This article is more than 1 year old

Generative AI is coming for jobs

There's been plenty of chatter surrounding generative AI and assessment - but what about the impact on careers? Mark De Freitas has some opening gambits
This article is more than 1 year old

Mark De Freitas is a Careers Consultant at UCL Careers

Understandably, a lot of the early focus regarding Chat-GPT in higher education was on the challenges and opportunities it presents regarding academic research and assessment.

Following the release of GPT4 and Bard, these concerns have increased – centring around the speed with which AI is developing and the overall operation of large language models.

Throughout this debate, there has been little or no focus on the potential of generative AI to disrupt and transform careers education.

Bizarrely, Prospects, the UK’s biggest graduate careers website’s only article on the issue makes no reference whatsoever to implications for careers education.

Academics appreciate that generative AI presents a quantum leap in terms of opportunities and challenges. The release of GPT4 and Bard have increased the widely reported concerns that generative AI can be used to complete assignments and exams.

Having sampled Bard – currently, I don’t think it has the same capabilities as ChatGPT. In Bard’s own terms: “Bard is still under development, while ChatGPT is a more mature product”. But, the level of investment in AI R&D ($223bn in 2022 alone) means this position may change.

There has been some discussion among academics of the impact on the overall curriculum and competency based education. Overall though, Prospects approach is typical, with virtually no discussion of the transformative potential of generative AI for careers services.

This is all the more surprising given that this debate is so wide and moving so quickly – into areas such as data protection, and calls for a temporary pause on AI development.

Therefore, my question is a simple one – HE Careers Services – where are we in all this?

Transformative power

Large language models have the ability to be used across all areas of careers education. They can:

  • Provide detailed insights into specific sectors, employers and roles
  • Suggest the wording for applications, cover letters and CVs; and give feedback on them
  • Provide suggested areas for interview preparation, including specific questions (and answers)
  • Translate any of the responses they generate into other languages.

There are limitations for now, among them:

  • Its data set is primarily data on the internet as of September 2021. ChatGPT claims this data is “not static” and it has the ability to ‘learn and adapt from new conversations’. In my experience, ChatGPT’s ability to do so is very limited
  • ChatGPT’s difficulties in accessing data from PDF’d sources (as with Company Annual Reports)
  • The fact that some (especially initial) responses can be generic and occasionally simply wrong!
  • The need to re-define and limit the parameters of a question to ensure the response is as detailed as possible. Essentially, to challenge and “fine tune” the answers received.
  • Many of the potential careers applications for ChatGPT are available on other paid-for platforms. Also, some of ChatGPT’s claims (e.g. to review CVs and applications, against a specific job description) are, for now at least, no better than those other platforms.

A single, adaptive platform

I believe the existing subscription based providers of specialist careers support will be looking anxiously at ChatGPT with its potential to steam-roller their business. More importantly, in circumstances where at peak times, students cannot secure appointments, ChatGPT has the potential to (and I believe will) transform careers education, advice and guidance.

Currently (at least) free to use, the beauty of a platform which looks to answer any career question and enables 24-hour interaction is simply incredible. ChatGPT’s ability to fine tune its answers as a result of user interaction is central to this.

The fact that ChatGPT is already being used by students so widely as a study aid underlines its potential. There are, of course, difficulties in terms of over-reliance on one platform but for years, careers services have engaged in endless debate regarding aspects such as how to:

  • Scale up to meet student needs (e.g. through being incorporated into the curriculum)
  • Provide interactive, personalised responses and up-to date labour market information (LMI)
  • Free up staff time so that common questions can be addressed without tying up staff

Individual companies have then looked to meet specific needs by carving out and selling niche products (e.g. for CV reviews and LMI). I believe this typical HE careers model is tired and does not adequately provide for the needs of the students we serve.

More significantly, it is about to be disrupted and torn apart by AI and we are not even in the debate regarding it.

