The first useful jobs for biological computing are likely to be narrow. Current experiments are strongest where adaptation, nonlinear dynamics, biological realism or direct access to living neurons matters more than raw arithmetic throughput.
Neuroscience and drug research
Commercial platforms can expose living neural responses to stimulation and compounds without requiring an intact animal. This is already a clearer use case than replacing conventional processors.
Adaptive control
Closed-loop tasks such as Pong and cart-pole test whether biological networks can change their responses when feedback depends on performance. Robotics is an obvious later target because control systems face continuous, noisy sensory streams.
Reservoir computing
Brainoware and 3D-MIND use neural dynamics as a physical reservoir. This is attractive because the substrate performs a nonlinear temporal transformation while conventional software handles training and readout.
Algorithm discovery
TBC is pursuing a different route: use living neurons experimentally, then translate useful dynamics into software adapters for conventional AI models. This can create commercial value before wetware itself is practical to deploy.
Disease-specific computation
Patient-derived neural tissue could eventually combine computation with disease modelling, allowing researchers to ask how a particular biological network learns or responds to perturbation. That overlaps strongly with personalised neuroscience rather than general-purpose computing.