In The Beginning of Infinity, David Deutsch asks the reader to consider "what it would have taken for scientists [in 1902] to forecast, say, carbon-dioxide emissions for the twentieth century:"
On the (shaky) assumption that energy use would continue to increase by roughly the same exponential factor as before, they could have estimated the resulting increase in emissions. But that estimate would not have included the effects of nuclear power. It could not have, because radioactivity itself had only just been discovered, and would not be harnessed for power until the middle of the century. But suppose that somehow they had been able to foresee that. Then they might have modified their carbon-dioxide forecast, and concluded that emissions could easily be restored to below the 1902 level by the end of the century. But, again, that would only be because they could not possibly foresee the campaign against nuclear power, which would put a stop to its expansion (ironically, on environmental grounds) before it ever became a significant factor in reducing emissions. And so on. Time and again, the unpredictable factor of new human ideas, both good and bad, would make the scientific prediction useless. The same is bound to be true – even more so – of forecasts today for the coming century. (p. 439) (emphasis mine)
Human creativity makes even our most well-reasoned long-term forecasts fundamentally unreliable.
This fact matters most when we rely on long-term predictions to make present decisions. For instance, many people decide what to work on by looking at today's most "important" problem. Some common answers in 2025 America include AI, AI safety, longevity research, defense tech, space, reshoring manufacturing, climate tech, crypto, etc. It is essential to recognize that your decision on the most "important" problem fundamentally makes a prediction about the future, and all predictions are conditional on current knowledge and will be overturned by unpredictable new ideas.
My sense is a better approach in deciding what to work on is following your curiosity. "Important" problems experience hype cycles, like the one currently ongoing with AI. True impact comes from persevering through problems even when they are low status. When the "importance" measure of a problem inevitably fluctuates, following importance rankings will push you toward career switches, whereas curiosity-driven work will allow you to persevere through hype troughs. Curiosity is a more sustainable guide than externally-mediated measures of "importance" which will inevitably fluctuate due to human creativity.