Proteins are the building blocks of life, and their shapes dictate their functions. After cracking protein folding prediction last year, AI researchers now aim to design entirely new proteins to specification. Rather than tedious lab trial-and-error, algorithms suggest protein arrangements with desired traits for anything from laundry detergent enzymes to biologic drugs.

DeepMind's Breakthrough Pioneering the field, DeepMind revealed AlphaFold in 2021, able to predict unknown protein shapes more accurately than expensive lab techniques. They recently unveiled AlphaFold DB - open-sourcing structures for 200 million protein sequences. This trove of data now fuels designing custom proteins.

Accelerating R&D Specialized AI models can create protein blueprints targeting specific functions like catalyzing chemical reactions or binding to antigens. AI is expected to slash the typical antibody drug development timeline by years. Startups like Insilico Medicine pioneer such real-world applications.

Combine AI and Experiments Of course, biological activity remains complex with unknowns. After using AI to propose protein designs, confirmation still requires real-world synthesis and experimentation. With AlphaFold reducing guesses to test, machine learning makes the full process vastly more efficient.

What Are the Risks? As with any dual-use technology, protein engineering does carry risks if misapplied. well-intentioned work might accidentally yield dangerous organisms. Monitoring and governance becomes vital as capabilities improve.

Redefine Bioengineering? Further out, some researchers speculate advanced AI could redefine bioengineering - or expose unintended consequences. As algorithms grow beyond human understanding, what emerges from unleashing such generative models into the building blocks of life? The implications are profound either way. For now, all eyes look ahead with guarded optimism.