Martin Shkreli argues that the primary bottleneck in drug development isn't finding new molecules, but the immense inefficiency caused by poor communication, irrational decision-making, and misaligned incentives across numerous human departments. He believes AI's greatest contribution will be optimizing this complex organizational process rather than just improving discovery.

The paradigm for drug development is shifting from being "first" or "best" in a category to being "last in class." Using AI, the goal is to design a molecule with such intentionality and specificity that it becomes the final, definitive therapeutic for a disease, rendering subsequent improvements unnecessary.

The catastrophic failure rate in drug development isn't just bad luck; it's a structural problem. It originates from the very first decision: researchers, biased by existing literature and simplistic models, fixate on a single biochemical target, ignoring the body's complex, multi-faceted nature.

Unlike coding, where AI models get immediate feedback on whether code runs, drug development faces immense delays. A biological hypothesis can take a decade and hundreds of millions of dollars to test in the clinic. This lack of rapid validation checkpoints is a core obstacle for AI's ability to learn and reliably improve drug target selection.

While AI for novel drug discovery has lofty goals, its most practical value lies in accelerating development. This includes applying AI to de-risked assets for new indications, improving delivery methods, and designing faster, more effective clinical trials, which is where the real bottleneck lies.

While AI is on the verge of cracking preclinical challenges, the biggest problem is the high drug failure rate in human trials. The next wave of innovation will use AI to design molecules for properties that predict human efficacy, addressing the fundamental reason drugs fail late-stage.

Beyond accelerating timelines, AI's real value lies in its ability to design molecules for targets previously considered 'hard-to-drug.' These models operate on different principles than traditional lab methods and are indifferent to historical challenges, opening up entirely new therapeutic possibilities.

The current, tangible breakthrough for AI in drug discovery is not identifying completely novel biological targets. Instead, it's rapidly designing effective molecules for known targets that have historically been considered "undruggable," compressing years of screening work into a month.

The immediate goal for AI in drug design is finding initial "hits" for difficult targets. The true endgame, however, is to train models on manufacturability data—like solubility and stability—so they can generate molecules that are already optimized, drastically compressing the development timeline.