The claim that there are no new ideas in AI, only new datasets, is both provocative and partly true. As Jack Morris argued in his recent post, many of the most important AI milestones have been driven not by theoretical leaps, but by new sources of data. He puts it succinctly:
“The breakthroughs weren’t big ideas; they were new ways to learn from new kinds of data.” from blog.jxmo.io
That framing resonates with history. The ImageNet dataset unlocked computer vision. Web-scale text collections made Transformers viable. Reinforcement learning from human feedback reshaped how models align with our preferences. Time and again, progress has come from scaling access to structured, abundant inputs rather than from fundamentally novel algorithms.
But the story is incomplete if we stop there. Innovation in AI is multidimensional, and algorithms still matter. DeepMind’s AlphaEvolve has already demonstrated the ability to generate algorithms beyond human expertise, producing approaches to matrix multiplication and optimization problems that surprised even expert researchers, as reported in Wired. If data were the only driver, we would not see systems out-innovating decades of human design.
As Dr. Rumman Chowdhury cautions:
“Innovation stems from human minds, not AI. Don’t delegate your thinking to machines.” from Business Insider
Her warning underscores the deeper point: humans frame the problems, interpret the outputs, and decide which ideas matter. AI extends our reach, but it does not replace human originality. Even when AI generates unexpected solutions, the spark of innovation is in how people direct, interpret, and apply them.
Morris’s post highlights an important truth: data has been the engine of AI’s rise. But the full picture includes algorithms, architectures, and above all, human creativity. Ignoring these other drivers risks oversimplifying how innovation actually happens.