Melting Psychology and Deep Technology with Amir Konigsberg

Share on linkedin
Share on facebook
Share on twitter
The multifarious tech operative talks about his beginnings, empowering startups with AI, and the blueprint for disruptive innovation.

In Israel, the arguable startup capital of the world, you might be fortunate enough be a startup founder and attract investor funding, or just join a promising venture in any capacity. But rarely do you have an opportunity to do both, or act from both ends of the startup spectrum. Introducing Amir Konigsberg, in a candid interview, he sheds light on unseen corners of the startup game and the recipe he’s conjured for commercializing deep tech innovations and becoming a trusted advisor to many in the process.

In reverse chronological order, Konigsberg most notably co-founded Twiggle, an e-commerce search technology growth startup with more than $35 million funding, with investments from the likes of Alibaba, Yahoo and The Naspers Group, wherein he acted as CEO for the first four and a half years of the company but recently relieved himself to President. Twiggle, originally conceived by Konigsberg and his co-founder Dr. Adi Avidor, marries Natural Language Processing (NLP) with e-commerce to embed human context into internet search. What was once largely limited to strict matching of keywords and conformation to search engine’s architecture became enabled by intuitive free language expression. Like searching for a ‘brown formal jacket suited for mild weather’, which would accurately return relevant results as opposed to a randomized order of jackets. The magic of it stems from a breakthrough in Natural Language Understanding (NLU) conceived and supported by the 25+ patents filed by Konigsberg, Avidor and their diligent Twiggle algorithm team. But equally, from a fundamental understanding of human behavior and psychology, since both psychology and technology contribute to building powerful information retrieval systems.

“All my life, I’ve been interested in the intersection of human psychology and deep technology” explained Konigsberg. “I’m interested in the information systems that understand human thought processes and language, and can improve people’s belief formation, knowledge acquisition, and decision making. More generally, I’m interested in designing and tailoring automated solutions to interface with humans.”

Konigsberg earned a PhD from Princeton University and The Hebrew University of Jerusalem in Rationality and Interactive Decision Theory, combining the fields of Philosophy, Psychology and Game Theory. His doctorate covered how to update one’s beliefs based on changing information, a Bayesian approach to belief updating inspired by Robert Aumann’s seminal Agreeing to Disagree. He completed his dissertation under game theorist and behavioral economist Eyal Winter, but drew inspiration from Israeli Nobel Prize winner Daniel Kahneman, to whom he owes gratitude for guidance in his Princeton admission. But Kahneman’s work didn’t consummate with Artificial Intelligence.

His professional momentum took motion in 2005, working in strategy for Google, first in London and then in Tel Aviv, under Meir Brand, as the second Google employee in the Emerging Markets, and later proceeded to join MySupermarket as the Chief Commercial Officer. “The startup was truly unique at the time. It allowed consumers to shop in aggregate through multicart shopping. It provided multi-product shopping comparison, which is something shoppers need in order to compare the price of a cart in aggregate across different stores. But it is very difficult to do because retailers sell similar products by different manufacturers and brands, which is a challenge for software to identify due to the lack of common criteria by which to compare. This required an algorithmic supplement inspired by the decision making process that humans apply when comparing for related items.”

Following MySupermarket, Konigsberg worked for General Motors after the 2008 financial crisis where he worked on human-machine interfaces for autonomous and semi-autonomous vehicles. “Very often people think the major challenges involved in getting self driving vehicles on the road involve perception – LiDAR, RADAR, etc… – but in fact, the passenger experience and human machine interaction is also a critical and very challenging issue that needs to be dealt with for self-driving cars to become commonplace. For example, a self-driving vehicle may need to hand control back to the human driver because of unforeseen problems with the self-driving system  – e.g., because of lack of visibility due to computer perception problems. How to hand back control to a driver while she is not focused on driving is not straightforward; the driver will be in one of many cognitive states and this will determine how she will respond to alerts or notifications about taking back control as driver. And this means that when designing systems for interacting with drivers, many things have to be taken into consideration such as the timing of the notifications based on the predicted state of the driver and the way in which these notifications will be delivered. And to be able to do this you also need to design systems to detect the state of the driver, and so on.” Konigsberg also led projects on search interfaces within the Autonomous Vehicle group. “I worked on voice-based search mechanisms where the informational retrieval needs to be tailored to the passive state of driving. It’s a massive area, evidenced by the fact that Google and Apple, two companies that until now haven’t been focused on automotive, have an operating system dedicated to automobiles. This is likely due to the fact that very soon, you’ll be spending more time in the car, focusing less on driving and more on entertainment, work, and other activities that can be done when attention is freed up from the need to drive the car. It’s a super fascinating area and I hope to revisit it one day.”

