► 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.
Comments