We’ve heard it all before: artificial intelligence (AI) is the future. But as of now, we are still far(ish) from building technology that can develop its own knowledge. Machine learning can imitate the way humans learn, thereby gradually improving its accuracy.
Machine learning vs AI
In its simplest form, artificial intelligence is a field in which technology can think for itself. According to Stuart Russell and Peter Norvig, this can be defined in four potential ways:
- Systems that think like humans
- Systems that act like humans
- Systems that think rationally
- Systems that act rationally
I know what you’re thinking: “What’s this got to do with data?”
The answer is a lot. Although our current technology doesn’t match the definitions from above, subfields of AI are relevant to what we use today. From speech recognition to self-driving cars, machine learning is used to innovate technology in many ways. And data is critical to making this technology work.
So, what is it?
Machine learning is a branch of AI that uses data and algorithms to imitate how humans learn. It’s programmed to look at data in a certain way and, with advanced models, build on that data.
The learning system of a machine learning algorithm can be broken down into three main parts:
- A decision process: The algorithm takes in a selection of data and produces an estimate about a pattern within it.
- An error function: If there are known examples, this function will compare its results to evaluate how good the guess was.
- An optimisation process: The algorithm looks at where mistakes were made and updates the decision process to enhance future outcomes.
This system can run on repeat to constantly optimise results. In short: looking at data, learning from data, and building on those learnings.
How does it affect data optimisation?
AI tools like ChatGPT and Salesforce are making headlines, reeling businesses in with their ability to accelerate long and tedious tasks. This trend is also making waves in the world of data optimisation.
New machine learning tools are emerging to correlate data in the right way more efficiently. This can be used for customer targeting and supporting different levels of experiments.
Take a multi-armed bandit experiment, for example. I imagine a future where CMS providers have intelligent content management supported by machine learning. Content editors can create multiple titles, upload multiple images, or even add alternative components or layouts to a page. Add a sprinkling of machine learning, and your CMS will automatically and continuously test the best combinations, even identifying cohorts of users and showing each their own most effective combination. Your website may never look the same on two different devices!
In this instance, a machine learning tool can help trial smaller rapid changes that aren’t usually tested due to time or resources.
When is it worth it?
To really make the most of machine learning, you need to have a big supply of data to build on. Big data sets will help find opportunities to really make a significant impact. The less data you have, the less helpful machine learning tools will be.
Another thing to consider is that these tools are not a ‘one size fits all’. Many factors play a part in what can make them successful: your industry, the type of data you’re dealing with, and the results you’re trying to achieve. Depending on these, there’s a different approach to match the problem you need to solve.
Is this going to change the way we work?
Short-term? It depends.
No need to worry; the robots aren’t taking over…for now. Commercial machine learning isn’t capable of original thought, so it needs someone operating the levers.
The most significant shift this technology will bring is the skill set required when working with data. Machine learning removes the human step of analysing large data sets to identify patterns, which allows us mere mortals to focus on analysis.
Machine learning tools will only be as valuable to your company as those using them. Yes, they might be shiny new toys, but they will not magically find solutions independently. You need to know your business problems and apply the right tools to solve them.