AlphaFold: the moment AI starts changing science, not just back-office
In November 2020, DeepMind announced AlphaFold 2 results at CASP. Why this matters more than most AI news - and what it changes about how we understand the technology.
At the end of November 2020, DeepMind announced the results of its AlphaFold 2 system at CASP - the Critical Assessment of protein Structure Prediction competition. The results were significant enough that the competition organisers called it a "solution" to the protein structure prediction problem - a problem biology had been working on for fifty years.
I am starting with context deliberately, because without it this sounds like another "AI breakthrough" - and there have been many of those. This is different.
Why this is not just another benchmark
Predicting the three-dimensional structure of a protein from its amino acid sequence is not a game task or a language understanding benchmark. It is a fundamental biology problem, and solving it has direct consequences for drug development, understanding disease, and the capabilities of biotechnology.
Before AlphaFold, experimentally determining the structure of a single protein took months of laboratory work and significant resources. This created a large gap between what we knew about sequences (millions of them) and what we knew about structures (orders of magnitude fewer).
AlphaFold did not just narrow that gap - it changed its nature.
What this means for understanding AI
I pay attention to how business leaders and founders receive news about AI. Most often the conversation stays in the category of "automating something" - replacing operators, classifying documents, chatbots.
These are useful applications. But AlphaFold shows a different dimension: AI as a tool for scientific research, a way to solve problems that previously required laboratory experiments.
That is a fundamentally different class of impact. Not "doing faster what a person used to do", but "solving problems that were not solvable before".
Why the time horizon matters
AlphaFold's results will not change anything in your business tomorrow. That is worth saying honestly. The path from protein structure prediction to a new drug or therapy takes years of clinical trials and regulatory process.
But the thinking horizon matters. If you lead a company in pharmaceuticals, agrobiology, materials science, or chemistry - AlphaFold changes the medium-term landscape of your industry. Not quickly, but genuinely.
And the broader conclusion: AI as a tool for scientific investigation is a direction that will produce results across different industries over the next ten years. This is not hype. It is a slow, long, and structural process.
What to track
A few questions worth keeping in mind:
- Are there problems in your industry where computational prediction could replace expensive experiments?
- Is anyone in your team tracking AI applications in your specific domain - not in a general sense, but in an applied one?
- How do you distinguish "an AI development that matters to me in three years" from "an AI development I can skip"?
AlphaFold is the kind of event that will be called a turning point in a few years' time. Right now is a good moment to understand why.