The four pillars of good AI

By

Home > News > The four pillars of good AI.

At FIDO we believe in good AI. But what does this mean? What makes good AI? We believe that there are four pillars that determine the quality of AI: truth, accuracy, speed and intelligence.

Good AI is based on verified data, and must give repetitive accurate results in real time. It should learn on a feedback loop and start to create associations you may not otherwise have seen.

Only, when all these four pillars are present, is AI truly effective.

1. Truth

Let’s start with truth.

AI must be built on absolute truth. That means verified data. This must be established in order to create your training library. A training library built on unverified data is essentially built on sand and likely to be flawed and biased.

United Utilities’ openness and cooperation with us through the Innovation Lab provided us with the access to verified data that we needed in order to train FIDO AI.

All of the 1.8 million plus files that have been fed into our training library have been physically verified, so we know without a doubt that our AI is built on absolute truths.

2. Accuracy

This enables a high degree of accuracy. Accuracy is essential to good AI.

Even the best and most experienced humans can only detect leaks with 60% accuracy. With FIDO AI, it’s over 92%, and with the AI constantly learning through a feedback loop, its only improving.

When implemented, AI can eliminate human error. It ensures that you know with confidence what is and, importantly, what isn’t a leak, reducing false positives.

3. Speed

Good AI also has to have speed.

Compared to a human, who can analyse around six files an hour, FIDO AI can analyse thousands. We analysed 37,500 files for Thames Water in 2hrs 30. This means you can get your analysis in real time. By making it automatic, we can rapidly send engineers out to the right locations, without waiting for a human to make that decision.

4. Intelligence

Finally, good AI must be intelligent.

Unlike humans, AI retains the data from previous leaks and learns from it. This enables it to go beyond identifying a leak – it can also identify leak size and pipe material.

Intelligent AI turns your leaks into assets as it builds up an accurate data repository, meaning that you can start to predict where leaks will occur.

AI is the future of leak detection. But not just any AI; good AI. Good AI which stands on the four pillars of truth, accuracy, speed and intelligence.