Arshak Navruzyan is the founder of Launchpad.AI, a provider of artificial intelligence and machine learning solutions to enterprises. He founded Fellowship.AI an applied machine learning fellowship program, and is a cofounder of Platform.AI. He has delivered AI solutions for some of the largest enterprises in the world and multi-billion dollar quantitative hedge funds. Previously, Arshak served as the Chief Technology Officer at Sentient Technologies. He has also been in technology leadership roles at Argyle Data, Alpine and Endeca (Oracle). He attended the University of California, Los Angeles, where he studied Russian literature and art history.
Q. Your family is originally from Armenia. What brought your family to the United States, and where did you settle and grow up?
A. I knew there was no real future for me in Soviet Armenia. I was rather precocious for my age, and it was my initiative that brought us to the U.S. That was 1988 just as the Soviet Union fell apart.
Q. What were your first impressions of Los Angeles?
A. Disappointment, I sort of expected the U.S. to be more futuristic, high tech and livable.
Yerevan [the family’s hometown in Armenia] is a proper city with architecture, stone buildings, parks… LA is a hodgepodge of random buildings, malls, designed for cars and only has Griffith Park. But shortly afterwards, I discovered the public libraries and would spend most waking hours there. I was starved for books. In the Soviet Union, everything was contraband. People passed around hand-copied versions of books. Here you could read whatever you wanted. I basically learned English by spending all my time at the L.A. and Glendale public libraries.
Q. What were you reading? Was there anything systematic about how you discovered books?
A. I was an adolescent so top of my list were books you absolutely couldn’t read in the U.S.S.R. that were considered “degenerate” for political and moral reasons like Camus, Sartre, Beckett, Joyce, Jean Genet. I don’t read any of these authors now, it’s not my taste. I prefer Chekhov, Turgenev, Gogol, but you know how it is with forbidden fruit.
Q. I note some existentialists among your early reading. How much of the thrill of reading them was that they were forbidden back in the U.S.S.R. versus how they might have resonated with the fact that you were in your adolescence, meaning, a lot of us seem to go through an existentialist phase when we are young. Why have you gravitated back towards Chekhov, Turgenev, Gogol?
And I know that you developed a love for visual arts. What did that stem from? What art and films were you into in your younger years?
A. I was thinking about this recently, if you want to do something sophisticated with your audience you need to have a good shared frame of reference (to get jokes, speak shorthand,). My frame of reference matches these authors.
Scorsese and Coppola talk about it in this interview. They said their frame of reference was classic Hollywood, today’s generation has a frame of reference based on TV sitcoms (which makes it challenging for them to make good movies that the audience will get)
Even in the visual arts I gravitated towards things that I am somewhat indifferent to now. As an adolescent I think you look for things that are very different, so at that time I was into Peter Greenaway, Sergei Paradjanov, [Andrei] Tarkovsky, in other words, visually striking, postmodern narratives–the cinematic equivalent of the purple and pink mohawk.
My taste has evolved to filmmakers like Eric Rohmer, Yasujirō Ozu, Louis Malle and to a certain extent Fellini and Antonioni that tell ordinary human stories, but with the subtlety and depth of understanding of Chekhov.
Q. Yes! Like how indebted to Japanese samurai films and WWII movies George Lucas was! And to Flash Gordon too…!
A. So I think as with most things, my love for the visual arts started with a really good teacher. Someone that made paintings and sculptures seem really significant like the key to unlock a past or even present civilization.
If you’ve seen Simon Schama’s “Power of Art” series, I think Schama does this very effectively. You want to understand the French Revolution, all you need is to hear Schama talk about Jacques Louis David.
I started watching this BORIS play yesterday. It’s amazing, unbelievably original. I would love for everyone to see it to really understand Putin. But it won’t work because outside of Russia. No one reads Pushkin, so the jokes will be missed, and they don’t know Russian history, so the context doesn’t exist. This is something we can say about [AI] LLMs [large language models], they don’t have enough context.
Q. Who was the teacher who influenced you, and was there a sudden event, like a specific work that struck you?
A. I forget his name now. It was at university, I was taking Greek / Roman art and architecture as an elective. I didn’t have high expectations of the class. He basically had this idea that Roman art was propaganda, statues of Augustus, and this completely changed my viewpoint.
Q. I was surprised to learn that you were not a computer science major back in school–but that you studied art history and literature? What were you originally planning to do with a college degree?
My plan was to become a museum curator and for a while I worked and volunteered at LACMA [Los Angeles County Museum of Art]. I quickly figured out the career opportunities were very limited, and so I pivoted to tech, which I also loved.
Q. Why did you want to become a museum curator?
I wanted to be a curator because I thought I have good taste and continue to privately curate, books, music, films for a circle of friends.They say if you love something don’t do it professionally.
Q. What kind of things are you curating for your friends?
A. Mostly Russian and French cultural artifacts that aren’t available in translation. (YouTube [Closed Captions] CC auto translate is only like 30 to 40 percent accurate.) Also, having some background information helps. I’ll sort of set the scene why I’m excited about a particular director and writer and watch it with them so they get it. It’s rewarding for me because trying to explain it to others deepens your own appreciation and understanding.
