With the rapid evolution of AI technology and related industries (see my article “2024: An Industrial Revolution”), we will see many changes in the job market.
In the previous industrial revolution, automation displaced many factory workers. These individuals eventually found other jobs, such as selling insurance, driving taxis, writing computer programs, and so on. With today’s industrial revolution, the jobs being replaced are in office buildings and software companies (the manufacturers of computer code). The displacement of white-collar jobs is another significant event that will unfold in the coming years.
Let’s take a look at jobs disappearing. While it may look depressing, new jobs are also being created. We will discuss them later.
First, human translators will go away as AI translation becomes available everywhere through text or speech, in real time. Personal assistants who do basic Internet research, scheduling, and travel booking will be replaced by AI. Copyeditors will be completely replaced as AI does an excellent job in grammar checking. Creative editors’ jobs will be replaced as AI can be asked to give suggestions for the next plot or writing prompts.
A large amount of marketing jobs will go away, including marketing copy writing, marketing visual creation, and social media management (scheduling social media posts, creating posts, and so on). Most of these processes can be automated.
Many customer support jobs will disappear as the sophisticated AI assistant can answer user queries and help users complete tasks. Some other obvious jobs replacements are: Many front desk roles, such as answering phones, taking orders, and scheduling appointments, will disappear. Similarly, account management positions, including accounts receivable and customer outreach, will become automated."
1. Disappearing Jobs in Software Industry
Database engineers, particularly those specializing in ETL, will experience job reduction as Text-to-SQL becomes the norm, allowing people to generate SQL queries through natural language. Correspondingly, data analysts (or data scientists performing similar functions) will lose jobs as AI can easily generate dashboards and analyze results from queries. In other words, executives can directly query data and obtain business information themselves, eliminating the need for intermediaries.
UI designer and mobile developer jobs will decline as screen interfaces yield to speech interfaces. Similarly, the demand for frontend engineers who build pages for websites or mobile apps will decrease. There will also be a decline for UX designers, as newer interfaces shift towards speech interfaces. Traditional UX designers may not be as effective in this aspect. How do we improve user experience with speech? That is the job of AI engineers or ML engineers.
Backend engineering jobs will remain stable for the next two years as companies transition towards automation. This means developers are still needed, but their job functions will shift towards using OpenAI API or other LLM APIs. These engineers will focus less on writing code and more on crafting prompts and performing integrations.
Many management jobs will disappear as software applications become more streamlined. With fewer coders and fewer steps in building applications, a top executive can quickly access the work done by engineers. This is because they can interact with the product-building process through words, and ask questions about it. Each company will become flatter, driven by the speed of innovation and the need to respond to the market.
In addition, many enterprise processes will be automated, from HR to management, employee training, and internal communications. This will lead to greater transparency and easier communication. Such automation will reduce the need for middle managers. As for motivating workers and ensuring project completion, it becomes less top down. As AI workers are highly educated and skilled, they will seek more self-initiative, and fewer managers will be needed.
There will be a decline in programming language training, as the usage of many languages will diminish. We will converge to a few languages, such as Python and its related languages (e.g. PyTorch). While old languages like Java or JavaScript may still be used, their prevalence will decline. Consequently, training and education for programming languages, other than those related to AI, will see a decrease.
2. New and growing Jobs
In 2024 and onward, we are entering an AI boom. It will drive every area of business and every aspect of our lives. More jobs are being created for understanding AI, improving AI, building AI applications, creating AI devices, and maintaining AI infrastructure. The real 'brain' of AI sits in the cloud, which is a cluster of thousands of GPUs.
There is a boom in jobs in the GPU sector and cloud deployment sector. This means cloud engineers, as well as GPU programmers, will be in high demand.
There will be many edge AI devices, ranging from child companions (toys) and personal AI pins to home appliances. We will enter a more connected world, where our on-body AI devices are always connected. With the push of a button or a voice command, these devices will deliver anything from the Internet, sifted and analyzed by this AI assistant. This will create a host of new jobs. They include:
Edge device engineers who understand both small hardware and AI.
Product managers who understand the pipeline of AI products, the basic technology, and market trends.
AI engineers who build AI applications. They typically have a Ph.D. or a Master's in AI (which provides foundational understanding) and know how to fine-tune models.
Infrastructure Engineers who maintain a company's AI pipeline.
Workers who ensure network connectivity and communications.
