From an AI perspective, the cost, power, latency, and performance limitations of the current AI chip technology are presenting the most significant barriers to advancing the industry as a whole.
How AI will Transform the Economics and Sustainability of the Agriculture Industry
Q&A with Bob Beachler, VP of Product | Untether AI
Tell us about yourself and Untether AI.
My name is Bob Beachler, and I am the Vice President of Product at Untether AI. I have a long history in the semiconductor industry building chip technology that makes our most advanced technology work. At Untether AI, I have been focused on building chips that will make artificial intelligence (AI) a reality across various sectors including precision agriculture.
Our company makes energy-efficient, high-performance AI acceleration chips that enable new frontiers in AgTech robotics. Our chip architecture is optimized for running AI tasks faster and more efficiently in farm machinery such as tractors, combines, drones, robotics, etc. We have developed chip technology that is six times more energy efficient than the leading competitor through our novel At-Memory compute architecture.
Untether AI provides the AgTech industry with a disruptive mix of higher performance, improved efficiency and lower-cost semiconductor technology to realize the full potential of AI.
What’s your overarching vision on how AI will transform the economics and sustainability of the agriculture industry?
It is clear that old solutions addressing agriculture have run their course. AI will play a critical role in replacing old approaches and lore that have been passed down through generations. From addressing labor shortages, and soil burnout, to increasing the overall productivity and profitability that can be achieved from a plot of land, AI is poised to radically transform the economics of the agricultural industry.
However, in the last few decades, despite advancements in farming equipment, much of agriculture still relies on intuition, tradition, and manual labor. The use of herbicides and fertilizers has led to challenges like over-spraying and soil degradation, while labor shortages further complicate crop harvesting. These challenges have set the stage for a new transformation in agriculture—the Fourth Agricultural Revolution— driven by AI technologies that promise to enhance precision and efficiency in farming operations.
What are the technological challenges facing agtech right now?
From an AI perspective, the cost, power, latency, and performance limitations of the current AI chip technology are presenting the most significant barriers to advancing the industry as a whole. The AgTech industry is relying on the same chip technology that was introduced over 25 years ago to improve the PC gaming experience!
Who knew the same chip powering your late-night gaming marathons would also end up zapping weeds on a farm? While those GPUs were fine to battle the weeds of yesteryear, it’s no longer a sufficient technology based on where we’re at today.
It is old, costly, inefficient, and power-hungry. The outdated GPU has gone far in demonstrating what’s possible in applying AI to AgTech; however, compared to where technology is today, the solution is falling behind. Relying on chip technology from 25 years ago will simply not innovate the industry.
Understanding AgTech solutions are relying on old chip technology – which is making it impractical for real-world agriculture – is there a solution to this problem?
We have built an advanced technology that will fundamentally transform the capabilities of today’s AgTech robotics. Our novel At-Memory compute architecture is optimized for AI computation. This is the underlying technology that allows us to build the best-in-class "edge-optimized accelerators for AI inference"– a technical term referring to the chip technology used in farm machinery to make it faster, more precise and energy efficient. This innovation shifts the focus from exploring "what’s possible" to delivering "what’s practical" in real-world applications.
This new technology delivers significantly improved performance with much lower power consumption and lower latency than what’s being implemented in AgTech robotics today. As an example, the current GPU solution takes approximately 2.6 weeks to weed 450 acres of farmland. With our chip technology in robotics weeders, we can reduce the weeding process from 2.6 week to four days!
Your company is working on technology that will reduce weeding time of a 450-acre farm from ~2.6 weeks to ~4 days? Can you explain more about this?
In today’s AgTech solutions, a robotic weeder that uses lasers to kill weeds currently employs 24 GPU processors to provide the computing power needed to address the task. However, the performance of this weeder is roughly 0.5 mph, which verges on the point of being impractical from a cost, power, and reliability standpoint. It is certainly not a practical speed for the average US farm.
By engineering an edge-optimized AI accelerator such as Untether AI’s speedAI inference accelerator into a robotic weeder, it will deliver 2.3 times the performance from a single chip, at a tenth of the power of 24 GPUs. The higher AI performance allows for enhanced weeding performance, increasing productivity while lowering input costs.
Laser weeder robotics built with edge-optimized inference accelerators like our family of speedAI products ensures AI is no longer a bottleneck to weeding a 450-acre farm.
More details on this analysis can be found by downloading our whitepaper - AI in Agriculture: Driving the disruption and transformation of an industry and society.
What’s holding back technological advancements?
The simplest answer is the AgTech industry must take robotic innovation beyond gaming GPUs. However, there is low awareness in the industry of edge-optimized AI accelerators. This technology is at the forefront of revolutionizing what’s possible for AgTech robotics and it will disrupt the current paradigm of old GPU-based solutions.
To make AI applications in agriculture a reality, the industry must provide a collaborative and open architecture platform that allows farm machinery equipment designers to evaluate and deploy solutions that are readily optimized for their machinery.
Untether AI is leading an industry-wide initiative dubbed the “AgBox.” The intent of the initiative is to reduce the barriers to the adoption of more advanced technologies and accelerate the practical use of AI in AgTech.
How can we make AI in agriculture practical and economically viable?
As any farmer knows, it starts with using the right tool for the job. In theory, you can hammer in a nail with a screwdriver if you try hard enough, but it’s more practical if you use a hammer. The same holds for employing the right AI processing engine for the job.
Many different technologies are still coming together to fully realize the vision of the fourth generation of agriculture. We need the right partners to come together to make AI in agriculture a reality. We must foster collaboration between technology developers and the agricultural community. A community-wide collaboration is needed to kick-start and actually begin building optimal AgTech robotic equipment that isn’t using technology from 25 years ago. It’s an incredible opportunity to work with a company at the forefront of leading the transformation of AgTech robotics.
The content & opinions in this article are the author’s and do not necessarily represent the views of AgriTechTomorrow
Comments (0)
This post does not have any comments. Be the first to leave a comment below.
Featured Product
