Over a million developers have joined DZone.

4 Artificial Intelligence Pitfalls

DZone's Guide to

4 Artificial Intelligence Pitfalls

There is no point-and-click, off-the-shelf AI software that 'makes my data center work better.' Organizations that share this feeling will be set up for failure.

· AI Zone ·
Free Resource

Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.

No one could argue that life hasn’t been more convenient since the initial breakthrough of artificial intelligence (AI) in 2012. The world is moving faster and faster, driven by technology and innovation. From speech recognition to facial recognition allowing you to unlock the new iPhone with your face to cars self-navigating the streets in Silicon Valley, AI is all around us. And if a seating chart is telling for a tech company's priorities, Google’s AI researchers were recently moved to sit near the boss in its Silicon Valley office. However, with the hype and noise around AI, it’s easy for companies to take missteps along the way when trying to take advantage of the technology. As it’s always a good idea to exercise caution when deploying an AI-driven service, here are a few common pitfalls to watch out for:

1. Being Shielded by the Hype

Adopters need to look beyond the hype to accurately judge AI's benefits, as it’s far too easy for organizations to underestimate the time, knowledge, and data required to effectively implement AI systems. In many cases, organizations encounter a major pitfall by handing over decision-making power to AI too early after implementation. It’s common to be blinded by the hype and jump fully into AI, but it’s important to give AI technology the time to learn and grow into its environment. It’s also critical to build up the AI system’s success before it’s trusted to make decisions and take actions.

2. Lacking IT Team to Manage AI Effectively

While AI is dominating 2018 trend predictions from analysts including Gartner, Forrester, and IDC, it is important to remember that it’s not a new technology. This is where the fundamental truths of networking from RFC 1925 are important to remember. Specifically, Truth 11 comes into play: Every old idea will be proposed again with a different name and a different presentation, regardless of whether it works.

A major pitfall is lacking an IT team that has the expertise to effectively manage the AI system and interpret insights to their maximum value. AI in data centers gives organizations powerful capabilities and insights, but without a team to manage the system and leverage those insights, organizations will be less likely to take full advantage of AI.

IT organizations of all sizes need more and more automation to keep up with the growth of compute and the increasingly distributed nature of compute resources. This does not mean that you need incredibly complex algorithms based off of neural nets to get this increased efficiency. Having effective data collectors that feed into a condition system is a great way to get good data into the system. Coupling that with technologies like state machines that take action on the changing and relevant conditions is a very effective building block for creating a self-healing datacenter. The key throughout all of this, however, will be ensuring organizations have an IT team that can manage these AI systems and bring insights to the information gathered.

3. Trying to Keep Up With Google

Organizations also can’t fall into the trap of comparing themselves to Google. Google used the DeepMind AI engine to make their data centers more efficient. They incorporated a system of neural networks, but effectively using neural networks requires a firm grasp of the mechanics. You have to have extensive training and huge test sets to validate the data before you ever put it into production. Developing and utilizing neural networks the correct way requires a ton of expertise and computing resources, something that Google clearly has. The same cannot be said for most other IT organizations today, but that doesn’t mean to scrap your AI ambitions.  

4. Using AI to Answer All of an Organization’s Problems

It's beneficial to use caution when deploying an AI-driven service. While there is this notion that AI is here to save the day, it certainly helps — but is not the end-all, be-all answer to the problem. There is no point-and-click, off-the-shelf AI software that “makes my data center work better.” Organizations that share this feeling will be set up for failure. It’s key to incorporate AI into facets of your organization, but utilize other key data center technologies alongside it. 

Your machine learning project needs enormous amounts of training data to get to a production-ready confidence level. Get a checklist approach to assembling the combination of technology, workforce and project management skills you’ll need to prepare your own training data.

ai ,machine learning ,deep learning ,google

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}