How AI will change jobs as we know it

With all the hype about ChatGPT and Artificial General Intelligence (AGI) this year, a lot of people don’t realize what’s coming.

This is like the personal computer in the early 1980s or the Internet in the mid-1990s, it is just the beginning. ChatGPT is like the Apple MacIntosh in 1984, and things are about to get a lot more interesting.

However, a lot of people overhype it and say jobs will be getting replaced left and right by AI, and more people will be unemployed.

Not so fast! With every new technology, it will be used to make our jobs more productive, rather than eliminating human labour altogether. Many new types of jobs will be created as well.

The introduction of the personal computer did not replace writers, designers, engineers, or accountants. It simply made them perform their jobs on the computer instead of pen and paper, albeit much more efficiently. However, that means people who want to pursue these career paths have to learn how to use a computer or risk being eliminated.

Humans have an amazing ability to adapt. Most life on earth operates on the principle of adapt or perish. The same goes for technology. It is what drives evolution and human technological advancement.

Here is what I believe will happen to jobs in the coming years with the advancement of ChatGPT and AGI (Artificial General Intelligence).

Let’s begin by looking at what ChatGPT is and its capabilities.

I won’t go into too much detail, but ChatGPT is just the interface for GPTs, or Generative Pre-trained Transformers, which are essentially massive deep neural network models (more than 1 trillion parameters for the latest GPT4) that are trained on huge amounts of text data (like hundreds of billions of words) online. They are then tuned specifically to perform certain tasks, such as sentence completion and generation based on given prompts/instructions, aka a chatbot. It is basically a massive memorization of existing text on the internet by a computer. You can even train it to memorize your own set of texts.

Because it is essentially a memorized model, they are only good at performing specific closed-system tasks that are being trained for. For example, giving known answers to questions, writing emails and essays, as well as completing standardized tests and exams on various subjects. It is only good for regurgitating and summarizing existing knowledge.

ChatGPT’s so-called “creativity” is merely giving re-iterations of existing knowledge. Try making ChatGPT write a poem, a story, or a novel about something. It will always give you some generic, templated response borrowed from existing works. It does not have the ability to give you something that it didn’t have data for.

There are similar AI tools out there for image and video generation (such as Midjourney and Synthesia), as well as sound/music generation, that were trained on image, video, and sound data from around the Internet, which generate images/videos/sounds based on a prompt. Even though these give great results for what we ask for, and they do so by synthesizing existing images/videos that the models have memorized, they cannot come up with something completely novel that they were not trained for or memorized. Conceptually, they are developed exactly the same as ChatGPT, except they used different forms of data (images/videos/sounds vs. words/text).

Now, ChatGPT is just one of the very first versions of what could become something much more powerful — Artificial General Intelligence (AGI). Memorizing and regurgitating existing knowledge worldwide is just the very first step to AGI.

Ideally, an AGI could learn to accomplish any intellectual task that humans or animals can perform. Alternatively, AGI has been defined as an autonomous system that surpasses human capabilities in the majority of economically valuable tasks.

Now you can train more advanced AIs that have a greater ability to memorize and synthesize existing knowledge and data. However, in the short term, I do not believe we have created the ability for intelligent agents to truly become “creative”, i.e. coming up with novel concepts that were not merely iterations of existing concepts.

Now how is that going to impact jobs as we know it?

Machines will always beat humans at closed-system type problems or games. They don’t have to solve them, they just have to make fewer mistakes.

— Garry Kasparov

I think by now, it is very clear what the limitations are for this current generation of AI tools including ChatGPT. They are good for performing defined tasks they are trained for which rely on existing knowledge of the world, albeit at a greater accuracy than humans.

They can give you answers to generic questions, write a software program, help you write an assignment or exam, and build a customer service chatbot, along with many other clearly defined tasks.

However, to ask for AI to come up with new concepts, whether it be a new mathematical or scientific concept, a new form of art or music, a new culinary dish, or even a decent fiction novel, is asking for bulls***.

So right off the bat, any job that requires some degree of creativity will not be going away — People who create new forms of content, arts, science, and businesses. To truly innovate, one cannot always be drawing from existing ideas. I doubt AI will have that kind of ability anytime soon.

Also, as I’ve written before about the harms of virtualization, we cannot eliminate direct human contact without severe consequences (poor mental and physical health, more de-personalization, etc.). AI is just another step forward in the increasingly virtual world that we live in.

