Spence Green of Lilt On The Future Of Artificial Intelligence
Machine translation: People have been working on machine translation for 70 years and so much progress has been made just in the past few years– even more than the last 15 that I’ve been working on it. These systems have really advanced and are now being utilized. There are a lot of cases now where you can use machine translation systems without any human being involved whatsoever.
As part of our series about the future of Artificial Intelligence, I had the pleasure of interviewing Spence Green.
Spence Green is Lilt’s Co-Founder and CEO. Prior to founding Lilt in 2015, Spence served as a fellow at XSeed Capital, worked in software and research at Google, where he worked on Google Translate and developed a shallow syntactic language model for improving the English to Arabic translation. He was also a technical lead, project manager and software engineer at Northrop Grumman, where he worked on various large projects including a national air defense system and avionics packages for naval aircraft.
Spence graduated with highest distinction from the University of Virginia with a bachelor’s degree in computer engineering. He also earned both a master’s degree with a distinction in research and a PhD in computer science from Stanford University.
His research area is the intersection of natural language processing and human-computer interaction, and he has published papers on statistical machine translation, statistical language parsing, and mixed-initiative systems and given talks on translator productivity.
Thank you so much for joining us in this interview series! Can you share with us the ‘backstory” of how you decided to pursue this career path in AI?
I became interested in language about 15 years ago, around the time that Google Translate was released. I had been programming since I was a kid, but I had gotten bored with building deterministic systems, which are systems that just execute a series of procedures. I became really excited about machine learning systems, like Google Translate, which are probabilistic. I left my job and went to grad school to start working on machine learning.
What lessons can others learn from your story?
Find something that you’re interested in with a high impact that doesn’t feel like work. There’s an opportunity to convert your passion into a business or public need.
Can you tell our readers about the most interesting projects you are working on now?
Our team has built a neural automatic review and correction system that learns to predict corrections to a translation. This project won best paper from the North American Chapter of the Association for Computational Linguistics, a top conference in our field.
None of us are able to achieve success without some help along the way. Is there a particular person who you are grateful towards who helped get you to where you are? Can you share a story about that?
My parents. Thankfully, I was born into a family that valued education. My parents made sure that paying for a good education was a number one priority at the expense of things they could have done for themselves, like going on vacation.
What are the 5 things that most excite you about the AI industry? Why?
- Machine translation: People have been working on machine translation for 70 years and so much progress has been made just in the past few years– even more than the last 15 that I’ve been working on it. These systems have really advanced and are now being utilized. There are a lot of cases now where you can use machine translation systems without any human being involved whatsoever.
- Large language models: Everyone is really excited about these. They can be used for many different applications that didn’t work previously, such as text generation. With large language models, you can now fluently generate text with a given prompt.
- Foundation models: With foundation models, you can train one large model and then specialize it to a specific task. You don’t have to train different architectures and different models for each task. You can start with a base model and then specialize it to a variety of different tasks.
- Multilingual models: Historically if we wanted to translate from one non-English language to another, we would pivot through two cascading systems. For example, if we wanted to translate from Korean > Arabic, we’d pivot through two cascading systems: Korean > English >> English > Arabic. Today, we’ve seen that LLMs can be trained on mixed, multilingual text, and we can also train MT systems on concatenated parallel text. Multilingual models achieve state-of-the-art deep neural networks, especially for low-resource language pairs. There are vast opportunities for applications of multilingual models.
- Opportunity for impact: Lack of access to information is a limiting factor for many in their pursuit of a better life. AI has the potential to fundamentally change the volume of information available to all, and provide access to a better life for millions.
What are the 5 things that concern you about the AI industry? Why?
- The AI industry was overhyped for a long time. I think that claims of sentient beings and level five self-driving cars did a disservice to where the technology was and resulted in a major distraction from the scientists actually doing the work.
- I think some of the hype around AI has turned into disappointment because the more grandiose claims have not come to pass.
- Hopefully, more attention will be paid to where machine learning systems are having a real impact, such as machine translation which is transforming how we communicate. It’s not as exciting as killer drones, but it has a higher impact on humanity.
- The explainability of machine learning models. People ask how these large neural networks make predictions internally, which is pretty complicated to explain. I think a lot of people are working towards being able to better explain how neural network systems work, whereas, in the old linear models, they would just tell you what they thought and were more readable.
- I think that having systems that we don’t totally understand can be worrisome, but it doesn’t necessarily keep me up at night.
As you know, there is an ongoing debate between prominent scientists, (personified as a debate between Elon Musk and Mark Zuckerberg,) about whether advanced AI has the future potential to pose a danger to humanity. What is your position about this?
Anything people build has the danger of posing a threat to humanity. That sensationalized debate is disconnected from what scientists are actually doing and thinking. I mean, you can use a nuclear bomb to blow up a city or you can nuclear fission to generate power. Anything that people can build can pose a potential threat, but I don’t think that the possibility of killer robots is something we should be worried about right now.
What can be done to prevent such concerns from materializing? And what can be done to assure the public that there is nothing to be concerned about?
The public tends to worry about systems they don’t fully understand, including advanced AI systems. The more we can explain these systems, concepts and functionality in an uncomplicated way, the better.
How have you used your success to bring goodness to the world? Can you share a story?
It’s not my success, but the Lilt team’s success. The work that we’re doing to enable people to communicate is extraordinary. For example, we work with a public school on the East coast to help parents read what their students are having for lunch every day. While it would have been cost-prohibitive to have people do that work, having a machine learning system that facilitates communication is making the world better for those families.
As you know, there are not that many women in your industry. Can you advise what is needed to engage more women into the AI industry?
While this is not my own observation, I do believe that it’s an education pipeline problem that happens early on in the educational journey. We have the same number of graduates in computer science amongst women in the United States today as we did in the early 1980s. Programs and initiatives that incentivize increased interest and engagement in computer science for girls and women at an early age — both within and beyond their academic environments — are critical for building this early pipeline, and ultimately for engaging more women as professionals in the AI industry.
What is your favorite “Life Lesson Quote”? Can you share a story of how that had relevance to your own life?
“The world belongs to the discontented.” — Oscar Wilde. I think that if you have the constant mindset that the world could be better than it is, then that becomes an inspiration to action to change things for the better.
You are a person of great influence. If you could start a movement that would bring the most amount of good to the most amount of people, what would that be? You never know what your idea can trigger. :-)
I humbly disagree that I am a person of great influence. Lilt is a movement to make information accessible to everyone, and that was the movement I chose to start.
How can our readers further follow your work online?
This was very inspiring. Thank you so much for joining us!