Midam Kim, ML Engineer at ServiceNow | BotCast #14

Meet Midam Kim, ML Engineer at ServiceNow, who joined the BotCast #14 to answer our usual questions on building with Voice AI.

Who are you?

My name is Midam Kim. I'm an ML engineer at ServiceNow. I'm a lifelong lover and learner of speech and speech AI.

How did you end up using voice AI?

First of all, I fell in love with speech when I was nine or ten years old, when I was first learning English. I learned how to distinguish R and L while listening to a pop song, and that's how I fell in love with speech. And I wanted to do speech AI since I learned about AI. ServiceNow is making voice agents for everyone, so that's how I ended up working here.

A moment of joy and pain using voice tech.

Joy and pain of building voice technology: every day, every time, just like parenting. It's a great joy because I love this kid so much. And it's a great pain because it doesn't know how to speak like us yet. I am a linguist, so I try to teach everything I know to this kid so that it can really help us.

What lesson would you share with other builders?

It's voice AI technology, so you would think that you need to learn a lot about technology and engineering. And that is very true. But at the same time, it's speech AI. So you need to know a lot about human language, how we communicate with each other, and what is the best way to help each other during communication using speech. And it's called linguistics. So learn linguistics, or hire linguists, and that will solve an incredible amount of problems that you didn't even know existed. So learn linguistics.

Where do you see the industry in 12 months and in 5 years?

In 12 months, I think and I hope we are done, or very close to done, solving alphanumeric problems. And I think we will have a clearer idea about how to use speech-to-speech models for a natural conversation while doing tool calls, so that voice AI can help us solve problems better in a natural conversation. That's my prediction for the next 12 months of voice AI.

In five years, I hope that voice AI will be just like human communication, so that it learns how we speak and who we are in real time while we are talking to it. The audience of this YouTube video will be learning how I speak, who I am, and what kind of words I will say at the next moment. That's how human beings communicate, and I think voice AI will learn that in five years. In Korean (I'm from South Korea) it is called nunchi. When you have good nunchi, you become a good partner in communication. I hope voice AI will be there in five years.