Graduate employers (as with HE generally) were not ready for the release of ChatGPT. Employers have little cause for complaint as they had no reservations using algorithms in recruitment. Thankfully, this is becoming increasingly regulated/restricted. ChatGPT levels this playing field, providing students with a powerful tool to compete with employers. The democratising nature of a free platform enabling this should not be underestimated.

Joining the debate

In some areas, decisions have been taken swiftly, (e.g. The Guardian ChatGPT use in IB 27 Feb) but it is a struggle to keep pace with AI developments. This has led to wider calls for a temporary pause on AI development .

I think there are certainly needs to examine the way models like ChatGPT aggregate data and comply with existing regulation, but I agree with the view that a call for a moratorium is misplaced.

Careers services though are not even in the debate. One of the very first questions I put to ChatGPT, was ‘Would you like a cup of coffee?’. This was to gauge its response. Given the transformative nature of ChatGPT, it’s time to gauge the response of university career services. So, Careers Services:

  • What are we doing to investigate and explore the use of ChatGPT and AI more generally?
  • Is ChatGPT/AI the tool we have repeatedly looked for in terms of enabling us and our students to cope with demand, especially at peak times?
  • What does the availability of a currently free to use platform mean for specialist providers of CV screening and LMI etc.?
  • Should we be concerned at ChatGPT’s data aggregation model? Alternatively, given our students are already using it as a study aid – is this of little/no concern?

It looks to me that we are collectively in need of a large cup of coffee and wake-up call!

Currently, the full implications of ChatGPT within careers education are not being discussed and are certainly, even at this early stage of its development, not being fully realised.

6 responses to “Generative AI is coming for jobs

    1. Great article Mark! In answer to one of your questions about feeling concerned. I see this as an opportunity to try new things, embrace the changes and do what we do best- offering personalised and contextualised guidance to individuals with all the information at hand. But equally I’m not quite sure how to harness this tool to best effect. Did careers services have the same questions when google surfaced? How did they resolve them? I asked my students in the classroom how they are using it- we spent time exploring together what it CAN do, and why students are using it. Interestingly, many don’t want it to write their work for them but prefer to to feed it ‘prompts’ which help them to find resources quicker so they can do ‘deep dive’ into topics that interest them. Perhaps it’ll be the students themselves who end up teaching us the best way to use Chat GPT?

      1. Hi Clare,

        Many thanks for your comments. I think you’re right to see ChatGPT in an incredibly positive light. I think you’re right too that among other things students are using ChatGPT for initial research. Both you and Susan have raised really interesting points regarding ‘personalized and contextualized guidance’. I’ve therefore gone into some detail in my answers to you both.

        Regarding students showing us how best to use ChatGPT – one aspect I think we might need to recognize is that the empowering nature of ChatGPT means that students don’t need us in their interactions with Generative AI. We have always seen ourselves as ‘gatekeepers’ with guidance being the preserve of a select trained group. However:

        Firstly, I think we should acknowledge that most of the time we are not providing guidance as such. Most of my time in appointments involves student requests for information and advice.

        Secondly, I think we collectively need to be honest with ourselves and take a step back to look at the ‘traditional careers model’ in the way that you are doing. Typically, this is a series of fairly ‘standard workshops’ with a very limited number of one/one and small group appointments.

        Finally, this model does not meet student demand, let alone student needs. And this is in the context of international students paying £25k to £35k/year for an undergraduate degree. I only work with one course at UCL School of Management, but this is far from the norm. Across the sector, students simply cannot access appointments.

        As a sector we have always placed an emphasis on personalized and contextualized guidance. But, this is largely not something we do, and universities can’t afford (or at least don’t fund) the resources to enable it. The other aspect is that Generative AI is developing at such a pace that I wouldn’t rule out a time when it will be able to provide what we would term guidance. For instance, the difference in the performance of ChatGPT4 to ChatGPT3 is phenomenal. Sitting behind this is a 571x increase in the number of data points.

