Autotech Ventures is excited to announce our investment and partnership with Latent AI following their recent seed round of funding.
Autotech Ventures is a venture capital firm focused on transportation tech and backed by 30+ transport corporations and financial limited partners. We are excited to partner with Latent AI to help deploy ML applications inside vehicles, at manufacturing facilities, and across the transportation sector.
The Latent AI solution is a critical enabler for the current explosion of ML applications to be practically deployed on all types of hardware and products.
Why the Edge?
Machine learning as an enabling technology is just getting started. There are many open-source tools and frameworks making it easier for engineers and developers to create exciting new applications — and the use cases are exploding.
The challenge has been that AI training and inference is computationally expensive and has been limited to running on GPUs or in the cloud. For example, in the last 5 years, Nvidia, a GPU provider, has experienced a 20X increase in their stock price while commodity chip manufacturer, Intel, has only seen 2X growth of their stock price.
The value creation gap between Nvidia and Intel suggests that there is a massive opportunity if you can deploy high-quality AI on any commodity chip. That is exactly what Latent AI does.
Latent AI optimizes AI algorithms for accuracy, power, memory, and latency requirements, and can compile the model onto specific hardware based on these requirements.
In other words, Latent AI’s Adaptive AI™ technology brings AI to the Edge allowing any edge environment (from cloud to microdevices) to have full ML capabilities – regardless of hardware target, architecture, framework, or OS. Latent AI doesn’t require internet connectivity and ensures personal data is not shared in the cloud.
Why a Software Solution vs. New Hardware Platforms?
Specialized AI chips for certain high-value tasks like autonomous driving already exist, but even these chips require optimized algorithms. However, there are thousands of different types of processors on vehicles today. Rather than replacing each of those with an AI chip, Latent AI allows those processors to effectively become “AI chips” themselves.
We previously invested in DeepScale (acquired by Tesla) and Xnor.ai (acquired by Apple for $200M) focused on algorithmic improvements and proved that algorithms developed by these startups can help bring sophisticated AI to any device.
Latent AI goes further by allowing customers to develop their own proprietary algorithms and use cases, then Latent AI compresses and compiles those algorithms on the optimal hardware. We are excited by the prospect of a highly scalable tool that puts edge AI capabilities in the hands of any interested customer.
Why Edge AI is Critical for Transportation Applications
For example, in Autotech Ventures’ sector of focus, automakers are deploying AI algorithms for automating vehicles, to detect quality in their manufacturing facilities, to detect potential onboard mechanical failures before they happen, protect against cyber-attacks, monitor drivers and passengers to improve safety, and probing the outside environment. Some of these non-autonomous AI applications are described in this TechCrunch article.
Again, this is critical for automotive, logistics, and manufacturing use cases. Electric power is limited on many vehicles and for electric vehicles, power consumption for AI applications directly reduces the vehicle’s range. Areas of low or no connectivity are ubiquitous across the transportation network. Drivers are sensitive to having in-cabin images shared outside the vehicle. Edge AI processing, via Latent AI, can mitigate these risks and enable the continuous functionality of these features.
We’re excited to see the explosion of edge AI in the automotive, manufacturing, and logistics sectors and looking forward to Latent AI enabling new edge applications in our field.
by Jeff Peters, Partner, Autotech Ventures
Jeff Peters brings to Autotech wide-ranging engineering, economics, and investing expertise in the transportation industry and beyond. He has a passion for partnering with startups that have defensible business models backed by strong founder-market fit – especially expert founders bent on deploying novel technology in under-served market segments.
Jeff has a Ph.D. in Applied Economics, MS in Economics, MSE in Transportation Engineering from Purdue University – where he co-founded TEDxPurdueU, and BSE in Aerospace Engineering from the University of Michigan.