At the time of the conversation, Melisa Russak was a lead research scientist at Writer. Her story of independently rediscovering fundamental machine learning algorithms aired on Chain of Thought as a crossover from Writer’s Humans of AI podcast, a look at curiosity-driven research and accidental discovery.
Latest episode with Melisa Russak · EP 41
The Accidental Algorithm | Humans of AI Crossover with Writer's Melisa Russak
Melisa Russak, Writer ·
Episodes
What the conversations cover
These notes summarize the episodes featuring Melisa Russak, including other speakers in panel conversations. Follow a source link to read the discussion in context.
EP 41 The Accidental Algorithm | Humans of AI Crossover with Writer's Melisa Russak
- Melisa Russak built what turned out to be K-means clustering without knowing the algorithm existed. As a high-school math teacher in China, she wrote a classification system to recognize her students’ handwritten Chinese characters — only later discovering, in her words, that “people are working already more than one hundred years on this problem.” Episode takeaways
- Russak says she chose to study Chinese alongside mathematics precisely because it was “maximally different” — “a completely different system” with no alphabet that pushed her “completely out of your comfort zone.” She wanted learning she could see compound day by day, unlike pure math where “you can spend ten hours on thinking about a math problem and have no results.” Episode takeaways
- Russak rediscovered embeddings the same accidental way. Trying to build software that tracked her internet activity to answer “who am I?,” she had to represent both text and images so a computer could compare them. Her team “spent a lot of time trying to design those features ourselves” — before, as she puts it years later, “it’s all about embeddings… you just need to learn good embeddings.” Episode takeaways
In their words
When I joined, we had two people or three people in NLP team. Right now we have several teams doing NLP. And I remember my first task was training a model for grammar. It was not obvious what an LM can actually do the job well.
Melisa Russak The Accidental Algorithm | Humans of AI Crossover with Writer's Melisa Russak · 14:27 Transcript excerpts; ellipses mark omitted words. Some automatically selected quotes have filler words removed.