The discussion argues that AI is valuable in focused areas such as protein binding, tumor analysis, patient selection, and experimental design, but biology remains difficult to predict because it involves dynamic interactions across cells, organisms, and environments. Unlike chemistry, biological outcomes are shaped by mutations, immune responses, pharmacokinetics, and context that current models do not reliably capture. The central bottleneck may therefore be measurement: researchers need richer datasets, better assays, improved instruments, and validated systems such as organoids before increasingly powerful models can produce dependable results. AI co-scientists could help generate hypotheses, troubleshoot experiments, and design new tools, but they cannot replace physical testing or solve the challenge of translating animal and laboratory findings into successful human therapies. The speakers also challenge biosecurity efforts that focus primarily on speculative AI-enabled pathogens, emphasizing more immediate risks from outbreaks, agricultural disease, laboratory failures, and weak infrastructure. They support proportionate DNA screening, customer verification, and oversight while warning that excessive regulation could suppress innovation and weaken biotech competitiveness. Progress ultimately requires sustained funding for both biological research and practical safety measures.