Last few examples have been Vakhtangov Theatre’s production of Eugene Onegin, the moral tales of [French film director / writer] Eric Rohmer, Harold Pinter’s Victoria Station.
Unless you’re my friend, chances are you will live your entire life not realizing how great these works are.
Q. How strong were you on the science and quantitative side? How challenging was the pivot? Did you bother to change your majors or was it mostly changing your coursework? What were your interests on the computer science (CS) side in and coming out of school?
A. I never took CS classes. Didn’t find them interesting. I taught myself, playing around with early versions of the dialup bulletin board systems (BBSes) and (“worldwide web”) www before browsers were invented. I got a job at the UCLA Office of Academic Computing doing dialup internet support for students and faculty. That was the turning point. From there I went to EarthLink Network.
Q. How did you get from the early BBSes to where you are today in AI?
A. I went from BBSing to where I am today in both a linear and nonlinear way. BBS had forums, not unlike what Slack has today. They had app share features, not unlike that of the Apple App Store. What’s amazing is how far things have progressed in my lifetime.
Q. What common thread, if any, do you see between your interest in art and what you do in AI? Where is the line between your interest in curating art and creating art?
A. I was in Business Intelligence for a while working with Endeca, Qlik, Tableau and realized that manually sifting through data will not scale, and you need something intelligent–that isn’t just a business analyst drawing conclusions from data. I don’t have much interest [nor] probably the talent and courage for creating. I’m happy to be an advocate for others.
Q. Tell us about Launchpad.AI. What is it, and how and why did you start it? How is Fellowship.AI part of it?
A. Launchpad is the commercial arm of the Fellowship. Fellowship started in 2015 with the agenda of expanding the community of qualified [AI] practitioners through the hands-on experience of real world use cases, Launchpad was created in 2017 to commercialize some of the research work we had done. The initial focus was on trading through an exclusive partnership with a hedge fund. We pivoted to enterprise when we added Launchpad.
Today Launchpad is largely an AI consultancy focused on delivering resources and solutions to large enterprises. We continue to stay deeply engaged in research through the Fellowship, which is a further differentiator. Our people stay more cutting edge because of the research.
Q. How hard a “sell” has AI been when enterprises generally are still on the ground floor of understanding what is capable with AI?
A. We don’t hard sell, because we tend to work with industry leading organizations that clearly recognize the value of AI.
Q. Please talk about some of the more interesting projects / applications of AI you have executed?
Production AI is still fairly complex, and often there is a mismatch between what an organization wishes to do versus what they can realistically afford to do. We’ve been around for 7+ years now. There is no lack of interesting use cases: systemic trading strategies, EV battery lifespan prediction, optimizing medical claims, asset management signals from earnings calls, discovering the next blockbuster athletic shoes from consumer preferences…
Between commercial and Fellowship research, it might be easier to list what we haven’t worked on (military and sin industries).
Q. I know you are very familiar with OpenAI and its projects, but what was your reaction to ChatGPT when it became more generally available?
This is what keeps me so excited about Launchpad, I’m always learning. ChatGPT is a pivotal moment in that it takes AI out of the hands of researchers and engineers and invites the individual to experience its power directly–similar to the PC revolution that democratized compute. OpenAI is not Apple, however, they are throwing off interesting and useful artifacts on their [artificial general intelligence] AGI quest. The name of the prototype already tells you that they’re not a consumer products company.
Enterprises are starting to be concerned on many levels about depending on ChatGPT, about what information leaks into these models during inference. I think enterprises will be inspired to move more quickly because of what ChatGPT instigated, but they’ll probably look to instructionally fine tune open source models like LLaMA.
Q. What does the curator in you think of how generative AI is going to change the creation of art? What will the role of humans be?
A. I go to a lot of contemporary art museums and see [something like an] artist who has quilted a provocative slogan into a blanket. This is commercialization of work that belongs on a freeway overpass not in a modern art museum! It’s not mind expanding, shocking, inspiring, terrifying, the way real art is.
As [Martin] Scorsese argues, movies based on Marvel comics are not cinema. I argue that word art isn’t art. Maybe generative [AI] can get us out of this rut, maybe it will get us even deeper into it. If I knew the answer I would be the next [art dealer] Larry Gagosian–who coincidentally has a succession problem because great taste isn’t easy to pass on.
Q. How possible is artificial general intelligence (AGI)? How far or close are we to achieving something like it?
Depends what we mean by AGI. If the definition is accurate and instantaneous retrieval and summarization of knowledge, I think we are only 5 to 10 years out. If our definition is more broad–that includes scientific discovery, real creativity, real-world decision making, I think it will take much longer, 50 to 100 years, perhaps.
To train AIs with these advanced capabilities we need to give them a lot more access to the real world (let AI run labs and blow things up) and rethink our notions of privacy (let AI listen to your dinner conversations and even difficult arguments with your spouse). Otherwise, AIs trained on the Internet will continue to exhibit the intelligence of the Internet.