Literature/202504281141 roadblocks in the future of sdls

• Source: [[@canty2025Science acceleration and accessibility with self-driving labs]]
• Tags: #sdl #self-driving-lab #challenges

There are a limited number of successful SDL's in the field, which limits the perception of value for funders.

Looking further ahead, there are severe limitations in the workforce. Most academic curricula do not include trainings for future technicians, users, or developers of automated systems. One of the challenges is that there are not enough pedagogical tools to easily include the topics in the coursework, but also because trainers and teachers are from a different generation, and may not see the value in automation.

Automation is a topic that requires inter-disciplinary knowledge and many departments are just not equipped to handle it. Chemistry is detached from robotics, computer science is separated from material discovery. Moreover, programmers have little exposure to instrumentation, and the other way around.

Today, the lack of specialized people means that there is a large involvement of users as developers and technicians. This limits the development of properly designed systems, since the "quick-fix" approach is still an option. On the other hand, the know-how is in the hand of individuals subject to poor working conditions.

An interesting discussion is IP regulations for ML discoveries, which is not a given. If a new compound is discovered using a ML-based SDL (self-driving labs), who will own the patent over it (if it's even possible to patent it?)

This may yield a low value proposition for industrial partners to actively engage in the development of SDL's as a distributed source of new knowledge creation. (See: 202504281112 SDL approach to distributed discovery).