How care can beat brute force in technology

Over the summer I dipped my toe back in the festival scene by booking tickets to a handful of outdoor shows.

When the weather’s good, I’m happy to make an early start, avoid the traffic and crowds, and settle in slowly. It allows for a relaxed arrival, helps with finding a good spot and means the picnic is more ‘tea’ than ‘supper’.

Meanwhile, in the background, we were serenaded by the sound engineers doing their customary checks.

Often, Donald Fagen’s classic album The Nightfly makes an appearance in these sound checks. Mostly it’s because it’s so well engineered that any distortions can be assumed to be in the local sound rig rather than in the source material.

The importance of the right controls in AI and other technologies

But buried inside that simple audio test are some profound observations that can be applied to the current and future state of the world. Hearing Fagen’s I.G.Y. track recently and given all the current attention around artificial intelligence, I was particularly taken by its reference to the benefits of technology being controlled by caring people.

Look up the song for yourself. They’re grand thoughts and, if nothing else, prove the vision of AI has been around for a while (the album is from 1982).

But for all the recent advances, are these still just aspirations rather than expected destinations?

It feels a long way from the current state of AI.

That said, perhaps it’s all too easy to be cynical and dismissive.

One of the features of interacting with any AI tool is the feeling of impartiality it brings. Human intelligence – colleagues, customers, suppliers – can all be political, ducking and diving for their own ends, implicitly if not overtly. In comparison, an AI chat can sometimes feel positively open.

And for all the (warranted) criticisms thrown at the heads of AI companies, I know many programmers within them do have good intentions and foresight.

Why context matters

And while it’s crucial for humans to be in control of big decisions, if the 21st century minutiae of your life could be handled for you, would that be such a bad thing?

As it always is, the crux of those questions is data or, more accurately, information. That distinction’s important. What matters isn’t just the raw values in fields and rows, paragraphs and texts. Instead, it’s the links to those elements that are critical. They provide the context for the values, showing why a measurement might have the value it does, what constraints applied and what followed.

Without a context, the values are simply that – raw numbers representing arbitrary measurements, something that might explain some of the AI ‘hallucinations’ you might have heard about.

But with the context, the values are much more, well, valuable (forgive the pun) and can help:

• Determine if the value was expected or not, and within a tolerance – perhaps helping with fair decisions
• To compare with what was seen before in similar situations – giving a considerate view
• Establish a benchmark for future processing – at a stretch, to ensure freedom from past mistakes

Such a context makes it sounds like you ‘simply’ need to collect as much data as possible and then refine it to identify what’s needed to provide ‘dimensions’.

The pitfalls of using brute force with artificial intelligence

Current AI learning techniques suggest brute force might be the answer. But this feels awfully similar to the early days of internet searching, where results were returned almost on spec. Instead, to identify the real value in data, you need more than just an awful lot of it. It must be collected and processed consistently, which, in turn, implies use of a framework. In part this is a ‘paper’ one to define the core elements of the data but it can also be ‘binary’ to help with its collection.

This is exactly how financial accounting systems are (and have been) designed and implemented – now using IT but before that courtesy of Bob Cratchit’s quill pen. Ledgers and journals define a common pattern by which all accounting activity can be recorded. And, from that pattern, rules for its presentation and validation (double entry) can be expressed simply and across all business processes.

But in other business areas there’s a tendency to use custom patterns and implementations. In effect this means reinventing the wheel for every project. And not just how to make the ‘wheel’ but before that rediscovering how to design it, and maybe even that one is needed at all.

How Quantiv can help

Our NumberWorks method and NumberCloud platform take a different approach. We borrow from – and extend – those accounting principles. We use them to define and collect information in many different business areas. We do this by using a common grammar to express the patterns but then use a custom vocabulary to accommodate the varying data present in different business processes.

And while our approach might not quite reach the vision referred to in Fagen’s music, we’d like to think the machines that result could be more virtuous, and so could provide better help with decisions large and small.

Contact us to find out more.