top of page
Writer's pictureMatt Strain

Company to Watch: Hume AI

► What’s Cool About Hume AI

Augmented Empathy! Hume is able to accurately detect and categorize a persons emotional state based on voice input. In their demo, you’ll see that they also respond to your input with impressively appropriate empathetic tones.


There are a wide range of uses, but at the highest level, being able to increase the quality of empathy and human connection is exciting.


► What to look for in Hume:

The good: Think about how this can be used for:

● Personal development and self-exploration

● Helping people be more empathetic partners, managers, and teammates, and

● Early detection for conditions that range from stress to diseases like Parkinson's


The bad: In the wrong hands, being able to monitor, judge, score, and manipulate others based on vocal input could lead to some very dark outcomes. Computational Emotional Manipulation will be a thing.


► Interesting Use Cases (and datasets)

What was most striking are some of the datasets they're collecting. Some of the ones listed on their site include:

● Depressed vs Non-depressed Mood: Uses video diaries to identify depressed moods. Imagine the uses for mental health awareness and intervention. Profound implications for early detection and support.

● Parkinson's vs. Non-Parkinson’s: Analyzes voice recordings to predict Parkinson's disease presence. Vital for early diagnosis and management.

● Alert vs Drowsy: Determines driver alertness through video analysis. Preventing accidents and enhancing road safety.

● Toxic vs Not Toxic: Classifies streamer speech as toxic or non-toxic. Helps create safer online environments by mitigating harassment and abuse.

● Attentive vs Distracted: Identifies attentiveness in virtual meetings. Important for improving engagement and productivity. (May spike stress in attendees).

● Self-Confident vs. Self-Doubting: Uses video clips to assess confidence levels. Valuable for personal development and mental health. Yes, potential dark side. 

● Good vs Bad Call: Evaluate customer service call quality. Useful for improving customer satisfaction and service standards.

● Song Genre Prediction: Predicts song genres from the GZTAN dataset. Help for music categorization and recommendation. (May not change the world)

● Best vs Worst Baker: Predicts baking competition outcomes. Entertainment-focused with minimal impact on well-being or societal benefit. (Because you can)


A shout-out to Alan Cowen and the Hume.ai team for pursuing AI for good.


Check out their playground to see these datasets.


7 views0 comments

Comments


bottom of page