A recent NPR report, by Katia Riddle discusses Ginkgo Bioworks, a Boston-based company attempting to automate parts of scientific research. The company is developing what its founders describe as a “lab of the future,” where AI systems and robots carry out experiments, while human scientists focus on guiding and supervising the work.
The concept goes back almost two decades. According to Riddle, it began nearly twenty years ago with four MIT graduate students who believed that biology would become programmable. Co-founder Jason Kelly says they felt that “programming cells would ultimately be more important than programming computers.” At the time, the vision seemed unrealistic. He recalls early years of limited funding, buying lab equipment secondhand, and struggling to keep the company alive. Interest and investment later grew as AI advanced, including early conversations with Sam Altman before he founded OpenAI that helped attract Silicon Valley support.
Pipetting robots
During a tour of the facility, Riddle describes a lab overlooking Boston Harbor filled with automated systems. One-armed robotic machines sit in glass enclosures, each assigned to separate tasks. A digital display tracks experiments through a color-coded schedule. An overhead rail system moves tools and samples between stations, including petri dishes containing live cells.
Ginkgo Bioworks works across several sectors, including pharmaceuticals, agriculture, and government projects. Current efforts include engineering microbes to improve fertilizer and designing proteins that could influence snow and ice formation. For now, human scientists still translate research ideas into instructions that AI systems and robots execute.
When AI begins to act like a scientist
More recently, the company has begun allowing AI systems to take a more active role in experimental design. Co-founder Reshma Shetty describes a key moment as seeing a lab notebook entry written directly by an AI model. In collaboration with OpenAI, her team used ChatGPT to help design proteins, giving the system responsibilities usually handled by trained researchers.
The results surprised the team. According to Riddle, the approach reduced protein synthesis costs by about 40 percent and enabled more than 30,000 experiments over six months. The findings were published, though they have not yet been peer reviewed. The founders emphasize that humans still define the research goals and constraints. Even so, Shetty says the shift has changed daily work, allowing scientists to focus more on designing experiments that robots can run continuously, including overnight.
Questions about access and risk
Not all experts are comfortable with this shift. Stanford bioengineer Drew Endy warns that AI tools could allow people with little scientific training to run complex biological experiments. He and colleagues have published concerns that such systems could be misused to produce viruses or other biological threats at scale.
Endy told NPR that he supports AI in scientific research but worries about security risks, including potential misuse by state or non-state actors. He argues that regulation is possible and necessary, but it must be developed before serious harm occurs. Riddle notes that biology has traditionally relied on strong expert gatekeeping. Endy cautions that AI could weaken those barriers and shift power toward a smaller number of groups or actors.
Riddle concludes by highlighting a broader debate about the future of science. Jason Kelly told Riddle that he expects scientific research to become more accessible, enabling a wider range of people to explore scientific questions. At the same time, he acknowledges that this shift may bring tension over how science changes when its tools become widely available.
Acknowledgement: This summary was generated using AI writing tools as part of my AI Explorations process. For more, see the original article linked below.