This week at Career Matters, we have been pondering the power of AI. We’re planning to introduce some new tools to our career methodology and are curious about to harness the technology to give our programme participants an even faster and more specific level of career clarity.
So when we came across Anne Hsu, a researcher in psychology and Artificial Intelligence, we took the opportunity to get the lowdown.
Anne is a lecturer in computer science at Queen Mary, University of London. She received her Ph.D. in Physics from UC Berkeley followed by a research position at the Gatsby Computational Neuroscience Unit at UCL, where she developed models of neural systems. Her work since has spanned the fields of machine learning, neuroscience, cognitive science, motivational and organisational psychology, and behaviour change/persuasive tech design.
Erica: Anne thanks so much for giving us the time to find out more. Let’s start at the beginning. How do you define AI to a layperson?
Anne: Colloquially AI is used to describe machines that perform functions we associate with human minds such as recognising sounds and images, navigating terrains, understanding language, and solving puzzles and games. In computer science, AI refers to systems that can perceive the environment and take goal-oriented action. AI research encompasses all fields that contribute towards this. Essentially this boils down to training machines to spot useful patterns in data. This is a very broad definition. In practice, what is actually considered AI evolves as computational tasks that are well-solved and what is colloquially considered AI moves towards what is new and not yet solved.
Erica: Why are people so interested in the potential for AI?
Anne: AI has actually been around for a long time. The term was coined around the 1950’s. The recent interest in AI has been from breakthroughs in benchmark performance in machine learning and its subfield of deep-learning. These breakthroughs were made possible through a combination of increased amounts of data, improved algorithms and significantly more powerful computer hardware that is able to apply more complex models to massive amounts of data.
Erica: So in effect, like mobile phones, the technology has been around for a while. But it is the increased utility of the tech that has meant it has become really valuable. What will it mean for work and business?
Anne: I think there is quite a difference in the issues that get brought up around general interest in AI vs. what it means for work and business. General interest in AI tends to be driven by a curiosity about mankind and society. It often provokes philosophical questions such as what it means to be an intelligent being, what defines us as humans and futurist visions of a world where general purpose AI’s mingle on par socially with humans.
These are largely still science fiction fantasies.
Even the splashy news stories from the AI labs where machines beat humans in tasks such as particular games, or generates human-like speech are very specific examples of AI work and are quite far removed from much of core AI that will transform work and business.
The primary business gains from AI will come from significant improvements in less glamorous tasks like better systems maintenance, stronger network security, improved fraud detection, more efficient supply chains and more accurate customer analytics.
Erica: So the primary and more immediate benefits relate to more effective data processing and analysis – and though there are many applications for this, this isn’t where the media focus lies. That tends to be in the ‘robots will replace us’ genre. Can you say some more about this?
Anne: There are different aspects to how people might be replaced by robots.
One aspect is social replacement, i.e., perhaps we will get so used to interacting with AI that we no longer respond the same way to each other as humans, or respond to each other less. While it is true we are all getting used to interacting with conversational agents, I don’t see it as taking away from our human interactions. Many technologies have made significant impacts on how we spend our social time, including the cinema, video games and the internet.
Except for some experiments in care homes and some areas of customer service, I don’t see the social impact robots being significant any time soon. Most viable conversational agents provide services the way a search or recommendation does, it’s not really replacing human social needs.
Also, I think there is even significant potential for future AI to help us interact more empathically and learn to connect more satisfyingly with other humans. It all depends on what the AI is programmed to do.
For example, an AI can be designed to engage in active listening, with the ability to point out the emotionally salient observations that were made and prompt its conversation partner to think about their feelings and needs. Experience with such attuned conversations can help people have more of such conversations with other humans.
Anne can you give an example of where/how this might play out?
The other aspect of replacement is in work itself. AI certainly will be changing the job market. Technology has always shifted the job market and it is true that many industries will be reshaped.
Hopefully, there will be fewer people having to do boring repetitive jobs and they can spend more time on tasks that require greater creativity and human expression.
Erica: I’m also excited by this idea that AI can free people up to explore pursuits that drive a greater sense of fulfilment and personal meaning – and you are right that technology has always been driving and shifting the work we do – and we don’t complain about the existence of the internet, or the ability to create a website or an app, so what is it about AI that generates such a strong level of discomfort? Perhaps it’s the idea that a different kind of intelligence might be working alongside us, or even ‘outsmarting us’!
How do you see people and AI interacting together successfully in the workplace – can you give some examples of how this might work?
Anne: Much of the interaction between people and AI is similar to how people have always interacted with traditional data analytics. The difference now is that the insights and predictions of AI models are much more specific than for traditional analytics, resulting from more sophisticated data input and algorithms. The scope of business processes being supported is now much wider.
AI helps recruiters to filter candidates, lawyers to find relevant documents, and accountants to audit contracts.
