Biopharma and AI - promise or concern?
What do you see as the primary reaction between the crossing of Biopharma and emerging AI capabilities? As these two intersect, do you see the future as one with more promise, or one with more concern? If you perceive greater concerns, what key factors might help alleviate those concerns for you?
The question as presented is too broad. In order to answer it one should specify the issues that Biopharma needs to deal with and see where application of AI might be relevant, such as the use of reasoning and knowledge, planning, learning, application of statistical methods, mathematics optimization, etc. It is unlikely that AI will identify and solve problems without being guided by human input.
Coupling AI with CRISPR technology through an open research collaboration and info sharing will speed the development of multiple viable individualized treatments for patients and revolutionize the healthcare delivery infrastructure and insurance business practices. Global tracking of treatments and efficacy will occur at a welcome rate as efforts to treat patients become even more individually specific to hteir needs.
I think you have to view this as another complimentary tool that we can leverage in certain areas of drug discovery. In my opinion AI and other data analytical tools can provide prescriptive guidance in the screening phase for lead identification. The data from multiple omics platforms can be efficiently integrated through modern analytical tools to unravel complex biological networks governing disease in humans. We have a better chance for customizing the drug for specific populations and eliminate the candidates in pipeline early on in the drug discovery journey that are likely to fail.
All these possibilities as listed depend on us having enough information to feed AI to deal with. Human intelligence can sometime "jump" over an "empty space" and come up with a solution beyond what is supported by data; call it an "inspiration", a vision. Can AI do this now, or will be able to do it sometime in future?
AI holds the promise of extrapolation with reinforcement via machine learning to fill in the gaps you mention to the chasm leaps are not so great. In 1968 we used computer optical pattern recognition algorithms to identify object via image scanners. Through allowing feedback to guide the algorithms we were able to add additional "learning" and increase accuracy and predictability. Given the history of the trajectory AI is on findingappropriate and valuable areas to enhance our ability, and the machines, to make better decisions. Adoption follows the accuracy and value curves.