If you are new to the world of Artificial Intelligence (AI) let me attempt to break down the mystery from its underpinnings to what seems to be some kind of intelligent magic box.
First, AI is not magic. It is a machine simulation of human cognitive abilities or human mental processes involved in acquiring knowledge. AI is built on existing technology and methods, new and old. Did you know that AI was first coined in 1956 by a fellow named John McCarthy and his team at the Dartmouth Summer Research Project [1][2]? So, AI is not new, or at least the concept. New and old technology made it possible to finally achieve AI as we know it today. Powerful computers, access to large amounts of data, new methods, and algorithms have come together to build AI. Without any one of these components, there would not be an AI system.
Algorithms are like common everyday recipes with detailed step-by-step instructions to create a fantastic meal. Algorithms are step-by-step programming instructions to create a solution to a problem. Software engineers have developed methodologies and processes to create well-thought-out algorithms to solve problems. One algorithm is called the Random Forest Algorithm, for example. It is a predictive algorithm that facilitates predicting outcomes from data. It is used for classification and regressive tasks and is very versatile in constructing decision trees during the machine-learning process.
What can AI do or accomplish? AI has already been revolutionary, especially in productivity and automation. We are entering a new era where human AI-augmented intelligence is a thing. AI will be weaved throughout the human experience into the future. OpenAI ChatGPT, Google’s Bard AI, Amazon’s Bedrock AI, X Grok AI, and many others are continuing to build on these AI systems. AI systems are going to continue improving to help solve problems and further augment human intelligence.
What can’t AI do or accomplish? AI does not understand what is right or wrong. It does not have a perspective, even though it may seem like it does. AI uses algorithms to analyze the data or user prompts in chatbots for word patterns based on the algorithm’s results and its training of the data. Note that programmers and designers create bias in the data chosen for the AI system, including filtering of the data. The secret sauce again is the algorithms that provide the result and the value back to the end users. Remember AI is only as good as the data it is trained on. Similar AI-generated images can only be as good as the images it’s trained on and how it interprets the objects in the images. The more images the better the AI results will be.
How can you use AI to improve your skills? There will be a realization that AI will eliminate many jobs and reduce many others. For the first time, humans will have to learn to compete against non-humans to generate value. It will be more important to learn or understand AI and how you can use AI to augment your skill sets and intelligence. While AI is amazing at some things, at many others it still gets it wrong. It is important to check the results, and if they turn out to be valid, use it as a reference to create or build something. We should be careful in replacing human intelligence with AI at any level. Otherwise, we make ourselves irrelevant and what matters to each of us for survival.
At the core of AI is Machine Learning (ML). As discussed before, algorithms are used to learn from and analyze data. There is a process called Deep Learning where neural networks are used to emulate the way a human brain processes information. The human brain contains approximately 86 billion neurons. “ChatGPT is one of the largest Large Language Models (LLM) containing 175 billion parameters often referred to as “artificial neurons”. The current version of ChatGPT 4 can process approximately 170 trillion parameters [3]”.
For instance, in a project I was involved with, we aimed to develop a machine learning model for predicting patient length of stay at a Swiss medical facility. This project leveraged two years of historical patient data, although ideally, a broader dataset spanning at least five years would have been preferable. The facility’s transition from a paper-based to an Electronic Medical Records (EMR) system marked the beginning of this data accumulation phase. Again, the more data to train, the more accurate the system becomes.
Therefore, ‘the more data, the better’ holds particularly true in AI development. Insufficient data necessitates the ongoing collection and integration of new information to refine the machine-learning process. In our case, the predictive model initially relied on the available two-year data set but was progressively enhanced with additional data over subsequent years, leading to improved accuracy and reliability. This continuous data feeding is crucial for the maturation and precision of AI systems.
AI’s practical applications are both diverse and revolutionary. In finance, it’s reshaping risk assessment and fraud detection. In my field, Healthcare and Life Sciences AI is assisting physicians in detecting skin cancer and other areas of problems that are difficult to solve, such as predicting the length of stay for the patient based on certain data sets to assist physicians in discharge orders. This approach assists physicians as an input to deciding on discharge orders. EPIC, a healthcare EMR system, is rolling out AI to assist physicians in administration work to help reply to emails while giving physicians options to use the result, edit, or rewrite the email. The hope is that it assists physicians in reclaiming administrative time for patient care. The automotive industry has also leveraged AI for autonomous vehicles and the safety functions of vehicles. AI will continue to create new pathways for innovation in every sector and in our personal lives.
AI will inevitably continue to grow in ways we have not thought of or cannot comprehend today. As a society, there will need to be ethical considerations for AI projects. There are valid issues of privacy, bias in algorithms, and the impact on jobs, which are just a few of the challenges we face. Government regulations will be necessary for AI from a public safety perspective. The EU is pioneering AI regulations, while the US is developing regulations for AI in the US. The EU AI regulations are aimed at safety and fundamental rights, and they are transparent [4]. There is a classification system that sets up a risk level where the high-risk systems are held to strict requirements and other restrictions, such as bans on social scoring and untargeted facial recognition. There are significant fines for noncompliant AI system development.
Artificial Intelligence is not just a technological trend; it’s a paradigm shift in how we interact with technology and data. As professionals, embracing AI’s potential and contributing to its ethical development is vital. The journey of AI is one of exploration and responsibility, and it invites us all to be a part of its narrative and evolution.
While exciting, AI will be shaped by our decisions and actions today. AI will present challenges in society. The best path forward for the individual will be to stay informed, engage in continuous learning, and participate in ethical discussions to help shape the future of AI.
References:
- Dartmouth College. (n.d.). Artificial Intelligence (AI) Coined at Dartmouth. Retrieved from https://home.dartmouth.edu/about/artificial-intelligence-ai-coined-dartmouth.
- Wikipedia. (n.d.). History of Artificial Intelligence. Retrieved from https://en.wikipedia.org/wiki/History_of_artificial_intelligence).
- EcoAGI. (2023). ChatGPT Parameters. Retrieved from https://ecoagi.ai/articles/chatgpt-parameters.
- EU Artificial Intelligence Act: deal on comprehensive rules for trustworthy AI https://www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificial-intelligence-act-deal-on-comprehensive-rules-for-trustworthy-ai