The other common example is customer service and question answering. Humans tag team with AI to answer calls. Many times when you call customer service you are served by AI as long as your query is systematic, but you are passed to a human when the AI can no longer help.
Another example closer to home is Professor Ashok Goel at Georgia Tech who has created an AI to serve as a Teaching Assistant to answer questions for his online module with hundreds of students.
Other workplace AI’s that serve as information and data insight portals are anthropomorphised with names and identities. I remember talking to employees in a consulting firm who said other employees did not accept the AI’s contributions until the AI was given a name and a conversational interface that allowed the employees to ask for information from the AI in natural language, rather than receiving the output of a set of computational algorithms.
Erica: Aha! So are we seeking greater rapport with our AI? And is it statistically relevant that the lawyers and accountants don’t need a name for their AI but the consultants do? I’m a consultant and I totally get that! (Laughs)
Let’s talk dark side now. Are there any risks in replacing humans with AI – if so, what are they?
Anne: There are risks. A key one is bias. AI’s will have all the biases that are present in the training data used. A classic example is in the case of recruitment. Just recently there was news that Amazon has scrapped its AI for internal job application sorting because it was biased against women.
If the labels for your AI training set are biased then your algorithm will be biased. Another risk for AI is that it can be black box, in the sense that it can formulate predictions without being able to explain why.
There is new research on AI algorithms that explain themselves and trying to define what that even means. In general, AI’s will have their own limitations. There will be flaws in the predictions and outputs.
We might not be as familiar with the scope of these limitations as we might be with that of humans. So there is the danger of putting too much blind trust into systems trained on data because the data can be biased.
Erica: That’s interesting. So the AI mimics the bias of the system it exists in – it doesn’t improve upon or eliminate existing design flaws. That seems an important factor to be aware of.
In every crisis there is also an opportunity. What opportunities might be created or realised by having AI in the workplace?
Anne: I think AI can empower the workforce. Employees will have the data to make better decisions. They will be relieved of routine tasks that can be automated.
The tasks that AI is far away from replacing are those requiring creativity, strategy, and emotional intelligence. The non-mundane tasks also tend to be the tasks that allow people more self-expression, allowing the potential for more meaning and engagement with work. For example, a nutritionist can have an AI help with meal planning and focus on emotional support and strengthening their rapport and relationship with their client, which in turn can increase the possibility of behavioural change.
There have been examples of AI achieving creative tasks, e.g., to create soundtracks, stories, or movie trailers. These results are interesting but also are far from continuously compelling and meaningful storylines. We’ve had electronic mixing for music for a long time, and I think AI could potentially further this way of creating by collaborating with the creative through suggesting certain snippets or patterns to consider. Same for writers, the AI can suggest an autocomplete sentence the way Gmail has recently released. For writing, if the writer wishes to sound more unique perhaps the AI can show what the expected sentence completions are so writers can choose to avoid these phrases if they want to write more interesting things.
Erica: This sounds like a great way forward – to focus on the collaborative process. Can you give some examples of what is currently underway in a variety of fields and why it is exciting?
I think one of the most impactful applications of AI is in the medical field. Just recently Google AI announced 99% accuracy in metastatic breast cancer detection. This technology would then highlight areas of concern for the pathologist to review for their final diagnosis. Now that AI can automate the more repetitive tasks in medicine, this frees up the physician’s time to focus on higher level clinical tasks.
Also, there is huge potential in personalised medicine. There are companies like Modernizing Medicine, which gathers the collective knowledge of tens of millions of patient visits to create recommendations and suggest diagnosis.
Erica: As a psychologist, what do you see as the evolving trends in our relationship with our work and the workplace itself?
There is an increasing demand for work to be more personally relevant. Younger generations buy less into hierarchical imperatives and seek personal fulfiment.
That concept was more foreign in my parents’ generation. Simultaneously, workplaces are getting more psychologically aware. Employee engagement is rising to the top of HR priorities. A huge amount of funding has gone into startups that aim to transform workplace culture. These include tools that help organisations to quickly gather employee views and act on them quickly.
Many workplaces have ‘thank you’ programs to prompt peer recognition. There are tools that use big data to help predict job matches, and help plan career trajectories and transitions. I hope these trends in employee engagement combined with AI automating the more mundane tasks can provide more people the opportunity to work that promotes growth and flourishing.
Erica: Absolutely. We’ve been focused on the importance of personal meaning and fulfilment in work since we began, but it has taken some years to feel that the client companies are fully alongside and understand the value and impact of addressing career aspirations and career conversations in a quality way. I’m excited to see how AI might partner with us to deliver even better results for clients.
To find out more about how Career Matters can help you retain, energise and engage your talent, please visit www.ericasosna.com.
Anne Hsu offers practical workshops and consultancy in the field of Artificial Intelligence and Machine Learning for business. To find out more, please visit www.annehsu.co.uk/.