AI is the hot new thing – in healthtech, and everywhere else, as far as I can tell. But I’ve noticed many people using the term “AI” to mean different things. I think this comes down to artificial intelligence being such a broad term.
What can AI do?
The major categories I’m noticing are:
- Automation
- Pattern recognition
- Generation
1. Automation refers to setting up systems, where things that happen on a regular basis can happen without human action. For example, automating an email to send every Monday, or every time a person takes a certain action.
2. Pattern recognition is when an AI is trained to identify specific terms or visual patterns. For example, pattern recognition is what allows AI to read MRI scans. This could also be considered “analysis”; identifying what to do based on a pattern.
3. Generation is any situation where AI is creating content, however it’s typically thought of in terms of chatbots or other formats where an AI might speak directly to an individual, without being reviewed and approved by professionals.
But “what can we do with AI” is rarely the question we should be discussing. The real question is: what should we do we with AI?
What should we do with AI today
More accurately, the question is “what should we do with AI today“?
Our technical abilities are developing rapidly. Already we use AI to automate work – think of sending batch emails based on broad triggers. And we use it to analyze patterns more and more, whether in “spell check” situations or MRIs.
But generation? That’s trickier. Our ability to design for the ethical considerations, such as designing for anti-bias and inclusivity, are expanding. But we aren’t there yet. An AI will still generate “hallucinations“, where they invent facts. They will tell patients misleading or inaccurate information – which can be frustrating at best, and deadly at worst.
AI can easily automate tasks that take humans a frustratingly long amount of time. It can partner with humans and assist in time-consuming tasks like initial pattern recognition and analysis, with humans finalizing the reviews. AI can even generate draft content in valuable ways that save humans’ initial brainstorming time.
Perhaps in the future, we’ll use AI to do all of our content generation. As content designers we’ll focus on training the systems. Imagine a world where a content designer could:
- Scale – working on evaluation criteria and writing 5-50 sample content strings, rather than 1000s of individual strings for personalized experiences (without needing to triple check each one!).
- Experiment – spending much more time testing out what users respond well to, and digging into why.
- Focus – writing more for audiences that don’t respond to what works for “most” people.
We’re not quite there yet, but we can get there in the near future if we build AI appropriately. AI is a tool, and a useful one. What’s key is to use the tool appropriately.
Marli Mesibov: Content Strategist & Writer 