As early as 1966, we built machines that followed simple rules, like digital parrots mimicking conversation (ELIZA). They could repeat, but not comprehend.
Meanwhile, in the lab, a different language revolution was unfolding.
Starting in the 1970s with Sanger's methods, we first learned to read the individual letters of life's code—the A, T, C, and G of DNA.
In the 1990s, machines learned to count with n-gram models, predicting the next word based on statistics. They understood the grammar of a sentence, but not the meaning of the story.
Soon, in late 20th and early 21st centuries, we were reading entire sentences and paragraphs—mapping genomes, profiling gene expression (transcriptomes), and identifying the vast library of proteins (proteomes).
Then, in 2017, the world was shattered by a five-word sentence: Attention is all you need. It was the revolution of the Transformer, and for the first time, it gave a machine the power to grasp context, connecting the beginning of a story to its end.
Humanity taught machines to speak.
Then, to write.
Now, to reason.
We generated oceans of data. But there was a problem. Each dataset spoke a different biological language.
We had powerful tools to analyze DNA. Separate tools for proteins. And yet more for cell expression profiles. They were brilliant, but monolingual. They were isolated in digital silos.
Biology's greatest insights were getting lost in translation.
Research was stuck in a slow, frustrating cycle:
Fragmented Models
Tools could only see one piece of the puzzle at a time.
Data Scarcity
Each biological "language" had a limited vocabulary, stunting AI growth.
"Wet-Dry Lab" Divide
A massive gap existed between computational prediction and real-world experimentation.