|
By Steve Berkowitz
In a previous blog, I discussed the first step in the critical analysis of artificial intelligence in healthcare—an operational or working definition of artificial intelligence. I would now like to take that definition and take the next step—develop a descriptive model of AI that would illustrate the evolution of artificial intelligence and help us better understand the future of AI in healthcare. The model I propose has three components. Each one can evolve separately as well as influence the progress of the others. The ultimate AI application or outcome is a function of the effectiveness of all three:
HARDWARE – THE RAPID DEVELOPMENT OF PROCESSING POWER When I talk about the evolution of the individual processing unit, I am primarily referring to the unprecedented progress that has been made in the past sixty years in terms of fulfilling the mantra: faster, smaller, cheaper. Moore’s Law, coined in 1965 by Gordon Moore, an engineer at IBM, stated that the number of transistors in an integrated circuit will double about every two years. This doubling process has continued unabated to this day. Processing speed and efficiency are essential in the ultimate evolution of AI, and until very recently, it has been the rate-limiting step in the development of advanced capabilities such as language processing and complex machine learning. It has only been in the last five or so years that we have had the processing capability to have such entities as GPT and deep learning applications. Now that this limit has been attained and exceeded, true advanced intelligence is operationally and commercially feasible. One can only imagine what will be possible as this capability continues to double. The recent development of quantum computers, for example, which rely on using quantum mechanics and subatomic particles to power the processing, is currently in the research stage, but it promises to further jumpstart this cycle of “faster, smaller, cheaper.” AI will operate more effectively, a million times faster and more efficient. CONNECTIVITY – INTEGRATION OF INDIVIDUAL UNIT INTO NEURAL NETWORKS AND THE INTERNET Put simply, not only is the individual processor becoming faster, smaller, cheaper at exponential rates as mentioned above, but each individual computer unit now has the capacity to become interfaced and connected to multiple units and form large computer networks. These neural networks have the potential to give each computer access to enormous amounts of data, including becoming part of the internet of things, or the “cloud.” The ability to work as large systems makes the AI process much more robust. All systems would have the potential to communicate with each other and learn from each other… and even influence each other. Neural networks using artificial intelligence resemble the human brain, in which thousands or millions of individual units become interconnected and organized into layers. The output of one of the individual entities is now the input of another entity. This synergistic interaction further empowers AI to do extremely complicated processing such as deep learning and to do so faster, smaller, and cheaper. SOFTWARE – THE EVOLUTION OF MACHINE LEARNING AND COMPLEX REASONING Now that the hurdles of requisite speed and power as well as the ability to interconnect on complex levels have been considered, the model now moves on to the actual application of learning algorithms and logic models. As algorithms and systems become more sophisticated, more complex logic that simulates or even exceeds human logic, are incorporated into the model. Supervised learning, where models, trained with labeled data sets and pre-set algorithms, can now progress to unsupervised learning, and even reinforced learning. This allows the AI to evolve and move beyond the initial data sets and algorithms. As I mentioned in my last blog, not only can this learning bear enormous benefits, but it also has the downside of potential negative effects of emergent properties, and even hallucinations. Not only can the model readily incorporate the most basic form of reasoning, deductive reasoning (going from the general to the specific), but it also can begin to develop more complicated patterns of inductive reasoning (going from the specific to the general). The potential further exists to incorporate or “learn” more advanced forms of the human reasoning process such as creativity, emotions, and even empathy. AI has not acquired these traits yet, but if the human brain can learn these traits, why couldn’t AI too? And do so more quickly and efficiently. I believe we will see increasing complexities of AI resulting in applications that combine deductive, inductive, and more creative forms of reasoning. Once again, we can only imagine what the ultimate applications will be able to accomplish as the three stages of hardware, interconnectivity, and software come together and mutually advance. MODELING THE FUTURE OF AI IN HEALTHCARE As healthcare executives better understand the application of AI, it is valuable to conceptualize the model into these stages in looking to the future. I hope this model gives us a better framework in which to analyze the evolution and future of artificial intelligence, particularly AI in healthcare. Putting the model back together, we see there are three major components: the hardware, connectivity, and the software. All three will continue to quickly evolve, and work collectively and synergistically to improve the ultimate sophistication of the application outputs. Products that now flood the media daily have resulted from the evolution and development of these three stages. Planning your next event? Get in touch with us at the Capitol City Speakers Bureau today to schedule your ideal speaker and make your event a success!
0 Comments
Leave a Reply. |
Archives
June 2024
Categories
All
|
RSS Feed