What to Watch in Robotics
Four founders highlight trends in robotics, planning for AI products, startup exit timing, integrating product with marketing, jobs, events, and more
It’s great to be back in your inbox yet again with Startup Pirate #92. This is a newsletter about entrepreneurship, technology, and startups. Made in Greece. If you haven’t subscribed, join 5,575 readers by subscribing here:
What to Watch in Robotics
Many people who lamented the seeming incrementality of technological progress in the 2010s would complain that we weren't getting the future we envisioned in the 1960s. "We wanted flying cars, instead we got 140 characters". Robotics represented another area of underachievement: limited versatility, extremely expensive, hyper-specific capability rather than solving more general problems. That may now be changing.
A lot has happened in the first weeks of 2024, from robots cooking shrimp (Mobile ALOHA) to making coffee (Figure) just by observing humans perform the tasks. Or semi-autonomously folding the laundry (Optimus). Last year, we caught a glimpse of the future foundation models and platforms for robots with multimodal LLMs and robot arms as a physical I/O device; RvT (NVIDIA), RT-1, RT-2, PaLM-E (Google), RoboCat (DeepMind), etc. We are probably a few years away from the ChatGPT moment for physical AI agents.
To better understand the state of the industry, I asked four founders to share the trend they believe is worth watching and what they are working on. I hope it helps us better identify areas of opportunity and topics worthy of further research. We are going to discuss:
The need for general robotic intelligence
Key components of the modern robotics stack
Applications in industries like agriculture
Why the robotics revolution should be accessible to SMEs
Let’s get to it.
Giving robots the sense of touch
Robotics has undergone significant evolution over the past decades. The advancements we currently observe with humanoids, legged and mobile robots, and companies having a production capacity of 10,000 robots annually are years in the making. We start seeing applications in the physical world that were previously unthought of being automated, from surveillance and inspection to manufacturing, the assistance of disabled humans, and more. As a robotics nerd at NTUA a decade ago, developing a sensor for a small robot to move like an animal, I certainly didn't expect this rate of progress that fast.
Converting robots from insensitive machines to intelligent tools is an uphill battle, and the modern robotics stack still requires significant improvements. We humans take signals from the environment through our sensory system, and these offer key inputs for how we further interact with the world. It’s a closed feedback loop. Typically, robots miss that level of responsiveness in tasks that require high agility and dexterity. Therefore, on many occasions, an operator is still required to refine the robot’s movements and adapt when the robot fails to detect changes in its environment or perform in ways it was not pre-programmed for.
At Bota Systems, we give robots the sense of touch through advanced, in-house developed force torque sensors with built-in electronics and AI. These are integrated into robotic arms, unlocking new skills that are critical for a number of functions, which have yet to be adequately served, such as polishing complex surfaces, assembling delicate parts, electric vehicle charging, and others. We consider force sensing a key enabler for automating tasks that require a detailed and accurate feedback loop between the environment and the robot, which is how we are going to unlock the next stage of the robotics revolution.
Klajd Lika, founder & CEO, Bota Systems
General intelligence for robots
Vinod Khosla predicts that within 25 years, there could be more than a billion bipedal robots performing daily tasks. He envisions the potential to liberate humans from the least desirable jobs, from factory workers to farm labourers and more. Today, industrial robots within advanced manufacturing, as well as humanoids, can only perform simple tasks (like moving something from A to B in a predetermined fashion); they don’t exhibit the general intelligence required to perform complex tasks in dynamic, real-world environments. General Robotic Intelligence (GRI) is still missing.
We need a platform that encapsulates the complex interactions of the physical world, allowing manufacturing robots and humanoids to adapt safely and accurately in unstructured, real-world environments, making them commercially viable. They can only do so by learning from vast amounts of physical world data powered by AI, yet existing teaching methods don’t scale to unstructured environments. They are based on control devices, virtual simulation environments (developed offline and then downloaded and tested on robots), or kinesthetic teaching and teleoperation methods, where the robot is physically guided through the task by humans. General intelligence for robots requires high-quality data that needs to be collected quickly, cost-efficiently, at scale, and agnostically of hardware to accelerate the trajectory of global robotic adoption.
At Acumino, we have developed a General Robotic Intelligence platform that enables connected robots to perform complex tasks across any industry with unprecedented efficiency and scalability. We do that through advanced human-to-robot skill transfer and proprietary robot models trained on very large physical world datasets, which support the execution of highly accurate and dexterous tasks. This platform provides an intuitive way for any company to translate human dexterity and expertise related to the execution of complex tasks into efficient, optimized robotic actions.
Minas Liarokapis, Chief Robotics Scientist, Acumino
Revolutionising agriculture with Robotics as a Service
The growing global population and the urgent need to feed people in a sustainable way have forced the agricultural industry to rethink how food is produced. Increasing crop yield, following environmentally friendly practices, and reducing reliance on human workers are now top-of-mind for most farmers worldwide. These have led to a rapidly growing agricultural robot market.
