Momenta's Take: ChatIoT

ChatIoT – Generative AI in industrial applications

Generative AI technologies are top of mind across industries and a key focus for early-stage investors. While traditional AI will describe, predict or prescribe something based on existing content, Generative AI differs in that systems can generate output that doesn’t already exist – whether a new image, text, narrative, or design, for example.

Most importantly, Generative AI empowers uniquely creative uses of the technology. Generative Pre-trained Transformer (GPT) technology (the underlying technology for ChatGPT) uses large-scale language technology that uses deep learning to produce human-like text and code. According to CB Insights, there was a record $2.6bn in investments in Generative AI startups 2022 across 110 deals.

The adoption of Generative AI is accelerating and is expected to impact industrial contexts significantly. Initial use cases are already seeing rapid adoption in marketing and sales (for creating outbound content) and customer service (automated chatbots).   Generative AI automated coders have improved software developer productivity by over 50% in some cases. Gartner forecasts that by 2025, more than 30% of new drugs and materials will be systematically discovered using Generative AI techniques, up from zero in 2022.  

For industrial applications, Generative AI has the potential to meaningfully increase efficiency and improve productivity and decision-making across manufacturing, logistics, energy, agriculture, and other sectors. Some of the most promising areas include:

• Product and Parts Design - Generative AI algorithms can iterate on and improve designs based on user-defined parameters, helping save time and resources. Additionally, Generative AI systems can be used to create more realistic simulations that can be used to optimize performance, identify potential issues and improve operational efficiencies.
• Materials Science – Generative AI can benefit the automotive, aerospace, tech, medical, and energy industries by employing inverse design, defining the desired properties first. The system works to discover materials that possess them rather than relying on trial and error.
  
• Code Generation for IoT applications - Large language models can be used to create, combine, or complete software code based on small snippets of code or descriptions. These capabilities can help both experienced and inexperienced developers build applications more quickly and efficiently, complementing the range of no-code/low-code tools that have become increasingly available.
  
• Robotics Programming and Control – Generative AI can generate control logic and commands for robots by capturing motion data (from humans or animals, for instance).   Generative models can create multiple individual part movements rather than having to program complete systems deterministically. They can also help robots interpret “horizon” goals while assisting in filling in the intermediate steps without human intervention.
  
• Improving Human-Machine Interfaces – Generative AI can help improve the user interface for SCADA systems. By understanding operator inputs, intent, and context, models generate relevant responses, provide real-time assistance, and provide more intuitive interactions.  
  
  

Similar to traditional AI, the successful adoption of Generative AI depends on choosing and training the appropriate models, which include transformer-based models, generative adversarial networks (GANs), and variational autoencoders (VAEs) – then selecting and training appropriate data sets that optimize for the specific application and desired outcomes. While it’s accurate to describe Generative AI as evolutionary, the potential impact of these systems is poised to be transformational.

_____________