPT is a friend and can save us time with the cognitively repetitive or labour-intensive tasks associated with Instructional Design.
?
What should we be careful about?
Spend an hour experimenting with content generating AI and you’ll recognise its ability to interact in a very human-like way- but with the power of the worlds’ most powerful supercomputers behind the curtain. It’s raises many questions, and of course some concerns for the future of work.
It will clearly replace some jobs, but it will likely also revolutionise every job – It’s already very clear that AI can write in all genres (from poetry, to reports to computer code) as well as, and in many cases, better than a human.
领英推荐
Time and efficiency: What do you do when an article that would have taken a week to write takes less than a minute, plus 30 minutes of editing? Blue collar employees have already felt the tremendous impact of robotics in recent years, is it now it’s white-collar employees’ turn?
Cost and Access: It’s free now, but not for long – ChatGPT is currently free, but as GPT4 becomes embedded into facets of everyday life, there will be micropayment fees that go back to the AI companies. It’s not like you can choose to support your local AI company either – this is a global organisation standing to profit enormously from every day small interactions with AI.??
Privacy and Intellectual Property: AI learns by digesting huge volumes of human-generated content; using all of humanity’s digital information to generate knowledge in a format that we value. It’s a black box in terms of privacy. We don’t know what happens when we upload an article for spell checking – presumably that information then becomes usable to the AI system. Likewise, we don’t know how information we enter now on an individual basis will be used in an identifiable way in the future- for advertising, surveillance, or other new and potentially nefarious purposes. As with any emerging technology, the full implications of AI are as yet unclear.
Already, artists whose work has been fed into AI tools as a dataset, have raised serious concerns about its ability to mimic their personal style, in some cases rendering their life’s work down to a simple and repeatable engine to reproduce in whatever form a user can dream up. The rise of deep fakes is leading us to question what we see, hear and read online. This will only become more prevalent as the tools progress.
Ultimately, AI in our everyday lives will become ubiquitous – like water that comes out when you turn on the tap, broadband Internet and so many parts of modern life, AI will likely become just a part of how we live our lives, and we will come to depend on it.
What’s on the horizon?
Chat GPT4 AI engine is here, and it’s going multimodal and will be capable of analysing image inputs, video, text, and audio in all languages, revolutionising content creation and translation services. The ChatGPT interface uses the GPT4 engine for text-based interactions, and so far, doesn’t output content using the multimodal capabilities. However, this will likely be added over time.
Open AI has now released Whisper APIs, enabling third-party developers to integrate GPT4 into their software in new and unimagined ways for more advanced and nuanced human conversation. From a workplace performance support perspective, organisations can integrate their policies, procedures and learning content into an AI-augmented platform… This will revolutionise knowledge management and training within organisations in the years to come.
AI is here, it’s revolutionising Learning and Development, get used to it!
As a leader responsible for the continued prosperity of the business and my team, I’m constantly challenged to think about where we need to head together to stay ahead of the fold during a period of disruption. Content generating AI has certainly ‘rocked the boat’ lately, but I feel through healthy experimentation and working together, we will be able to realise great efficiencies and strengths from the tool and be able to focus our brains on the complex challenges facing our clients. Embracing the technology and maintaining curiosity, as well as some healthy skepticism, will ensure we navigate the wave of change ahead and even deliver some new, unthought of innvovations in learning.
How do we transition into a healthy relationship with this new technology so that we can celebrate the best that ‘being human’ has to offer? I’m interested in how this tool might create space for us to think more deeply, solve the big problems, and remove the noise from our working lives. Will it move us closer to the much-discussed four-day working week?
For me, the answers lie in learning the tool has a lot to offer, but like any tool in our arsenal, it is the skills of the user, the questions we ask, and the information we put in that will ensure positive results we all seek.
What do you think?
I’d love to have a conversation with you about this, and to hear the reflections of the broader L&D community, so please share your thoughts in the comments, or hit me up for a coffee or a chat!
NOTE: Currently none of our client projects or storyboards are written using Chat-GPT. We are conscious that any information entered into the tool is subject to confidentiality and have used it with generic contexts and situations for our research.
Helping people get heard at work | Speaker | Author | Communication Dynamics Expert
1 年What a great article Simon Dewar. I'm very interested in how this is going to change our lives. I too, am frustrated by the lack of referencing and would never rely on it to provide me facts. Right now I use it to make my existing copy 'funnier' or to rewrite content in more interesting ways. I don't want it to think for me but I am interested in how it can level up my expression. Keen to follow this thread..
Strategic Solutions Consultant - BSI Digital Learning
1 年David Swaddle recently posted some content generated with GPT3 and Synthesia as an example of what's possible right now in terms of content creation - this example made in approximately 90 minutes. It's interesting to see what's possible right now. https://www.dhirubhai.net/feed/update/urn:li:activity:7039358134259187712?updateEntityUrn=urn%3Ali%3Afs_feedUpdate%3A%28V2%2Curn%3Ali%3Aactivity%3A7039358134259187712%29
I design resources to help people learn to do their jobs better. Instructional Designer | Facilitator | Training Wizard | 0417 732 217
1 年Fantastic article Simon Dewar, covering both the pros and cons of AI in our instructional design work. The biggest issue for me is the absence of references. How do we know what’s true, where does the information come from? Also, it’s a lack of respect for all the great researchers whose published work is used to write the answers to our questions. On the other hand, it may save us time and provide us with ideas and different perspectives and that’s absolutely exciting!
Artist at Lisa G Hunter
1 年Interesting article Simon Dewar. Thank you,