Best Uses of AI Outside a Screen

Waymo’s system drives paying passengers on public roads without a safety driver in several cities. It is one of the clearest cases of AI performing difficult physical work at scale.

AlphaFold made accurate protein-structure prediction dramatically faster and opened its database to researchers. It remains the benchmark for AI that changes scientific work rather than office software.

John Deere’s See & Spray cameras distinguish crops from weeds and target herbicide only where needed. The system can reduce chemical use without asking farmers to sacrifice coverage.

Paige analyzes digitized pathology slides to help clinicians detect and classify cancer. The system is useful as a second set of eyes in work where overlooked details have serious consequences.

Figure is testing humanoid robots on factory tasks that require movement, grasping and repeated physical judgment. We care more about those work trials than choreographed demonstrations on social media.

Zipline combines autonomous aircraft, routing software and precise delivery drops for medical supplies and retail orders. Years of real operations make it more persuasive than a drone demonstration.

Carbon Robotics identifies weeds with computer vision and destroys them using lasers. The LaserWeeder replaces repeated chemical treatment with a machine farmers can put directly into a field.

Anthropic’s hardware standard gives models a defined way to operate microscopes and robotic equipment while limiting unsafe actions. It addresses the unglamorous interface between intelligence and machinery.

Isomorphic Labs applies machine learning to protein structures and drug design, then tests the resulting molecules through pharmaceutical partnerships. The value will be measured in candidates, not chatbot fluency.

Form Energy uses software to coordinate long-duration iron-air batteries with changing grid demand. Intelligent control matters because the hardware stores electricity over days rather than a few hours.