AI-Driven Industrial 5.0 Reimagines Manufacturing and Robotics

Key Quotes

"Now foxconn is the largest manufacturer in the ict industry and we used to be a very much labor intensive and then we transformed to automation intensive with new generative ai technologies we think the ai intensive manufacturing is coming and with this new technology this is new and disruptive we'll have to work with industrial leaders or technology leaders like nvidia siemens and the friends at this table together to be able to catch up and apply the new technologies to our manufacturing facilities."

Young Lu explains Foxconn's evolution from labor-intensive to automation-intensive manufacturing, highlighting the transformative potential of generative AI. He emphasizes the necessity of collaboration with technology leaders to integrate these new, disruptive technologies into their facilities. This indicates a strategic shift towards an AI-intensive manufacturing paradigm.

"we think of it as more of an ai native factories that we want to build so it starts with the sensors right i mean the sensors that you all carry and that that we all carry we love the data and because we got so many sensors now everywhere humans cannot make sense out of this anymore so therefore you need to automate that and so as we build our factories we build them always twice we build them first in the digital world and we optimize them and we see how you know you place the machines how the material will be flowing how humans will be interacting with these machines and then we optimize them over and over and over again until we finalize think this is great then we build the real thing."

Peter Kutta describes the concept of "AI-native factories," where data from ubiquitous sensors is leveraged. He explains Siemens' approach of building factories twice: first in a digital twin to optimize design and operations, and then constructing the physical facility based on these optimizations. This methodology aims to enhance efficiency and address the limitations of human data processing.

"so for us we have this like kind of phrase we use when we're trying to give ai a body and to do everything a human can so that means everything from pre training like we we have to basically build like large scale data collection efforts of human like data to do training we we use nvidia there and then that means at test time when we're running policies on robot we're doing inference on nvidia gpus on robot without any network connection so we can basically run robots in full indian situations like doing work without any network outside network so so for us we think of ourselves like an ai business we happen to be building like these physical agents that are out in the world like similar to web agents basically just like in the physical world touching things."

Brett Adcock articulates Figure AI's mission to create humanoid robots capable of performing human tasks, emphasizing the need for AI with a "body." He details their reliance on large-scale human-like data for pre-training AI models and the use of NVIDIA GPUs for on-robot inference, enabling autonomous operation without external network connections. Adcock frames Figure AI as an AI business that creates physical agents.

"so if you go back to the core of what palantir is all about you have all these data systems you have erps mrps random software that people have built in the ecosystem and all of these kind of systems solve a specific problem they have different security frameworks and whatnot how do you provide the ergonomics for an autonomy stack whether it's for an autonomy stack whether it's for a factory worker whether it's just for a human to actually answer questions about that business and optimize it and that's really what the ontology is it's kind of that agility layer."

Aki Jane explains Palantir's role in integrating disparate data systems within an organization. He defines their "ontology" as an agility layer that provides ergonomics for autonomy stacks, enabling humans and AI to interact with and optimize business processes. This layer addresses the challenge of making complex, siloed data accessible and actionable.

"yeah well i think right now at least from palantir's perspective both across their half of their business is commercial half of their businesses global government and what i would say is look we always start with the big heavy cloud model um from from the frontier lab companies and there were proprietary models i i gotta be honest they're moving so fast and they're so effective um so we use them you know to solve any problem is the first thing we try but we're finding more and more especially as we move to the edge um we think about the need to sort of take those models train them on bespoke data and then move that towards kind of more of an edge inferencing architecture that solves specific problems either with a small language model or even something that maybe is just a traditional machine learning model for a lot of problems where data becomes critical."

Aki Jane discusses Palantir's approach to AI models, starting with large proprietary cloud models for their effectiveness. He notes a growing trend towards edge inferencing architectures, especially for specialized problems, which involves training models on bespoke data. This shift highlights the need for adaptable AI solutions that can operate efficiently at the edge.

"i think what really matters today is we see the core like horizontal technology stack both in hardware and software of ai come together to build like general purpose like almost like human like intelligence in the physical world like that's the key unlock like i can't stress that enough like people are looking past that in terms of cost and manufacturability and all this different stuff we're at a stage now where we got to go build like synthetic humans and it's an and it's incredibly hard and i think what we are proud about at figure is we've been able to now to show these like real like pockets of like long horizon intelligence done with neural nets."

Brett Adcock emphasizes the critical importance of the core horizontal AI technology stack in both hardware and software for developing general-purpose, human-like intelligence in the physical world. He asserts that this foundational capability is the key unlock, often overlooked in favor of discussions about cost and manufacturability. Adcock highlights Figure AI's progress in demonstrating "long horizon intelligence" using neural nets.