Artificial intelligence is the umbrella term for computer systems that can interpret, analyze, and learn from data in ways similar to human cognition. The field of AI is vast, encapsulating numerous subfields and applications related to machine intelligence. With AI, computers can perform a wide range of tasks—from playing chess to diagnosing cancer and virtually everything in between.
The term artificial intelligence was first introduced by American computer scientist John McCarthy in 1956 at a summer conference at Dartmouth College in New Hampshire. That conference is believed by many to have launched AI as a genuine field of research. In the ensuing decades, a number of inventions, discoveries, and experiments have led to the many ways AI turns data into insights, powering our society and influencing how we use computers every day.
With AI and machine learning, computers are programmed or “trained” to perform intelligent tasks. Tasks that are either “narrow” or “general.” Artificial narrow intelligence or weak AI pertains to specific, pre-defined tasks such as predicting the weather, recommending your favorite music, or even autonomous driving. Narrow AI can on its own transform the way we treat a particular process or task. Most of what we see today in terms of machine intelligence falls within this category and shouldn’t be taken for granted. Narrow AI is capable of analyzing massive volumes of data, thousands of times faster than people and typically with fewer errors. Narrow AI also relieves us of mundane tasks so that we can be more efficient with our time.
Artificial general intelligence (AGI) or strong AI is related to more complex functionality that is expected to match human level capabilities across multiple domains. Think about the very advanced AI systems you see in sci-fi movies where the interactions between people and machines are seamless and feel conscious. An AGI system can draw valuable insights from diverse data sets (e.g. images, text, audio files, logs) and use cognitive computing to perform functions that are indistinguishable from those performed by a human.
As described in one of our prior blogs, traditional data analytics and machine learning differ in several key ways, including structure, purpose, and benefits. Without diving into too many details, in short, traditional data analysis is descriptive and quite useful in explaining current or historical data while machine learning is predictive and capable of learning from data in ways that provide valuable insights and recommendations.
AI/machine learning is a dynamic process, often requiring algorithmic model training, validation, testing, refinement, and integration with other software components to create real value. Unlike many other engineering functions such as traditional software engineering, where you can create a solution based on certain known requirements, quality machine learning requires deep model and data exploration to arrive at something useful. Simply put, experimentation and embracing the unknown is par for the course in advanced AI.
Building models and proper orchestration are also core to success here. Added complexity sets in when the intended use case is multimodal and the data requires multimodal AI processing, creative ensembling of multiple models, and intricate queuing and software orchestration. This is where combinatorial expertise in machine learning, compute infrastructure, and software engineering is needed but currently in rare supply.
Then there are the 4 Vs of data which are important criteria for success in advanced AI initiatives. The 4 Vs include data volume, variety, velocity, and veracity. The road from data to insights can be patchy and long, requiring many types of expertise. Dealing with the 4 Vs early in the exploration process can help accelerate discovery and unlock otherwise hidden value.
It is also important to note that high accuracy and precision in artificial intelligence is the byproduct of rigorous scientific, engineering, and design efforts. This is where advance science meets art to deliver results. And the journey from ideation to implementation for even a single AI application requires cooperation with other contributors including those fluent in business, operations, legal, and cybersecurity—18 skills in all.
For a quick refresher on key AI terminology be sure to read the 53 useful terms in the world of artificial intelligence.