Prediction Machines, Prediction Markets, and the Future of Investing

Sometimes we come across a book that doesn’t just answer a question but changes it. Blaise Agüera y Arcas’s What Is Intelligence? is one of those books. Rather than asking whether artificial intelligence is becoming more human, the book asks whether we have misunderstood intelligence all along.

This may sound like a straightforward question until you realize the depth of author’s view. From the earliest bacteria navigating toward food, to the human brain, to today’s large language models, intelligence is presented not as consciousness or creativity but as the ability to build increasingly accurate models of the future. Organisms survive because they predict. Markets function because investors predict. AI succeeds because it predicts astonishingly well.

Agüera y Arcas is uniquely positioned to make this argument as he is one of Google’s leading AI researchers and a Vice President and Fellow at Google DeepMind, but his career has consistently crossed the boundaries between computer science, biology, neuroscience, mathematics, and philosophy.

One of the book’s most interesting themes is its rehabilitation of cybernetics, the largely forgotten intellectual movement of the 1940s and 1950s that viewed intelligence as continuous feedback between organisms and their environment. Cybernetics largely lost the battle to symbolic AI, where intelligence was defined by logical rules and explicit reasoning. History, however, appears to have sided with the cybernetic thinkers. Large language models did not become useful because engineers successfully encoded the world’s knowledge into rules but because they became remarkably good prediction engines.

For decades, financial markets have been built around the assumption that information creates an advantage. Investors searched for better data, faster news, superior research, or proprietary models. Increasingly, however, those inputs are becoming commodities. Every institutional investor now has access to earnings transcripts, alternative data, expert networks, research assistants, and increasingly capable AI models that can summarize, synthesize, and explain nearly everything available publicly.

If intelligence is prediction, then AI is steadily democratizing one of investing’s most valuable capabilities.

This creates an uncomfortable question. If everyone possesses increasingly sophisticated prediction tools, where does excess return come from?

The answer is unlikely to be faster spreadsheet models or better earnings summaries. Those are becoming abundant. Competitive advantage shifts toward asking better questions, recognizing structural changes earlier than others, understanding incentives, and identifying situations where markets themselves are changing.

Prediction, after all, is only valuable when the future differs from the consensus. This helps explain why investing increasingly resembles what economists call second order thinking. AI may become extraordinarily effective at forecasting outcomes based on existing patterns, but markets constantly evolve precisely because participants adapt. Once everyone predicts one outcome, prices adjust and the opportunity disappears. Investing therefore becomes less about producing forecasts and more about anticipating how everyone else’s forecasts will change.

That dynamic also sheds light on today’s explosion of prediction markets, quantitative investing, and agentic AI. These are not separate trends. They are manifestations of the same underlying force. More participants are using increasingly sophisticated prediction engines to compete over increasingly complex questions. Capital markets themselves become giant feedback systems where humans and machines continuously update one another.

Perhaps the book’s most important lesson for investors is more psychological than technological: For years we have debated whether AI is truly intelligent, whether it understands meaning, or whether it merely predicts the next word. Agüera y Arcas argues that these distinctions may be misleading because prediction itself is the foundation upon which intelligence emerges. Investors should probably stop asking whether AI “thinks” like humans and instead ask whether prediction at machine scale changes market structure.

History suggests that every major technological infrastructure eventually becomes invisible. Electricity ceased to be an industry and became part of every industry. The internet followed the same path. AI may be doing exactly the same. Eventually we will stop talking about AI investing just as we stopped talking about electricity investing. AI will simply become part of investing itself.

As prediction becomes cheaper, judgment becomes more valuable. The scarce resource shifts from computation to interpretation. Machines generate possibilities. Humans decide which possibilities matter.

The greatest investors have always been exceptional prediction machines, although they would never have described themselves that way. Warren Buffett predicts the long term economics of businesses. George Soros predicts reflexive feedback loops. Renaissance Technologies predicts statistical relationships. Venture capitalists predict which founders will reshape industries before the evidence becomes obvious.

Artificial intelligence does not eliminate those activities. It amplifies them while simultaneously raising the bar for everyone involved.

Reading What Is Intelligence? reminds us that perhaps the most important question facing investors is no longer whether AI will transform finance. The more interesting question is what happens when prediction itself becomes abundant.

History suggests that whenever something valuable becomes abundant, something else becomes scarce.

This article was originally published in the Value Hedgehog newsletter by Ehsan Ehsani.