        So, in short, I think you’re right that students are likely to develop the best ways of using Generative AI. But this will be from their own individual perspectives. I don’t think there will be one best way.

        Careers teams that are aware of this transformation will be investigating how to incorporate Generative AI into their overall careers education. I believe that students vote with their feet! Careers teams that are not live to this transformation might soon find themselves caught in the crossfire!

  1. I am currently utilising ChatGPT for careers content and sector information. I can curate articles at speed which are then ‘fine tuned’ and nuanced to support our target audience.
    I am also using ChatGPT as a dialogic tool to support students in their learning, understanding and decision making within a career context.
    The use of chatbots and AI to support career decision making and action planning has been happening for a while with CiCi the chatbot – https://careerchat.uk/
    Being able to scale’1:Many’ and harnessing technology to do it is critical, but I also think that ChatGPT is limited in its ability to deal with the personalised nature of students’ career thinking and decision making.

  2. Many thanks Susan,

    You appear to be using ChatGPT in a very similar way to me during discussions with students. Re generating articles/LMI, for me, the beauty of Generative AI is that it empowers students to do this themselves. This is not only useful in a careers education context – students will use Generative AI in the workplace so developing skills in fine tuning are essential.

    You’re right that Careers Chatbots have been around for a while. GPT1 itself dates back to 2018. However, I think what we are now seeing with Generative AI is something very, very, different. The interactive and adaptive power of these models is incredible. Just as importantly, access to careers education is being fully democratized.

    So, what does Generative AI mean for our interactions with students? Firstly, as I’ve mentioned to Clare, most of my work is not ‘guidance’ as such. I don’t think I’m unique in this. My work is largely helping students navigate the various hurdles in the recruitment process.

    What does Generative AI mean for students in their relationship with employers? Just last week, I was at a session where leading employers openly admitted that they were not ready for the release of ChatGPT and they are struggling to cope. Essentially, ChatGPT doesn’t just jump over the typical graduate recruitment hurdles – it drives right through them and smashes them into irrelevance!

    Of course, there’s a legitimate difficulty here for employers– how to identify the ‘best’ candidates from a mountain of applications? Employers have used AI for years; and at times possibly misused it. For instance, where was the consideration for the increased anxiety many students felt confronting algorithms analyzing their interview performance? Well, in my view, it was wrong for employers to use facial recognition technology in this way. This was especially in the way they did with little explanation of the criteria sitting behind it.

    Employers are now getting a taste of their own medicine! How do they beat the AI? Not just any old AI, but a student friendly AI that is free to access – more powerful, more agile, more adaptive and more capable than their own.

    Again, if we are honest with ourselves, I think we also need to acknowledge that ‘personalized’ provision is something we have only ever provided to the students lucky enough to secure appointments. Currently, I think you’re right that there are some limits on what Generative AI can do. Could Generative AI not just supplement our work but perform most, maybe all, of what we do? This area is developing so, so quickly; that I wouldn’t rule out a time when it could ultimately do this on a 24/7 basis – for any student, in fact anyone with internet access.

    Where do we sit personally in all this? I don’t think we should worry. We should embrace it – there will always be students that want provision on a face/face basis and not through an AI platform.

    As I’ve mentioned in the article, I think there might be some issues regarding data privacy. For instance, regarding adequate safeguards and how ChatGPT uses the data it gathers. Ultimately, should or will Generative AI be regulated in the same that the cookies are regulated? Probably. And perhaps more so given ChatGPT’s capabilities. Our students though have already signed up to it and are already using it. We need to be where our students are!

    This is the most exciting time I have seen. Everything is being re-drawn. The points you and Clare have raised are really significant.

    But, as a sector, where are we? As a sector, I think we are missing out on an opportunity to investigate a new tool with the power to level the playing field and radically overhaul our model for the benefit of our students.

Leave a Reply