Beewise Co-Founder Saar Safra: Automating Beekeeping with Robotics for Alive and Thriving Bees

I was new to Israel as I've lived most of my life in the US and got back to Israel because of my parents. A few months after I arrived, I've met Eliyah Radzyner, a beekeeper who later became a co-founder of Beewise. A beekeeper by profession, Eliyah is actually an entrepreneur but he doesn't know it. He had this great idea - autonomous beekeeping. He saw that in 2019 we were still doing beekeeping the same way as in 1850 and was convinced that with the help of software, hardware, or a combination of the two we could alleviate some of the stress the bees were having and automate the processes. Thus, we started playing with a few ideas. In the beginning, I was thinking: "Who cares about bees? One cannot build a startup around bees!" At the end of the day, I was just a software guy who knew nothing about bees. But I wanted to help Eliyah and decided to do some pro bono work for a few months. Over the next six months, I realized the amazing potential impact and opportunity that exist here. That's how I got hooked and started running the company.

Bees are crucial to the global food supply. 75% of all the food - vegetables, fruit, seeds, and nuts we eat - is pollinated by bees. It is incredible. One can of course survive by eating corn and rice. However, we know that a healthy diet is a prerequisite for longevity and long healthy life. So, eating fruit and veggies is important and this is why bees are critical. At this stage of the company, our vision is to save the bees of the world. We strongly believe that by using modern technologies like Artificial Intelligence, Machine Learning and Computer Vision applied properly, we can save all the bees of the world which are becoming increasingly extinct.

It is an intervention, you are right. We absolutely need one as without bees we are not getting the critical food supply. Without bees, there are no almonds, avocados, cucumbers, tomatoes, coffee, and cotton. So, we can let the bees go but we will also have to let go of things that we like to consume. We can decide not to intervene, but there will be consequences.

Yes, without question. I will give you an example. TVs used to cost a lot of money back in the day but they got cheaper and commoditized. Today a 42 inch TV costs only a few hundred dollars. Any product with enough demand gets cheaper when there is plenty of supply. However, with fruit and vegetables, it is not what we see. Today it is not much cheaper to buy an avocado. The prices fluctuate and they do so because there is no steady on increasing supply to close the gap for the demand. So there is more and more demand for fruit and veggies. This is a part of the educational processes that we are going through. An increasing number of people consume vegetables and fruit but the prices are not becoming commoditized, at least for pollination dependant crops. If you have an almond field and you need a thousand hives to pollinated it and you get only five hundred hives, you are going to get less product than you could have gotten. Because there are fewer bees, there is less pollination. Today in the US and Europe, they get about 46% of the actual pollination they need to optimize the production.

Protecting the bees is almost a national security issue as no country wants to be 100% dependent on the import of food. If you want to avoid that issue you need your own agricultural sector for which to exist you need an infrastructure of bees. Almost every country has an active bee industry to pollinate agriculture so that they have their supply of produce. It sounds crazy but this is how it works through the whole value chain. That's why countries with large bee populations like the US, Europe, Russia, Ukraine, Belarus, and many others have national bee protection programs.

Indeed, we are creating problems for the bees. On the one hand, we are dependant on the pollination dependant crops. On the other hand, the bees are experiencing a huge decline. We see about 35-40% of all the bees die in the world every single year. Why is that? Pesticides, pests, disease, climate change, and many other factors stress out the bees. For example, if you put your hives on the avocado field, the bees don't know where the limits of the field are. They just fly out for up to 3 km. They can fly to your neighbors who put pesticides on their field and your bees die. And you might discover that they are dead when you come back to visit them only in three weeks. We do a lot of harm to the bees. The bees have been living for over 100 million years ago but only in the last 50 years, we witness their increasing decline. In Florida, for example, they've recorded a 40% decline in the last year alone.

