We’ve all been there: stuck in some supermarket checkout line while an electronic eye repeatedly fails to scan an indistinct barcode. Now imagine a world without barcodes, where cash registers instantly recognize everything you’ve bought – including those individual pieces of fruit and veggies – without human involvement. A flawless self-checkout.
No need to imagine. Israeli start-up Edgify has developed technology that uses intelligent cameras embedded on cash registers and weighing scales to identify items without barcodes. The British-Israeli company, which focuses on distributed machine learning, has developed a platform for training Artificial Intelligence models on end devices that eliminates the need to send data to the cloud.
Following a successful pilot in recent months, Edgify’s innovative technology is expected to operate in most of Israel’s largest supermarket chain Shufersal by the end of the year. According to Edgify’s website at the time of writing, 32 tills across the world had already recognized hundreds of thousands of items, at an accuracy rate of above 95%. A new customer experience.
Stephen Hawking once predicted that Artificial Intelligence will transform every aspect of our lives, and shopping is no exception. The global market for AI in retail should reach $24 billion by 2027, from $3b. in 2020, according to Marketstudyreport.com. Other potentially lucrative fields include the automotive market, agriculture and healthcare.
The world, it appears, has passed the tipping point of AI, with once sci-fi technologies becoming increasingly commonplace, and previously unimaginable applications. The term AI has already become synonymous with a better user (= consumer) experience.
“Although our technology is independent and not unique to a particular sector, we decided to start in the retail industry, where we have already proved feasibility through cooperation with several leading cash register manufacturers and retail food chains worldwide,” said Ofri Ben-Porat, Edgify’s founding partner and CEO, on launching the first-of-its-kind collaboration with Israel’s largest food retailer.
Machine Learning and AI depend on data generated on end devices. The supermarket systems use Computer Vision, an interdisciplinary scientific field that examines how computers can gain a high-level understanding from digital images. (Yes – the cameras can “open the bags” to see what’s inside them.)
Ben-Porat endeavored to explain Machine Learning in laymen’s terms. “Say you’re walking with your newborn child and see a Chihuahua. ‘Look,’ you say, ‘a dog.’ Then you keep walking and see a greyhound, and say, ‘look – another dog.’ The child concludes that they are all dogs, with differences. If we only showed the child greyhounds, he wouldn’t call a Chihuahua a dog. The training of AI is very much affected by humans – we let the real world train. The machines actually do the work without needing the cloud – training in a distributed, collaborative way, like a child does.”
The retail world is changing, and corona has been a catalyst in the process, with many stores morphing into supermarkets of the future. This means accommodating new grocery shopping trends from panic-buying to delivery and pickup options, while coping with disrupted supply chains.
Nadav Israel, Edgify’s founding partner and CTO, notes that Israelis willingly adopt new ideas. “When we started to talk with Shufersal, I was surprised how much they wanted to look to the future. Their contact guys understood the technology – Israelis understand what they can gain from machine learning and are always looking for ways to maximize performance.”
Edgify emerged from technology developed by Ben-Porat and Israel’s previous startup Pixoneye, which identified consumer behavior on mobile devices. In 2018, the management team and investors recognized that the technology has much greater potential than the initial company’s idea, and Edgify was established the following year with $9 million in seed investment from prestigious investors such as Sony Semiconductor, Mangrove and Octopus Ventures. The company, with offices in Tel Aviv and London, is now preparing for another round of fundraising, with its technology already installed on market leaders such as DataLogic, Bizerba, NCR and StrongPoint in global retail chains.
“The potential of the solution we bring to the world of Artificial Intelligence training and the retail industry is tremendous, thanks to the huge amount of information that food chains generate today, which is transferred inefficiently and wastes valuable processing power in uploading data to the cloud, its deciphering and analysis,” says Israel. “Data-driven solutions are hard to achieve. For example, if you’re seeking a medical solution based on data about genes, in order to achieve accuracy you need a lot of data, and a lot of types of data. Hospitals cannot send this data to the cloud due to privacy issues. The only solution is instead of taking out the data, you take parameters that describe the data. Each hospital creates a model based on its data – but you cannot depend on data generated inside one hospital only.
