AI, like Atomic energy, has the potential of being a huge blessing and a huge threat. It provides our healthcare providers with sophisticated useful tools. The FDA has approved 520 AI Enabled Medical Devices and Applications in healthcare. Medical imaging in radiology can improve the efficiency of medical imaging. It relieves radiologists, who can be overwhelmed by the volume of imaging studies, freeing them to devote to time and energy to direct patient care and communications. However, there is reason for concern, given the growth of AI in healthcare, and the impact it will have on patients, their welfare, and the doctor patient relationship. AI provides healthcare providers with sophisticated useful tools. While most communication is based on body language and words exchanged, physicians are given increasingly little time to spend with patients. Will we be able to safeguard and maintain the core principles of bioethics: Autonomy; Beneficence; Nonmaleficence; and Justice? We also need to add principles regarding Privacy, Explainability, Bias, and Trustworthiness:
Trustworthiness involves Organizational and technical principles. There are two aspects to trust in the case of AI: (1) the organization deploying and operating the AI must be transparent, responsible, and accountable, and (2) the AI system itself and its data and output must be verifiable. AI systems developed must be intended for positive purposes, such as improved health outcomes, rather than negative purposes, such as furthering bias, or financial gain.
What about regulations and enforcement to help be sure that the principles are properly utilized. We need to establish guardrails, limitations. Freedom without limits is chaos. Every industry and organization has regulations. The creation and use of any AI system involves multiple layers of people and computer systems. Given that AI is used for good and bad, we want to be sure as to why it is created, who is going to use it, and what impact it will have on our world. The principles must be integrated throughout the creation, usage, and results. Besides technical qualifications, oaths at all levels should include ethical factors. Pertinent ethical laws must be complied with, including the General Data Protection Regulation (GDPR), and the International Organization for Standardization (ISO).
Will we be smart enough to address the need for regulations? Do we have the energy and political will to do something? As with the first Moon Landing, there must be an override button provided when things are headed for a crash landing. We need to raise and address these issues now. When machine learning systems advance much further, it might well be too late.