Testing and experimentation
There’s still no substitute for human intelligence to evaluate whether a product will resonate with humans. But to get started and get a rough idea of what you’re looking for, or for smaller brands to evaluate a project on a slim budget that might not otherwise have room for validation, generative AI can help gather signals quickly and on a budget. These tools can open up the world of market research for brands for whom it was previously out of reach.
This is another key use: Feeding human intelligence into generative AI models to gain scale and speed. Or, in some cases, to avoid withholding potentially lifesaving treatment. Abeezer Tapia, former head of commercial development at Alphabet’s healthtech research organization Verily, says that this kind of data can come in particular use in healthcare clinical trial scenarios by employing “digital twins.”
One area they are highly valuable ethically and medically: When trying to eliminate the need for people who enroll in a clinical trial, hoping to receive an experimental treatment for a disease like cancer, but who actually only get a placebo because they were assigned to the trial control group, he told the panel on a recent Ipsos webinar. By using a virtual replica of the patient’s biology aggregated from similar patient data, researchers are able to build predictive patient models, extrapolate results, and eliminate that need.
“Let's take ALS, a rare disease. You can now take a look at the last 10 years of ALS clinical trial data and load up your model with that information. And what some companies are doing out there today, which is pretty fascinating, is when you enroll an ALS clinical trial, you'll have someone who gets the therapy in the variable arm [receiving the treatment]. And then you can actually take that person, run them through the model to emulate what that person would have been like in the control arm [receiving a placebo].”
