Learn How to Learn: What Does AI Really Add?

According to the World Economic Forum's Future of Jobs Report 2025, 59 of every 100 workers will need training by 2030, and 11 of them are unlikely to receive the training they need. The contest here is not over who holds the loftier certificate, but over who learns faster and more deeply.

How much of skill is changing?

The same report expects about 39% of a worker's core skills to be transformed or become outdated between 2025 and 2030; the figure was about 44% in the 2023 edition and about 57% in 2020. These are estimates by surveyed employers, not a direct measurement of skills. The fastest-growing skills are led by AI and big data, cybersecurity and technology literacy, and the weight of curiosity and lifelong learning is rising. Of employers, 63% name skill gaps as the biggest barrier to transforming their businesses, and 85% plan to prioritise upskilling their workforce. What is needed is a skill that is renewed, not a certificate that is kept.

What works for learning in the first place?

Before the tools comes what research says about methods: in a 2013 review in Psychological Science in the Public Interest, self-testing and spacing study over time came out on top for usefulness, and summarising, highlighting and rereading at the bottom. Hence an AI that examines you is more useful than one that summarises for you. The article "How Should You Learn AI? What the Evidence Says About Effective Learning" gives the review in detail, with an experiment showing how an unrestricted assistant can harm learning.

What did the AI tutor add in the Harvard trial?

In autumn 2023 a randomised crossover trial compared two lessons for students of an introductory physics course for life-science majors at Harvard: one with an AI tutor at home, one in an active-learning class. The analysis covered 194 students. The median post-test score was 4.5 for the AI tutor against 3.5 for the class, the AI group's median gain was more than double that of the other, and students said they were more engaged with it.

The researchers attach important caveats. It was a single lesson in one course at one institution; the measurement was immediate and does not test retention over time; and the AI tutor relied on pre-written step-by-step solutions and on GPT-4, and the authors do not claim the result holds without that design. The lesson is that pedagogical design matters, not that any conversation with a model will teach you.

Four practical habits

This is a practical summary consistent with the above, not the finding of any one trial:

• Set a measurable goal: what do you want to be able to do in a month that you cannot do today?
• Ask the tool to question you before it explains, and to put the questions to you again after a few days.
• Break the skill into weekly steps and apply it in a small project instead of only watching.
• Explain what you learned to someone else, or aloud to yourself, and check your explanation against another source, since models can be confidently wrong.

School education has another side, which we discussed in the article "AI and Education: The End of the Age of Rote Instruction?"

And the certificate?

The evidence does not show that the certificate has lost its value, but that skills change faster than courses. A certificate shows that you studied; a skill is shown by what you make. Seeking knowledge is worship and effort: take the means in organising and reviewing, and ask God for success and blessing in your time.