Thread by @tleam

#PTL19 #Intro19 - First afternoon session today: “Artificial Intelligence & Machine Learning: Empathetic Use Cases” with Michelle Sipics, Steve Zemanick, @brianorrell, and @skrinak

#PTL19 #Intro19 - don’t do AI to just do AI. Use it to augment a human centric process and help undergo a digital transformation.

#PTL19 #Intro19 - a proposed process for successful AI adoption.

Up next is @brianorrell to talk about some practical applications of AI. First comes from the medical world and how ML helped identify correlations between a string of infection issues in a hospital and potential cause.

#PTL19 #Intro19 - AI/ML application helps not only to produce predictive results but also prescriptive. In this use-case they were able to predict instances of infection and then provide individualized treatment to patients.

#PTL19 #Intro19 - Important note on ease of access to data: if you want AI/ML to provide timely assessments you can't rely on manual collection. Real time pipelines are critical to produce actionable results.

... unless you move in a really slow industry. But who does nowadays?

#PTL19 #Intro19 - @skrinak now talking about how to make some of these AI/ML use cases a reality with AWS technologies.

#PTL19 #Intro19 - So, what's the difference between AI and ML?

"If it's in PowerPoint it's AI, if it's in Python it's Machine Learning" - @skrinak

Whoops! Mistyped and copy pasted a bit too much. If you’re interested in the “Artificial Intelligence and Machine Learning” talk at #PTW19 #Intro19 go ahead and scroll through this tweet chain!

Thanks @kadud for catching it!

#PTW19 #Intro19 - "What are the compliments/substitutes to cheap predictions? What's the strongest compliment? Human decision making and judgement is paramount. We don't want decision makers to provide input without good data."

#PTW19 #Intro19 - How can we overcome some organizational challenges to AI/ML adoption?

AI/ML are no different than any organizational shift: don't forget about change management.

AI/ML provides features and does tasks but does not fully replace jobs, it augments them.

#PTW19 #Intro19 - Data collection is hard... probably one of the most painful processes for a data scientist.