“Our magic is to condense all the data together into one global model. We know how to process the data locally, then take the model from each one [hospital, supermarket, end-device] and interactively build the combined model with increasing accuracy.”
This approach prevents an information bottleneck, says Israel.
“Privacy is less of an issue than bandwidth. Think of a large retailer like Walmart, which needs to manage all its traffic. Taking all that data out of an individual supermarket would encounter bandwidth limitations. Instead, you can process the data locally and combine all the knowledge into one model.”
The cloud is still the data-transfer solution for most problems, he notes. “Some healthcare companies use the cloud, but practically data starts to have limitations, especially when we’re talking about private data. The hardware’s getting stronger all the time – just look at your smartphone. People are realizing that hardware doesn’t cost a lot, and you can do the same processes on-site.”
Ben-Porat said that “the future lies in eradicating the cloud component of AI training. The models continue to learn and advance with the data that continues to be collected – the developed algorithm knows how to differentiate between types of apples and tomatoes without the need of a barcode.”
Israel foresees a bright future in the field of medical crossovers. “There are so many advantages to this approach, such as doctors giving better advice, personalized medicine, more efficient emergency rooms and imaging systems with real-time diagnostics,” he said.
In the not-too-distant future, CT scanning machines will offer comprehensive diagnoses without a doctor seeing the images, thanks to Artificial Intelligence. And that’s just one of the myriad ways AI will change our lives. Soon we will see driverless cars, voice automation in homes, face recognition instead of keys, home-delivery by drones, AI-based virtual assistants placing human-like calls while understanding the nuance and context of the conversation, and even robot surgeons. According to research by scientists at the University of Oxford, Artificial Intelligence will be better than humans at translating languages by 2024, writing school essays by 2026, selling goods by 2031, writing a bestselling book by 2049, and conducting surgeries by 2053.
Israel is at the forefront of the worldwide AI revolution, far outpunching its weight. Every week, it seems, another Israeli start-up launches an innovative project – and they cover the gamut of potential markets.
Following are but a few examples:
- Optibus is developing an AI-enabled platform for planning and running mass transportation
- Deep Instinct applies deep learning to cybersecurity
- OrCam harnesses the power of artificial vision to compensate for lost visual abilities
- Percepto offers a fully autonomous multi-mission drone
- Syte has a product discovery platform designed to transform the way people search and shop online
- AnyClip provides media companies with tools to analyze and catalog their video libraries
- Vdoo provides a platform for automating all software security tasks
- Sight Diagnostics aims to bring affordable and accurate blood diagnostics to the point-of-care
- BrainQ is developing AI-powered technologies to treat neuro-disorders
- EyeSight has gesture recognition technologies for touch-free interactions with digital devices using hand gestures
- Buildots transforms construction sites into a data-driven digital environment.
Why are Israelis coming up with so many new tech ideas?
“In my opinion, it has nothing to do with the army,” says Ben-Porat, who himself served as a paratroop officer. Neither does he resort to overused clichés such as ‘entrepreneurship’ or ‘innovation.’
“I think it has much deeper roots – it’s about the ability to try and fail at a small scale,” says American-educated Ben-Porat, who lives in London. “Nothing is massive in Israel, and they try again until they get it right. The notion of fear of mistakes is lot less worrying to Israelis than to our European or American clients.
“For example, our systems teach themselves before moving from learning mode to active mode. At Shufersal the approach was, “Let’s put it in and see how it works. We’ll figure it out as we go,” while in Europe they want everything working perfectly from Day One. There’s a higher level of confidence in Israel.”
“There’s a holy triangle of high-quality education in a geographically concentrated area producing talent. The fact that we can’t pop over to neighboring countries like in Europe means the circles are quite tight, and able to test in the local market. This forces Israelis to rely on technology and software. Putting all this talent together produces sparks. Tech start-ups are easy to export. This combination creates the platform for Israeli technology to thrive.” ■