Artificial intelligence has disrupted a myriad of industries in more ways than one. With the rise of AI subfields like generative AI, machine learning, and AGI, the rapid pace of these modern advancements has overhauled a global set of practices that’s more optimal, efficient, and cost-effective than obsolete models.
The aviation industry is quite late in the game. With safety as its top priority, will it ever break through the barriers of adopting the latest technologies and keeping up with the early movers?
Today, I am joined by
Omri: Thanks for being here, Uri. Can you please tell us about your background?
Uri: I’ve been playing AI and computers since I was a kid. Back then, in the 80s when I was in high school, my dad bought me a computer, and I started writing programs, trying to make the programs play chess. Since then, that’s what I've been doing. I feel blessed that I get to do the work that I enjoy so much until now. For the last 3 decades, I’ve been managing, initiating, and building projects that are about AI and thinking machines. At some point, I stopped and chose to go to academia, and learned about computational neuroscience because I was and still am a big believer that, to build thinking machines, we need to mimic the way the brain works. After my Ph.D., I started working in the algorithmic trading field where I was head of AI in a big company for more than a decade, applying techniques taken from AI into trading. After that period, my partners and I chose to apply the same techniques to more traditional markets like aviation, and that’s when we founded Fetcherr.
Omri: As an AI expert, can you share the differences between deep learning, generative AI, machine learning, and AGI?
Uri: These concepts are quite confusing. Let’s start with artificial intelligence in general. It talks about the ability of the machine to perform tasks that require human intelligence. It relates to the principle of mimicking the human mind. Humans learn from data they learn from their lives, and that’s where machine learning comes into play because you want to perform these tasks based on data points. So ML is a subfield of AI based on samples. Another more concrete field is related to how our brain works. Our brain is made of cells that communicate with each other. These are called neurons. The technique of architecting these solutions using neurons that are connected in the network is called a neural network, and that field has exploded in the last decade. ChatGPT is an example that is based on this.
General Intelligence is the holy grail. That’s trying to build thinking machines that can do everything. Generative AI is also booming. It’s a subfield where the output of AI is generating content. For example, generating videos, as in Deepfake; generating text, as in ChatGPT; generating images, as in Dall-E, or generating demand curves of airlines, as in what we do in Fetcherr. Generative AI is quite successful.
https://youtu.be/zCDtdBIUKu0?embedable=true
Omri: I wanna take a deep dive into how technology has impacted the financial markets, particularly in terms of results, optimization, efficiency, and how the operations side looks. Can you elaborate on that?
Uri: I’ve been working in financial markets and capital markets for more than a decade. Technology has dramatically changed these markets as it has allowed new business models to emerge. Automation is very apparent, and it’s not comparable to how these markets behaved before. Speed has been optimized – in exchanges trade can take nanoseconds. It’s also related to transparency and liquidity, whenever you want to buy some financial instrument, it’s always there. But what’s interesting is the dynamics of these markets. In capital markets, there’s always a buyer and a seller that is waiting to trade with each other. A few decades ago, the spread between buyers and sellers was quite high, and that added a type of friction in the market. Today, thanks to technology, the markets are getting more frictionless, which invites more traders to come in. As traders come in, the markets become smoother. It’s a process that feeds itself.
Omri: You mentioned, “frictionless.” The aviation industry has a lot of friction. How can they do better?
Uri: It relates to the history of these industries. Aviation is always about safety. And when that’s the priority, it has some consequences. When safety’s number 1, make it human. It means humans should be involved in the decision-making. And if that industry is very human decision-oriented, then it can’t deploy software frequently, unlike other markets. And whenever it’s deployed, it’s deployed in patches – very small steps forward. That’s another problem because if you’re putting a patch on a patch on a patch, and these patches aren’t talking to each other, then you have a big problem. If the tech structure of this market is fragmented, then the whole mindset of the solution is problematic.
Now, we’re in an era where technology can make huge leaps in the industry. It makes sense to take advantage of it and make changes. The consequences are not only to make things faster and less frictionless but also to make the whole market less volatile.
Omri: How can AI play an impact in a very traditional industry like aviation?
Uri: Once the AI gets into the tech infrastructure, it makes better decisions. If better decisions are made, it gets more revenue for the airline. It makes the communication process more frictionless because it cuts the middlemen off. The prices get reduced for the end consumers. Being smooth means being less volatile – it reduces the risks.
Also published here.