If you look at the device behind me, it has beehives on both sides with entrances via which bees come in and out to the fields and bring honey and pollen. In the center of the device, there is a robot that monitors the device with Computer Vision. We use a convolutional neural network to do that which has been trained by us for the past couple of years. The network converts images into data allowing us to identify the specific needs of the bees. Our Computer Vision algorithms can spot disease, a pest, a lack of water and food, honey ready to be harvested. Then, we have a recurrence neural network that does Machine Learning and Reinforcement Learning where we try various treatments and overtime and learn which ones were most successful. So, we apply antibiotics when we spot a disease, we take care of the pests, we harvest the honey, we feed the bees with food and water - all that being done by the precision robotics into every single hive in real-time in the field autonomously.

Think about commercial beekeepers which we deal with. Today, they visit the hives on average once a month. If you are managing ten thousand hives, they are usually scattered with a few hours drive in between. You spend on average 2-5 hours to get to the hives and then you have to treat them. Everyday beekeepers do that to a different set of hives. A month goes by when they come back to the original hives since there are so many hives and locations they have to manage. Beewise is not replacing the beekeepers. Instead, we complement them so that there is a beekeeper 24/7. It is called "eyes on hives". Beewise robot essentially replaces the beekeeper all those 29 days that the beekeeper is not there. Now the beekeeper could sit in front of their computer and see everything that's going on in real-time. Our solution allows them to distribute the treatments, intervene in robot's decisions, and choose between data-driven options. Beewise allows them to control five thousand hives instead of visiting a few hives once in a while.

Beekeeping has been going on for thousands of years. So, the basic knowledge was out there when we started. To train algorithms, we started with the basic protocol the beekeeper usually follow. On top of that, we put a neural net that learns which steps were more effective over a portfolio of hives toward the bee.

The bees love it more than the regular hives. The regular hive is essentially a wooden box invented 150 years ago, the last time when this field saw technological disruption. These wooden hives are out in the field not protected from heavy rain, excessive heat, cold, bears hunting for honey, or human theft. In Beewise device, the bees are much more secure and safe. The whole device is thermoregulated and would heat the hive when it gets too cold. When the beekeepers open the hives to take out the frames, they disrupt the bees. In our device, it takes about 4 minutes to take out the frame to make it seamless and soft for the bees so it doesn't disrupt their ongoing day.

The idea is that Beewise is a bee-centric system. Everything we do here revolves around the bees. The bee is one of our customers. You can do that only with the robot because the robot doesn't get tired, doesn't go on vacation, and does its job consistently well.

I'd say hardware is hard, we have a lot of robotics, mechanical and optical systems in there which are complex. On the other hand, our convolution network - our AI that takes pictures for Computer Vision is less difficult as we have lots of data. Every frame, every honeycomb has six thousand cells and we analyze each of them. What happens is when you take a picture of the hundred frames, you train your system on 600,000 instances. The Reinforcement Learning or the recurrent network takes a bit longer to train and this is where beekeepers come in as they get to say yes or no to our proposed course of action. As a result, they help us train the system allowing us to understand which actions worked better on a portfolio level as well as per beekeeper.

Yes, Beewise is fully operational and we already manufacture at the scale of a few hundred devices per year and hopefully, soon we'll be able to go from hundreds to thousands and tens of thousands and start delivering them globally. Currently, we are mostly shipping to the US and Israel since we are here. Hopefully soon, we will expand to Europe, most probably the Eastern block countries with high bee population like Belarus, Ukraine, Georgia.

The Covid-19 pandemic has impacted us a lot in terms of travel. With a device like this, you want to give your first customers a super professional experience with a high level of attention. With Covid-19, our travel was restricted which made it difficult to spend time with customers in the field. Everyone on our team including me already got vaccinated and we hope to travel more soon which should expedite our go-to-market and delivery.

This is startup number six for me and I've seen AI evolve a lot since the release of the TensorFlow library eight years back. AI has already changed things in enterprise software, eCommerce, security spaces. But when you take AI and through it into laggard 200 years old industries which haven't gotten through any evolution, you witness a huge impact right off the box. In our case, we've seen a huge improvement of 200-300% in honey and pollination. Strategically and philosophically, AI has a huge potential in many areas and we haven't even started tapping into many of them. My advice is to snoop around areas of agritech, industrial sectors, construction. Think about old tech. Crack those areas and you will see a lot of magic!