Konigsberg also co-founded Israel Brain Technologies with Dr. Rafi Gidron, Miri Polachek, and Shimon Peres with the goal of creating a global conversation on the commercialization of brain science and brain technologies. And the art of commercializing technologies is a key competency that Konigsberg has mastered.

In 2014 Konigsberg reconnected with Avidor through Gidron. He had worked with Avidor ten years earlier while they were both at Google. “We started Twiggle with the goal of helping machines understand shoppers better, so that people can find what they want to buy.” To accomplish that, they had to build technologies for NLU and knowledge creation. They chose to start by focusing their efforts within a specific domain, rather than compete with the lurking giant Google. “We went for depth rather than breadth, and choose e-commerce as our vertical because it had a dire need for improved search, upon which NLP could satisfy – and it hadn’t been for over 20 years because of the limitations of text matching and keyword search.”

Photo Credit: Gali Kaner

Why is NLP so important now? According to Konigsberg, “One of the holy grails of artificial intelligence is enabling humans to be able to communicate well – naturally, fluently, seamlessly – with machines. For this to happen, machines need to be able to understand people, and to also be intelligent enough to converse with them. When done well, NLU is very powerful. It allows you to converse with a machine, across multiple stages of a conversation while jumping in and out of varying contexts, for which the machine has sufficient background knowledge to communicate well and be useful. Recently, OpenAI released a model that’s able to respond to a text prompt, called GPT2, the followup to GPT. It was given one goal to be able to predict the next word of text, based on a corpus of 8 million web pages, and the results were astounding. It builds familiarity on large training sets of data. It’s really impressive and we’re going to see a lot more synthetic writing because of it.”

But even with a rich background in technologies across a number of industries, still, the fickle commercialization problem persists across the massive pool of entrepreneurial projects across Israel. “I like looking for the business case to which a very powerful technology can be applied, and not vice versa. In Israel, the typical startup builds technology and then finds the monetization aspect. And within Israel’s burgeoning AI ecosystem, commercialization is a skill set in dire demand. However, while exclusively focusing on Israel’s tech, a crunch in talent may stall the ecosystem’s progression, attributed to multinational corporation’s (MNCs) high compensation package bids, destabilizing the playing field.”

“We need to close the incentive gap between MNCs’ exorbitant salaries with the equity and fulfillment provided by competing startups. At end of the day, with aggressive competition for talent, there’s more friction involved in creating an AI-based company, which means these companies will have to radically prioritize their hires. Which may also also benefit the startups by forcing them to radically optimize their staff and be extremely resourceful in how they design their technologies. While you can’t hire as many AI professionals, you can spend more time screening for professionals with the skill sets fit for your problem or hire a few highly capable people with outstanding learning curves. But the challenge is growing.”

While Konigsberg took on the role of President at Twiggle in September 2018, he’s also taken on independent board member roles in AI driven companies to share his wealth of experience. He’s currently on the advisory board of Allegro.ai, a Samsung, Hyundai, and Bosch backed company offering the first true end-to-end AI product life-cycle management solution with a focus on deep learning applied to computer vision. And he recently joined as a Board Member to HourOne.ai, a venture backed company building powerful computer generated video technologies, paving the way for synthetic video creation at scale.

As a rule of thumb to any AI startup, he implores technical entrepreneurs to “find a commercially minded co-founder that can elucidate the vision of the company in a non technical language but also with technological oversight.”