#120 Natural Language Processing with Catherine Havasi – CEO
Catherine Havasi's dream and mission are to give computers the ability to think about the world more like a person and less like a machine. Equally important is putting this innovation into practice in a way that effects real business outcomes and transforms our everyday lives. Catherine has twenty years of experience in directing and doing cutting edge research, operationalizing it, and using it to drive real results. In many ways, Catherine is a product of the MIT Media Lab, which taught her that technology is most valuable when it dreams big, thinks about its users, and doesn't forget to make a demo as soon as possible.
In this episode, Catherine explains how she started in the data world. Growing up, Catherine was always interested in learning more about how people think. Catherine's curiosity impacted her career tremendously. However, Catherine realized that teaching a computer to be smarter is not the right way to figure out how people can be more intelligent. Working in psychology, Catherine learned that young kids can be very adaptable to what they learn. When kids start seeing patterns that are different, they pick them up and start using them very quickly. Patterns also work well with artificial intelligence – Catherine picked up on that similarity right away.
In cognitive science, people ask a lot of direct questions. Catherine was unprepared with how questions are asked in psychology because people in computer science are way more diplomatic. Then, Catherine explains her move to working with language. She has always liked to learn about various languages despite not speaking them very well. AI people tend to focus on areas that they do not excel at. Catherine says that it is a little nuts how fast NLP has been developing in the last few years. Loads of challenges are going on. These challenges make it hard for people to be on top of everything, especially for students, faculty, and entrepreneurs. A lot of research is slipping through the cracks, and possibly things are getting overlooked.
Currently, Catherine is looking at story understanding. Conversational AI is very transactional. If you want to do something fun like putting AI on stage at an improve comedy show, then the AI needs to learn more about relationships and less about answering questions. Catherine is looking at what it would take to have AI work in different places. For instance, they are looking to have humans, and AI works together to solve problems. Later, Catherine speaks about her research around creating healthy habits. It turns out that the motivation behind creating a healthy habit matters a whole lot less than how the habit is named. Concrete language is more indicative of later behavior. Stay tuned as Catherine speaks about her time at Luminoso and her transition to entrepreneurship.
Enjoy the show!
We speak about:
[02:25] About Catherine Havasi
[03:50] How Catherine started in the data world
[05:25] How did your curiosity have an impact throughout your career?
[06:20] What surprises did you find in psychology?
[07:30] How do you describe the culture of cognitive science?
[08:45] How did you move to language processing?
[09:40] How has the language field evolved over time?
[11:55] What are the differences in data for language?
[16:30] What are you working on right now?
[19:10] What else is in your previous research?
[20:30] How does research work in the field of medicine?
[23:00] What the relationship between healthy habits and language?
[28:30] What is the relationship between patient and doctor language?
[29:00] What other things in language help in medical relationships?
[30:25] Tell me about your time at Luminoso
[35:00] How was the transition to becoming an entrepreneur?
[37:30] How has your company evolved?
[39:40] When is your company launching?
Resources:
Catherine’s LinkedIn: https://www.linkedin.com/in/havasi/
Catherine’s Twitter: https://twitter.com/catherinehavasi
Catherine’s Website: https://www.catherinehavasi.com
Dalang Health: https://www.dalang.ai
Quotes:
“People with computers can do more, that either can do separately. How do you build an environment that is not just a phoneline/helpline but people can go to the computers if computers are around and can go to person if that person is around, or people and computer can work together to assist people.”
“Especially when it comes to AI focused design in general, we want to be figuring out not just how to replace people, but how to amplify people to do their job better and to reach more people, and those vulnerable people are adhered to by humans answering hard questions, where AI cannot directly deal with such questions, and human touch is necessary (example current crisis).”
“How much a change you want to make to your life/motivation matters a lot less, rather than how much a change can fit into your daily routine, matters a lot more, or attach to something that is already a habit. For a first week or so its motivation, but then 100% it is how it fits into your life.”
“It’s a little nuts how fast NLP has been developing in the last few years. There is so much going on that it is hard for people to be on top of everything especially for students and even faculty and even for entrepreneurs. There are models out there that are performing really well and they have these extraneous parameters, but they are hard to be onboarded , because although they are doing well, but we don’t know why they are doing so well. Lot of research is slipping through the cracks, and possibly things get overlooked or other times not known what is out there and why is it doing so well.”
We are now on YouTube! Watch the episode here: https://youtu.be/Gl0Sfjq5UQA
And as always, we appreciate your Reviews, Follows, Likes, Shares and Ratings. Thank you so much for listening. Enjoy the show!