AI/ML: Overrated or Underutilized?

Dispelling the mystery of how artificial intelligence and machine learning can improve the quality of life and productivity of humans versus replacing them

Ravi Pasula
AI/ML: Overrated or Underutilized?

During a recent virtual leadership offsite, we were discussing whether artificial intelligence and machine learning (AI/ML) is overused. The overwhelming opinion was “yes, it’s overused.” That response sparked a curiosity in me about whether we realize how pervasive the use of AI/ML already is in our digital lives. Working in data science, I am immersed in the use of AI/ML and believe it will continue to accelerate the advancement of many disciplines. While some of my colleagues may believe AI/ML is overused, ESG research shows that 45% of organizations currently have AI projects in production using specialized infrastructure, with another 39% at the piloting/POC stage, as organizations look for smarter and faster ways to gain value from data.[1]

Simplification is key to understanding what’s new to us. In ninth grade, I discovered that relying on standard definitions limited my ability to understand new concepts. When my math tutor asked for my definition of a circle, I gave the textbook answer. He immediately challenged me to think of it differently, revealing a completely new path to discovering math. So, let’s take the mystery out of AI/ML textbook definitions by using simpler terms.

Enable AI at Scale with NVIDIA and Equinix

Businesses are continuing to emphasize the importance of AI adoption as a game changer for the modern business. In this white paper, ESG looks at how Equinix and Nvidia enable AI at scale by leveraging digital well-connected infrastructure and state of the art systems and software for AI workload life cycle.


What AI/ML is and how it shows up in daily life

When I think about AI/ML, it comes down to patterns and psychology. AI/ML uses software and data to identify patterns and build decisions for acting on those patterns. An example of patterns if the day of the week and time of day most people watch their streaming content. Psychology is used in AI/ML software to understand the differences in how humans behave to influence those behaviors. For example, if a certain color is used for a button, will people click more?

Here are the textbook definitions of AI/ML:

  • Artificial intelligence is the science of using and analyzing data, algorithms and programming to perform actions, anticipate problems and learn to adapt to a variety of circumstances by taking new actions.
  • Machine learning uses data to train computers/machines on how to make decisions – in effect, enabling AI.

Let’s stick with patterns and psychology as we take a quick look at common uses of AI/Ml:

  • Zoom uses machine learning in various ways such as automated meeting note taking, allowing participants to fully engage in meetings without the distraction of capturing key information.
  • Netflix uses AI to analyze viewer habits and preferences and then predict other movies they are likely to watch. Title images or photos are customized to reflect those preferences when promoting movies.
  • TikTok uses an algorithm to determine which content is most likely to engage certain viewers and then serves up similar videos or those liked by people with similar preferences.

Today’s use cases represent a fraction of the potential for AI/MI

While much has been accomplished to date, we’re only in the early stages of what’s possible with AI/MI. For example, while there has been a great deal of buzz about robotics, its use has been focused on specific industries such as healthcare, logistics and manufacturing. What about the potential for improving productivity and assisting people in their day-to-day lives? Sure, there are refrigerators that will tell us what we need to buy at the store, but what about more practical assistance? Shopping algorithms used on recommendations engines could also use some fine-tuning.

While there have been advances in AI/ML in healthcare, such as X-rays and diagnostics, there’s much more work to be done. AI for radiology can increase the accuracy and speed of medical diagnostics and assist physicians to diagnose x-rays as well as radiologists. What if pharmaceutical companies could use AI/ML in their R&D efforts to discover the root cause of diseases and develop cutting-edge medicine to replace painful treatments like chemotherapy? Let’s look at a few examples of what companies are already achieving with AI/ML.