Pruning, weeding, spraying, and monitoring are indicative activities where robots can be involved, with recent trends focusing on vision-based robot navigation and decision-making systems that mimic human farmers. Yet, high upfront costs (over $200,000 in some cases), labour-intensive deployment (known as the last-mile problem), and the high technical know-how required to operate and maintain such sophisticated machinery leave most agricultural robots beyond the reach of the average farmer. These are problems my family faced running a mid-size vineyard business in Northern Greece.
I started Agroverse to make state-of-the-art robotic technology accessible to every farmer and spearhead Robotics as a Service (RaaS). At the core of this is the “Nest” — docking stations placed in convenient spots inside agricultural lands that serve not just as charging hubs but also as launchpads for autonomous operations. Upon receiving a request from a farmer, the robot autonomously navigates to the field, performs tasks like weeding or spraying, and then returns to its Nest to recharge. Farmers in the vicinity share the same robot, hence this service model allows them to pay on a per-acre basis and receive detailed reports of the work done, with the added peace of mind that comes from remote oversight by safety pilots via 5G in case of any issues. Thanks to machine learning models trained on multidimensional data from farmers operating tractors across various crops, the robot can be easily adapted to environments in which the machine has never been exposed.
Dimitrios Koutras, founder & CEO, Agroverse
Making robots accessible to SMEs
Factory floors and production lines are currently being reimagined thanks to robotics automation. And we should be thankful for that, as there are activities no person is meant to do. There is a reason for the high prevalence of musculoskeletal and chronic injuries among factory workers. Who can stack 20kg boxes in pallets that are 2m tall every day for years? Yet, such tasks are indispensable from the smallest to the largest production lines.
Industry leaders can afford to use multiple robotic arms, where each arm is responsible for pick-and-place the same object type 24/7, but things usually fall apart when there are product variations. That’s increasingly the case for production lines run by SMEs, which don’t produce the same product day and night.
We started Progressive Robotics to make robots accessible to small and medium-sized businesses and easy to use so that every person in a factory can “program” it in just a few seconds. Using machine vision with AI and online motion generation, we make robotic arms more versatile to handle various objects with different shapes and colours, enabling use cases such as dense packing, palletising, unpacking, sorting, etc. Robots automatically detect how to pick and place objects after teaching them via advanced controllers, human demonstration, and an easy-to-use interface. With applications in manufacturing and warehouses, we aim to provide solutions that even SMEs can benefit from and help them stay competitive.
Fotis Dimeas, CEO & co-founder, Progressive Robotics
Check out job openings here from startups hiring in Greece.
Rooted in Greece since 1997, defence tech company THEON, which develops night vision and thermal imaging technologies, launched an IPO on Euronext Amsterdam. (link)
End-to-end software platform for carbon project developers BlueLayer comes out of stealth with $10m in funding from Point Nine and other investors. (link)
MeduSoil raised 5.2m CHF to develop biomineral-based binder products in the construction materials sector. (link)
Audio studio for creators powered by AI Wondercraft raised $3m Seed. (link)
Sentify, a platform enabling users to extract actionable insights from LLM products, announced a $1.1m Pre-Seed. (link)
Mental health platform for seafarers, Wellness by Learning Seaman, raised funding from the Northern Greece Investment Fund. (link)
Nuuro announced funding from Genesis Ventures to revolutionise nutrition with AI. (link)
Future BrAIn, a startup that offers insights into organizational culture, engagement, and team dynamics, announced a Pre-Seed round. (link)
Eight new funds to invest in startups and beyond launching in Greece in the next few months. (link)
Choosing the right startup exit timing with Dimitris Glezos, founder of Transifex. (link)
Konstantinos Giamalis, Chief Product Officer at efood, on why product and marketing need to work together. (link)
How goal-setting and planning are different for AI products with Anastasis Germanidis, co-founder & CTO at Runway. (link)
Antonis Kalipetis and Paris Kasidiaris discuss the state of the Greek tech scene on Mikri Kouventa. (link)
Nine steps for business onboarding as a Product Manager from Joseph Alvertis, VP of Product at TileDB. (link)
AI, regulation, careers in AI, and more with Lilian Balatsou, Conversational Intelligence Chapter Lead at Vodafone. (link)
UX Greece welcomes Jakob Nielsen by UX Greece on Feb 14
How ElasticSearch empowers Transifex's Translation Memory by Elastic Greece on Feb 15
Meetup #8 by React 2 React Athens MeetUp on Feb 15
Fintech Forum by Mastercard on Feb 20
Meetup #4 by ProductTank Athens on Feb 21
Multi step flows with vuejs & Laravel by Athens Laravel Meetup on Feb 22
That’s all for this week. Tap the heart ❤️ below if you liked this piece — it helps me understand which themes you like best and what I should do more.
Thanks for reading,