Despite how many people fantasize about AI-powered humanoid sexbots, it just feels odd. And it feels wrong in many cases to be interacting with an AI or robot rather than an actual human being, for example, athletes in physical sports, getting a massage, starting a business, doctors, lawyers, engineers, teachers in schools, or customer service.

Furthermore, most of the world operates on capitalism. In other words, is it financially beneficial to replace human labour with AI? People have talked about automation for years and how it will replace jobs, but you still see a lot of jobs out there.

Even for existing jobs that can be easily replaced by AI and automation, such as most manual labour and service jobs, since it doesn’t make sense financially to do so everywhere, many of these jobs still exist today. And because of how expensive it is to build computing power and robots, many of these jobs will keep existing until it becomes significantly more cost-effective to replace them with AI. Don’t count on these jobs staying around forever though. Look at what happened to most farm labourers or factory workers with the introduction of machinery.

However, for many of the existing jobs out there, being able to use Artificial Intelligence will not only give you an edge over other workers, but will soon become a job requirement. For example, writers who used to work on pen and paper and typewriters were required to use a computer to do their writing once it came out. And eventually, they will be required to use AI to assist in their writing. The role of the writer will not go away.

On the other hand, I believe this will have profound effects on how we approach education, especially with how easy is it to cheat in academic settings using ChatGPT. I will discuss this in another blog post soon.

Finally, the proliferation of AI will certainly create new jobs and career paths just like automobiles, personal computers, and the Internet did in the past. Machine learning and AI engineers will only become more common to build and maintain AI systems, as well as people who work on prompts/instructions for the AI (prompt engineers).

Ok as an average person in the workforce, how do I make sure I can adapt to the rapidly changing job market influenced by AI and automation, so I don’t perish?

• Learn how to use AI as a tool. Just as personal computers, and to a lesser extent, software programming, have been introduced to many jobs in the past, ChatGPT and AI will simply be another tool at the workers’ disposal to make them more productive. Learn how to use them effectively in your job functions.
• Learn how to be creative and think outside of the box — Originality and creativity cannot be replaced by AI anytime soon. Many jobs that involve a high degree of innovation, such as research, arts and design, entrepreneurship, as well as content creation (e.g. YouTube) are relatively safe. Also, be innovative and think about how to apply AI to solve problems.
• Work with other people — As I said, there are many roles where it would be too problematic to replace humans with AI. One cannot have AIs as doctors, engineers, or politicians making important decisions that impact many people. You cannot have robots competing in human sports. You can enter one of these roles.
• Upskill — I believe lower-level, more repetitive jobs will continue to get automated as long as the cost justifies it. However, if you keep upskilling and becoming more advanced in any career/job, you will simply become more valuable and difficult to replace.
• Soft skills — At least in the short term, humans are much better at soft skills (e.g., communication, emotional intelligence, people management, etc.) compared to AI. AI can replace certain hard skills (e.g. programming, accounting, etc.), but most soft skills cannot be replaced yet. Stand out and work on your soft skills.
• Become the ones working on AI — You cannot have AI building AI anytime soon. If you’re working on AI, you’re not going to be replaced, and you will have lots of opportunities as the world continues to virtualize and run on AI. It’s not easy and it requires good mathematics and computer science knowledge, but just as more people have learned how to use a computer and programming in the past, people should be able to pick up machine learning and AI quickly.

At the corporate and societal levels, we should also encourage application, experimentation, and creativity in using AI, using it to solve problems. It should not merely be a means to replace labour and increase productivity.

But we should also understand the risks of AI. No new technology is risk-free.

Innovation is the ability to see change as an opportunity, not a threat.

— Steve Jobs

People online have said far-fetched things about the future of the world with AI. Some believe most people will be out of jobs and collecting UBI except for the select few working on the AI systems. Others believe we will soon live in a dystopian world run by AI and robots, where humans become slaves to AI masters. Some even say this will be the end of human civilization as we know it since AI will destroy the world. I believe all these scenarios are highly unlikely to be realized. It’s doomerism.

People in the past have made similar doomsday predictions about nuclear power, personal computers, and the steam engine. None of these scenarios have ever played out.

AI is simply a new technology at our disposal. It creates new opportunities. Throughout human history, with the introduction of every new technology, from fires to agriculture to the steam engine to nuclear power and computers, there will be profound societal changes. But as history has shown, humans will adapt quickly to the technology and learn how to use it effectively without destroying civilization.

As someone actively involved in artificial intelligence and working as a machine learning engineer, these are just some of my thoughts as I continue to observe the changes in the AI world.