Episode 17. Our guest is Glenn MacDonald who was Spotify’s Data Alchemist, building it into an algorithmic powerhouse.
We’re critically evaluating algorithms' effectiveness and why GenAI probably isn’t the best technology for many problems.
Some key insights:
#1. As Spotify's former data alchemist, I expected huge advocacy for hashtag#ML & hashtag#AI as a predictive technology. Instead, we must not play god with algos. They should be assistive tool to get people to where they're headed. Prediction leads to errors.
#2. You must be able to evaluate algorithms. Too often we're deploying fancy tech with no way to know it is performing better than an alternative. hashtag#GenAI has a huge risk of this because the assumption is that it solved everything. But the cost of deploying it is also very high.
"I think the main thing I've learned Is actually not to think about it as prediction, I think the thing that happens to you when you start thinking about things as prediction, and I think this applies to thinking about LLM, LLM outputs as predicting text. It also applies to A& R and music as like predicting hit artists. The moment you start thinking about it as prediction, you've sort of internalized sort of ugly idea that the future is kind of determined and you're just attempting to guess what it's going to be and thus profit by anticipation. And I think it's a lot more productive to not think about the future as something you're predicting, but it's something you're making. "
"I think a lot of the time we evaluate new tech against really Poor baselines, like against randomness or against the most popular things, or like you said, against just like our intuitive guesses. And in those contexts, sometimes the fancy tools seem like, Oh, they're clearly better. But then when you compare them against, Oh, what if we just did some math and you realize. Oh, the math's even better. It's a lot simpler. "
The episode is hosted by:
Arpy Dragffy Guerrero (Founder & Head of product strategy, PH1 Research) https://www.linkedin.com/in/adragffy/
Brittany Hobbs (VP Insights, Huge) https://www.linkedin.com/in/brittanyhobbs/
Glenn McDonald is a music evangelist, algorithm designer, software engineer and technology strategist. He created the music-exploration website Every Noise at Once, and for 12 years was the Data Alchemist at the Echo Nest and Spotify. He has written about music online since before "blog" was a word, and his first offline book, You Have Not Yet Heard Your Favourite Song: How Streaming Changes Music, is available now from Canbury Press.
00:24 Meet Glenn MacDonald: Spotify's Data Alchemist
01:50 The Evolution of Music Discovery
08:39 The Role of AI in Music and Beyond
13:29 Challenges and Future of AI in Music
29:14 Navigating AI in the Workplace
31:25 Designing User-Friendly Algorithms
34:59 Challenges with Algorithmic Recommendations
39:42 Evaluating AI and User Testing
47:41 The Future of Music and AI
Thank you for listening to the Design of AI podcast. We interview leaders and practitioners at the forefront of AI. If you like this episode please remember to leave a rating and to follow us on your favorite podcast app.
Take part in the conversations about AI https://www.linkedin.com/company/designofai/
And subscribe to our newsletter for additional resources https://designofai.substack.com/
Spotify’s former data alchemist: Evaluating when & how to use GenAI