You’re doing cloud-based AI and machine learning wrong

Rackspace Technology just announced the results of a global survey that reveals that the majority of organizations lack the internal resources to support critical AI and machine learning initiatives. Indeed, 34% of respondents reported artificial intelligence projects that failed.

The larger issue is the misapplication of AI and ML for applications where these particular technologies are contraindicated. This has been a problem since the advent of neural networks and AI, which is much longer than you think.

AI on public clouds is just too easy and cheap not to leverage, so it’s being used with business applications where the process of learning or finding patterns is not a requirement. When AI is the shiny new hammer, every application looks like a nail.

Applications that are good candidates for AI or ML are those that need to determine and assign meaning to patterns. Think of the systems employed now on factory floors to determine product quality using image recognition and automation, or fraud detection programs in banking that look at transaction data.

A second problem is the lack of training data to support the use of AI and ML. Data teaches the AI engine to assign meaning to patterns, and your AI engine is only as good as the training data available.

These days enterprises often don’t understand where the training data is located, what the single source of truth is, or what the data means. Data is everything in the world of AI; knowledge is derived from data. If you don’t have a solid data source, and you don’t have an excellent understanding of the meaning of the data, AI won’t work for you.

Copyright © 2021 IDG Communications, Inc.

Source link