Global EN

Generative AI is Probably Not Going to “Steal Your Job”

Shane O’Hanlon

DevOps Engineer, Synechron , Belfast

Artificial Intelligence

But it promises to increase productivity and lead to a new workplace paradigm 

Since the release of ChatGPT in November 2022, we have entered what can only be described as an “AI Revolution,” sparking a level of intrigue and concern that probably has not been seen since the mass adoption of social media. Social Media has changed how both consumers and businesses communicate, created an entirely new medium for advertisers and drastically changed how we interact socially. It has also raised concerns around privacy, misinformation and has been seen to have a mental health impact on its users. As AI technology continues to rapidly evolve, there has been a growing apprehension that these technologies could replace human jobs, leading to widespread unemployment akin to the industrial revolution.

But is this anxiety over job loss a true reflection of the capabilities of AI? Or will generative AI tools instead increase our efficiency, reduce those mundane day-to-day tasks and complement our skillsets?

Understanding Artificial Intelligence 

Artificial Intelligence is a very broad term referring to systems that exhibit capabilities normally associated with those of humans; tasks like language comprehension, pattern recognition, problem solving and decision making. This is achieved using Machine Learning, where machines are trained repeatedly to learn from datasets and improve their responses, without what we would call “traditional programming.”

GPT is a model that was developed by Open AI, with the latest version being GPT 4. This uses a machine learning approach known as “Transformer Neural Network”, which has been shown to be excellent at interpreting language context. It uses a set of mathematical approaches called attention or self-attention, to detect how sequences of data elements influence and depend on each other. This technique was first described in a 2017 paper from Google. GPT was, and continues to be, trained on swathes of internet text using this machine learning technique. It is then able to generate new and original content with a humanlike tone. This ability to understand and create text based on the data it was trained on has made it a formidable tool when assisting humans -- from creating original content to writing code.

A true AI Job Apocalypse? 

The “AI Job Apocalypse”, as it has become known, has stirred up considerable debate and of course a lot of concern. The accelerating development of these technologies has led to a widespread fear that it will lead to mass unemployment. Predictions, however, by the World Economic Forum’s “Future of Jobs Report 2023” paint a much less negative picture. 75% of organisations surveyed expect to adopt AI in some fashion. The adoption of AI is expected to drive job growth in more than half of surveyed companies, offset by expected job displacement in one-fifth of companies. AI and Machine learning specialisations top the list of fast-growing jobs and is determined to be the third highest priority for workforce development and training programs. They also predict that AI, economic and social conditions over the next five years will result in 83 million job losses with 69 million job creations: resulting in a net loss of 14 million jobs. This is far from an “Apocalypse.” 

The primary role of AI in the workspace is to automate the more mundane 
and repetitive tasks allowing human workers to put their attention 
to the more complex, creative, and strategic aspects of their roles. 

The primary role of AI in the workspace though, is to automate the more mundane and repetitive tasks allowing human workers to put their attention to the more complex, creative, and strategic aspects of their roles. Furthermore, the integration of AI into the workplace will likely enhance productivity allowing increased business growth and job creation. Using customer service as an example, AI can handle a lot of the simple queries freeing up human agents to handle the more complex customer issues. Similarly, manufacturing processes can be enhanced with AI, resulting in more automated systems, enabling humans to focus on strategic planning and tasks requiring intricate decision making. 

New job creation amid necessary human input 

Previously non-existent roles are also being created, particularly in the ethics space. These roles will ensure that AI systems are designed and used in a manner that is ethical.  As AI systems grow, these roles will become increasingly important with tools integrating both into business and our everyday lives. New roles will be created to train these large models and ensure they understand context, and can make the right decisions. Roles like this underscore the fact that while AI can automate tasks, there is still a significant need for human input.  It's humans who decide what problems AI should solve, design the algorithms, gather, and select the training data, interpret the results, and make strategic decisions based on those. Humans are needed to provide context, ethics, common sense, and strategic direction -- elements which are currently beyond the capabilities of AI.  

Reinforcement Learning from Human Feedback (RLHF), a machine learning training method that combines reinforcement learning with feedback from human interaction, is critical to the continued evolution of LLMs (Large Language Models). In this approach, a policy is learned by the AI from human demonstrations, then a reward model is created based on human rankings of different actions or trajectories. This reward model is used to further fine-tune the policy, creating a cycle of learning and refinement. This allows the AI to leverage human knowledge and judgement, thus improving the efficiency and efficacy of its learning process. A study by OpenAI, the creator of GPT, stated that RLHF can reduce the toxicity of LLMs by 72%, increase an LLMs helpfulness by 30%, and improve the factual accuracy by 10% (Lee et al., 2023).  Another study by Carroll et al. (2022) used RLHF to fine tune a predefined LLM to generate Python code from natural language descriptions; this LLM showed improvements in the syntactic and semantic correctness, readability, and efficiency of the generated code when compared to another model that was only fine-tuned with supervised learning on a large dataset of code. These examples, among many other studies, have shown that RLHF increases the quality of an LLM’s output; human input is crucial to the continued success of AI. 

People with the ability to understand, manage and innovate with these technologies will be at a significant advantage. But this does not mean everyone needs to become an AI expert. 

The rapid integration of AI into industries highlights the importance of adaptability and continuous learning; humans must evolve. People with the ability to understand, manage and innovate with these technologies will be at a significant advantage. But this does not mean everyone needs to become an AI expert. Even a basic understanding and training on how to use these tools will be invaluable. Education and training will be critical. And this will need to be more than just technical training; the cultivation of a learning mindset and the willingness to integrate the system into day-to-day work will be crucial. Like Computer Science, which has been integrated into curriculum at both primary and secondary level education, adding AI to curricula should be a key strategy. Meanwhile 'soft skills', like problem-solving, creativity, emotional intelligence, and strategic thinking, will become increasingly valuable in an AI-driven world. While AI can automate routine tasks, these higher-order cognitive skills are areas where humans still outperform machines. 

Looking forward: Potential not fear  

While this rapid advancement of AI, especially generative AI like GPT, has sparked a lot of concern around job displacement, close examination reveals a far more nuanced reality. Our understanding of this revolutionary tool should not be rooted in fear but in the potential it holds to improve our lives, our jobs, and our businesses. For the most part, AI will work alongside us freeing us to focus on the more complex, creative, and exciting parts of our roles. While there will likely be some job displacement, along with shifts in our educational systems, and offering reskilling programs as we integrate AI into our lives -- just like we have experienced with the invention of the computer and the smartphone – this will ultimately lead to more fulfilling roles. Perhaps tools like this will lead to an entire paradigm shift in how we work and for how long. 

The Author

Rachel Anderson, Digital Lead at Synechron UK
Shane O’Hanlon

DevOps Engineer, Synechron

Shane O’Hanlon is a DevOps Engineer in our Synechron Belfast office. As a Principal Consultant with Synechron, he brings a wealth of expertise in the Cloud and DevOps space, having spent the last eight years here as a Cloud & DevOps Engineer. He holds a Master’s Degree in Computer Science and has worked for the last 17 years within the fintech sector. Shane has expertise in DevOps best practices, design and implementation of cloud solutions, and has been working alongside the Synechron AI Practice as an AI Ops specialist. 
  
You can email him at: shane.ohanlon@synechron.com

See More Relevant Articles