How businesses are solving AI/ML challenges

Healthcare researchers at Deep Minds used an AI system to solve a nearly 50-year-old protein-related challenge. AlphaFold, a deep learning system, can now accurately predict the structure of protein within the size of an atom, and do so within days versus months or even years. This system will also reveal what the protein does and how it works.[2]

Another company, OpenAI, is developing what’s called artificial general intelligence (AGI), a form of AI that will be capable of doing anything humans can do. While OpenAI has achieved promising results like acceptable-looking text, their latest release GPT-3, has just 175 billion parameters–which is a thousand times smaller than the human brain, with its 100 trillion parameters, or synapses. While not expected for several years, GPT-4 is expected to be 500 times the size of GPT-3 with about 100 trillion parameters–the equivalent of a human brain.[3]

Advances in AI/ML for robotics are driving the evolution of more sophisticated functions–to augment humans rather than replace them. Collaboration between humans and robots is expected to become a reality with improved sensors, better AI flexibility, and improvements in voice recognition and analysis technologies. Robots will complete routine tasks, giving people more time to focus on what matters to them. For those who require home assistance, robotic companions will eventually provide services such as personal grooming and household chores. Another area of focus will be developing more robotic capabilities to address the shrinking manual labor force.

Applying AI/ML across Equinix

Early on, Equinix Chief Information Officer Milind Wagle recognized the potential for using AI/ML in various functions across the company–to improve productivity and enable data-driven decisions. Planners use AI to forecast power and space capacity in Equinix International Business Exchange™ (IBX®) data centers to ensure customer requirements for specific megawatt thresholds are met. In finance, the use of algorithms has eliminated manual approvals for around two-thirds of the transactions through workflow automation. The marketing team uses AI tools to build models for identifying and winning new customers through analysis of around 100 different input types. Finally, natural language processing (NLP) is used for intelligent ticket routing. These are just a few examples of how AI/ML is currently being applied at Equinix, with more to come.

Exploring the risks of AI/ML

Like anything that can be used for good, AI/ML needs guardrails to constrain the power of AI/ML to protect humanity from the downsides. Left unmanaged, the use of AI/MI can open the potential for:

  • Decision-fatigue from the intake of too much information.
  • Privacy concerns and worker productivity penalties from the disruption of constant notifications.
  • Unintentional biases that result from using outdated data or criteria for evaluation.
  • Malicious intent that influences model behavior for harm.

To protect people from these risks, organizations will need governance and regulations for how AI/ML is used in business and everyday life. AI marketplaces that are blockchain-enabled will become crucial sources of data, allowing companies to track the lineage of data and AI models.

Digital infrastructure accelerates the advancement of AI/ML

Running AI/ML software requires massive amounts of compute power and data–close to where the data is being generated. Take self-driving cars for example. Low latency is critical when it comes to transmitting data to and from these cars, to ensure the necessary reaction time and avoid collisions. Powerful hardware can be provisioned quickly in colocation facilities such as Equinix IBX data centers–directly from Equinix or our partners.

Deploying digital infrastructure on Platform Equinix® and using Equinix Fabric™ for software-defined interconnection enables customers to run their AI/ML applications at the edge and seamlessly transmit data to and from multiple clouds via cloud on-ramps. Automated Bare Metal as a Service makes it easy to replicate digital infrastructure from one of our 240 IBX data centers to any of the 18 global locations where Equinix Metal™ is live–for an edge deployment.

A developing ecosystem of AI solution providers, including hardware, storage, data management  and security providers, makes it easier for customers to access AI as a Service solutions, such as NVIDIA AI Launchpad at Equinix.

We make it easy for customers to ramp up the distributed infrastructure they need to launch and deliver AI/ML capabilities to end users–no matter where those end users are located, by deploying distributed AI Infrastructure and Applications.

To learn more, read the Platform Equinix Vision Paper.



[1] ESG, Enable AI at scale with Nvidia and Equinix, Commissioned by NVIDIA and Equinix, Mike Leone, ESG Senior Analyst, January 2022.

[2] Techstory, DeepMind’s AI Solved a 50-year-old Protein-Related Challenge, Chaavideep Singh, October 31, 2021

[3] Towards Data Science, GPT-4 Will Have 100 Trillion Parameters — 500x the Size of GPT-3, Alberto Romero, September 11, 2021.


Subscribe to the Equinix Blog