What AI Professionals Worry About As AI Clears The Hype Curve

By Allison Proffitt

July 26, 2018 | When it comes to a hype curve, AI is coasting along it. Artificial Intelligence is now a catchall buzzword for many different technologies—machine learning, deep neural networks, natural language processing, image or pattern recognition, voice-capable digital assistants. Having AI capabilities is sometimes a job for marketing as much as scientists.

But that’s not to say that these technologies are hype. Researchers have been developing and using machine learning, natural language processing, deep neural networks and more for years. And many of these technologies are yielding very practical returns on investment already.

This spring, Bio-IT World surveyed 124 professionals from pharma, biotech, academia, and more about the biggest opportunities and concerns associated with AI. Nearly 70% of the survey respondents were from the United States; about 60% fell into the 40-59 age range.

Almost 70% of survey-takers reported using some AI-related technology in their jobs. Machine Learning was the most-used with 87% of respondents reporting use; natural language processing came in second place at 47%.

We asked respondents if they have concerns about AI in their roles. And 20% of respondents—a significant number—reported none at all. But nearly half of our survey takers are concerned about data security and nearly half expressed concerns with the technology’s accuracy. Some checked the box for concerns that the technology would make errors, but many more wrote in their own concerns around bad data models, bad training datasets, poor understanding of how the technology works, and no auditing tools to test it.

There’s no escaping it: the technologies are hyped and rushing into applications can be foolish, expensive, and potentially dangerous.

The biggest hurdle to implementing AI technology in research is, “the selection of AI approaches [appropriate] to solve complicated problems based on an understanding of both these problems and the AI approaches themselves,” wrote one researcher.

And yet, 99% of the survey takers see promise in the technology. Almost unanimously researchers agreed that AI-related technology will assist with drug development within three years, and they saw applications in many different areas: target discovery, biomarker discovery, patient stratification, diagnostics, and more.

With careful clarification of how the technology works and what we can—and should—expect it to do for us, AI can clear the hype curve and deliver.