Artificial Intelligence (AI) technologies develop at a breakneck pace. While Amelia stands at the cognitive vanguard, IPsoft isn’t resting on her laurels. To keep Amelia on the industry’s cutting edge (if not a few steps beyond it), we tap the expertise and imagination of our Research and Development team, including Senior Software Engineer Fiona Hou.
Hou is currently helping to drive the R&D team’s Natural Language Understanding (NLU) efforts, which leverage Machine and Deep Learning technologies to discern user intent and capture detailed information through conversation.
For the latest edition of our series, “The Real Women Behind Amelia,” we sat down with Hou to learn more about her work and experiences with IPsoft. In this wide-ranging conversation, we speak with Hou about the potential of AI, how the R&D team decides which features to work on and how she independently pursued a Machine Learning education.
How does the R&D team decide which features to work on?
Some customers want to improve or change one feature or another. We also keep up on what the rest of the industry is doing. New AI research papers publish almost daily detailing new solutions and technologies. For example, some paper might describe how some team was able to improve NLU accuracy by one or two percent – that may not sound like a lot, but because some accuracies are already around 90 percent, increasing it by two percent means the error rate would drop by one fifth.
Once I got to IPsoft, I had the opportunity to work on projects that were quite out of bounds.
That’s a significant difference and we want to keep Amelia at the forefront. We also receive regular guidance from Dr. Christopher Manning, a NLP (Natural Language Processing) professor from Stanford, who helps keep us up-to-date on the latest AI research. (Click here to learn more about Dr. Manning’s research from a fascinating live conversation with IPsoft CEO Chetan Dube from year’s DWS.)
How do you explain to friends and family who don’t have a tech background about what you do?
I ask them, “Do you use Siri, do you use Alexa? That’s what we’re doing.” I’ll then explain that the difference is those products are working in an open domain – they can talk about everything. Ours is more of closed domain. We’re working within specific industries, so Amelia has to be more specific; she goes deeper and is better suited for enterprise solutions.
How would you differentiate the work of R&D versus the Cognitive teams?
That’s a good question. I feel like we are building the core, and the cognitive team is building on top of us. So I would say that we build common features that can be reused, while the Cognitive teams build functions specifically for each client.
What are some things you are excited to see Amelia do in the future?
I mean, there’s big potential right? She’s always getting closer to human. If you look at the top-ranking AI, they’re all steadily improving. Every day different institutes publish papers about how their technologies are getting closer and closer to the abilities of human beings. I would say that in the near future, AI’s going to provide a lot of significant assistance to human beings by executing simple abilities much better than we can.
What lead you to your current role at IPsoft?
I was working as a developer – mainly in Java. I was in the banking and refinancing sectors. I got increasingly interested in AI and enrolled in some online courses to learn more. I started with Andrew Ng’s Deep Learning courses through Coursera. I was particularly impressed by the potential for Machine Learning. From there, I continued pursuing my education on my own through more online classes.
So you’re self-taught in the AI space?
In NLP, yes. I’m relatively new to this, but now I have the golden opportunity to practice it.
What do you like best about working here?
I worked for a few different AI companies before coming here, and they were developing things like AI-guided search that were somewhat interesting to me. But once I got to IPsoft, I had the opportunity to work on projects that were quite out of bounds. We’re really competing with all the major players in the industry who might be more well-known. I think of it as using the best of the technology. We have a lot of dedicated and smart people here. I’m glad to be part of the team.
How do you think the industry can get more women in STEM fields in general, and AI in particular?
I would tell young women that so much of this industry is new, so anyone who really puts in the effort to keep on top of the latest research can succeed. It’s very accessible. There are so many online classes and resources — just take some time to learn it. If you want to go to AI, it doesn’t matter if you are a girl or you are a boy or you’re from China or you’re from America. It’s all about if you are you willing to put in the effort to learn. I’d emphasize how STEM opens many opportunities, but I would also let them know that this field is fun and interesting.
Five Fast Fun Facts
What is your favorite depiction of AI from science fiction?
I like Iain M. Banks’ “Culture” series. Stanislaw Lem’s Fiasco. And Kubrick’s 2001: A Space Odyssey.
Name a technology you’ve recently come in contact with that made you think, “Wow, I’m really living in the future.”
There are many breakthroughs in different areas. The one that pops out of my mind right now is AlphaGo Zero. If AI can learn by itself, there’s really no limit.
Name a notable real-life scientist/engineer/innovator whose contributions inspired you.
Andrew Ng, definitely.
A thing your co-workers would be surprised to learn about you.
I went to a special detox diet camp where I didn’t eat anything for five days. It’s all doctor-monitored. You have to follow a special diet for two weeks beforehand and a special diet afterwards. It wasn’t painful at all.
Your last meal would be?
Anything as long as it’s super spicy.