Strangely, there will be a decline in the number of AI algorithm researchers. This is because building foundational models becomes very expensive. Very few companies or universities can afford a large GPU cluster for experiments. Fundamental algorithms are not needed that much as the model behind AI become unified. There will still be some companies or organizations building open-source foundational models, but their number will be very limited.
Certainly, there will be ongoing work on improving LLMs (reducing hallucination, preventing adversarial attacks, etc.), designing agents (integrating LLMs with action), and training and improving agents (determining the best actions to take). But the number of existing researchers in the AI domain is sufficient to advance this field, thanks to the immediate sharing of knowledge on platforms like arXiv and the open-source movement. This is also due to AI's self-learning automation. If AlphaZero is any indicator, we will see AI programs conducting research independently. They will seek self-improvement, using another AI for feedback. This possibility was demonstrated by the Reinforcement Learning with AI Feedback (RLAIF) method and weak-to-strong generalization from OpenAI.1
While the number of pure AI researchers may plateau or decline, there will be an increase in hybrid researchers with a minor or double major in AI. For example, a biology researcher who understands how AlphaFold works and its predictions for new proteins. A medical researcher who collaborates with an AI program to find a cure for cancer. A brain scientist who uncovers images seen by a person by analyzing brain recordings. An economist who relies on AI for large simulations of economic activities and human behavior. A building architect who consults an AI program for drafting initial designs. Every research field will be impacted by AI, meaning researchers in these fields will need to collaborate with AI researchers or train themselves in AI.
AI Ph.D. programs will continue to grow, not just for training researchers, but for training highly skilled AI workers. The field of AI has become so knowledge-intensive that it requires Ph.D.-level training to fully understand the underlying methods and to optimally improve a model in production. AI Ph.D. graduates will continue to be in high demand in the job market for at least the next five years. This trend is reminiscent of the biological field, where Ph.D. education is commonplace, and post-doctoral training is starting to become a requirement in many research labs.
There is a growing need for AI education as this technology advances rapidly, and AI becomes increasingly prevalent in our lives. More people, including both students and professionals, are enrolling in AI classes. Even professors need to update their knowledge, as many older technologies have become outdated. This includes almost all existing machine learning classes that discuss technologies from before 2017. Computer vision has been transformed since 2020, following the advent of vision transformers and the application of diffusion models to image generation.
This growth in AI education is very similar to the boom in biology education in the 1990s, driven by advancements in DNA sequencing and drug discovery. In the 2000s, the boom in the Internet led to large enrollments in computer networking classes. Many people moved into fields like building Internet routers, data centers, and so on. In the 2010s, mobile development professionals were in high demand as smartphones proliferated around the world.
There are growing jobs in AI consulting. Many companies don't know how to integrate with current LLMs and related services. AI-related jobs include OpenAI API integration, writing effective prompts, fine-tuning, and building LLM-powered systems from scratch.
3. Other jobs (long-term)
We are still at the beginning of the robotic age. In the foreseeable future (2-3 years), robots will enter households. With the rise of AI companions and robots, we are likely to see a declining need for pets (dogs and cats). The reason we keep pets is for love and companionship. However, an AI companion can talk with you, experience life with you, and provide support for you. If you have an AI, like "her," accompanying you everywhere, telling jokes, and engaging in conversations, would you still want a dog? When a physical form is truly needed for companionship, such as walking, running, or playing, we will see robotic pets unlike anything before. They will be agile, run gracefully, and even converse with you. Imagine a talking dog (or talking cat)! This shift could lead to a decline in jobs in the pet industry, from pet food to pet services, and animal hospitals.
As we are going through another industrial revolution, the future of jobs is both exciting and scary. Are you ready for the change?
Resources:
Check out our previous articles:
A list of major AI trends in 2024, and how they will impact businesses and our lives.
A review of major AI development in 2023, with in-depth analysis and references.
Announcements:
Join our next meetup on understanding VideoPoet on Friday, January 19, 2024
Lee, Harrison, Samrat Phatale, Hassan Mansoor, Kellie Lu, Thomas Mesnard, Colton Bishop, Victor Carbune, and Abhinav Rastogi. "RLAIF: Scaling Reinforcement Learning from Human Feedback with AI Feedback." arXiv preprint arXiv:2309.00267 (Sept 1, 2023).
Burns, Collin, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold Aschenbrenner, Yining Chen et al. "Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision." arXiv preprint arXiv:2312.09390 (Dec 14, 2023).