Looking back at the progress of AI just this year, I’m still surprised at the misconceptions people have about how AI works.
For years, we thought of AI as a machine built for one task. One system could read handwriting. Another could understand speech. Another could play chess. Each improved until it reached human skill, and some went beyond it.
That was the old story of AI. Today’s story is different.
We now have what experts call general-purpose AI. Large language models can write, solve math problems, study images, explain science, and help make software. Tests published through October 2024 showed leading models reaching very high scores, even on some exams made for experts.
The key point was not that AI could perform at a PhD level in one narrow field. It was that one model could show deep skill across many fields. Its real-world judgment was not always equal to that of a trained expert. Still, the range of skills was new.
A year later, the systems are better again. This progress did not come from software alone. It also came from better computer chips.
For many years, chip power grew at a steady pace. Then, around 2010, the pace of AI computing began to rise much faster. Nvidia and other companies started using graphics processing units, or GPUs, for work beyond video games. The same chips that could draw rich images could also handle many calculations at once.
GPUs helped mine crypto coins. More importantly, they helped train AI on huge amounts of text, pictures, sound, and other data. This opened the door to deep learning and, later, today’s language models.
As long as chips, data, and learning methods keep improving, AI will keep gaining new skills. Problems that seemed central one year can become less serious the next. People once talked about AI “hallucinations” all the time. Those errors have not vanished, but newer systems manage many tasks far better.
AI can now write useful software. As a software developer, I find that deeply ironic. Many of us expected programmers to build the machines that would replace other jobs. We did not expect software work itself to be among the first kinds of work transformed.
That brings us to the hardest question: jobs.
The old belief was simple. Robots would first replace low-paid factory workers. Office workers, lawyers, artists, engineers, and managers would be safer.
The evidence now points in another direction. In many studies, higher-paid jobs have more tasks that AI can perform or support. This does not mean every exposed job will disappear. Exposure is not the same as replacement. It means the job contains tasks AI can help complete. That difference matters.
Some jobs work well with AI. A worker may use it to move faster or make better choices. Other jobs contain routine tasks that a company might give fully to a machine. Human intent will decide much of what happens. So will the choice to keep a “human in the loop.”
An International Monetary Fund study showed that architecture, engineering, science, insurance, and computer programming are highly exposed. Office support, information services, law, accounting, and head-office work also face pressure. These are common jobs in business centers such as Makati and Bonifacio Global City.
Building care, landscaping, veterinary work, and transportation are less exposed for now. Even self-driving tools may support drivers before replacing them, especially in places like the Philippines.
Businesses should therefore stop asking, “Which workers can AI replace?” They should ask, “Which tasks should machines do, and which choices still need people?”
Consider elevator operators. Some buildings still pay a person to press a button that passengers can press themselves. Other buildings already choose the elevator and floor before a passenger steps inside. Keeping an old job may feel kind, but it can also keep a worker trapped in work with little room to grow.
History is full of such change. We no longer employ many switchboard operators. Refrigerators replaced workers who cut ice from frozen lakes. A “computer” was once a person paid to perform hard calculations. That worker might once have laughed at the idea that a machine would take the name.
Yet it did. AI will not make people useless. But it will make some tasks, and perhaps some job titles, fade away. Our duty is not to deny this change. It is to prepare workers, redesign jobs, and share the gains.
The machines are improving quickly. Our policies, schools, and businesses must learn just as fast.
Dominic “Doc” Ligot is one of the leading voices in AI in the Philippines. Doc has been extensively cited in local and global media outlets including The Economist, Channel News Asia, South China Morning Post, Washington Post, and Agence France Presse. His award-winning work has been recognized and published by prestigious organizations such as NASA, Data.org, Digital Public Goods Alliance, the Group on Earth Observations (GEO), the United Nations Development Programme (UNDP), the World Health Organization (WHO), and UNICEF.
If you need guidance or training in maximizing AI for your career or business, reach out to Doc via https://docligot.com.
