Behind the Talk: Generative AI for Intelligent Supply Chains

Slow, manual supply chain planning processes can be a thing of the past, with AI Agents taking on repetitive tasks that aren’t a good use of human capacity.

Sep 18, 2024

Last week, I spoke on Gen AI innovations in supply chain management at ASCM Connect 2024: North America. My full presentation is available for download. To dive deeper, I cite the publicly reported use case examples below.

Delving In

[2406.07016] Delving into ChatGPT usage in academic writing through excess vocabulary

Recent large language models (LLMs) can generate and revise text with human-level performance, and have been widely commercialized in systems like ChatGPT. These models come with clear limitations: they can produce inaccurate information, reinforce existing biases, and be easily misused. Yet, many scientists have been using them to assist their scholarly writing. How wide-spread is LLM usage in the academic literature currently? To answer this question, we use an unbiased, large-scale approach, free from any assumptions on academic LLM usage. We study vocabulary changes in 14 million PubMed abstracts from 2010-2024, and show how the appearance of LLMs led to an abrupt increase in the frequency of certain style words. Our analysis based on excess words usage suggests that at least 10% of 2024 abstracts were processed with LLMs. This lower bound differed across disciplines, countries, and journals, and was as high as 30% for some PubMed sub-corpora. We show that the appearance of LLM-based writing assistants has had an unprecedented impact in the scientific literature, surpassing the effect of major world events such as the Covid pandemic.

On 30 September 2012, a convolutional neural network (CNN) called AlexNet achieved a top-5 error of 15.3% in the ImageNet 2012 Challenge, more than 10.8 percentage points lower than that of the runner up. Using convolutional neural networks was feasible due to the use of graphics processing units (GPUs) during training, an essential ingredient of the deep learning revolution.

2014 [1406.2661] Generative Adversarial Networks
2015 [1506.02640] You Only Look Once: Unified, Real-Time Object Detection

Generative AI

2015 [1503.03585] Deep Unsupervised Learning using Nonequilibrium Thermodynamics (Diffusion)
2017 [1706.03762] Attention Is All You Need (Transformers)
2022 Introducing ChatGPT | OpenAI

Why Now?

2024 Supply Chain Executive Survey Infographic | Business Wire
2024 Supply Chain Executive Survey E-Book | Blue Yonder

Let the Machine Do the Job

2018 The route to no-touch planning: Taking the human error out of supply chain planning | McKinsey

How GenAI is Impacting Supply Chain Management

Suppliers

Supplier What If Simulation

Using Gen AI to Optimise the Supply Chain - Interview with Michel Morvan - Cosmo Tech

Notable Disruptions

Factories

Harnessing the Power of Databricks Mosaic AI for Image Generation at Rolls-Royce

Warehouse

Covariant Robotics Foundation Model

Transportation

United Airlines Hopes Oversharing About Delayed and Canceled Flights Will Win Your Trust - WSJ

Today United sends crafted messages to passengers on about 20% of delayed flights, according to Jason Birnbaum, the airline’s chief information officer. The figure is growing as the airline enlists AI to help with the messages, crafting messages off the same info the storytellers collect.

New C.H. Robinson Technology Breaks a Decades-Old Barrier to Automation in the Logistics Industry
Transform Supply Chains with Generative AI | C.H. Robinson
How CH Robinson is using AI to automate logistics tasks | Trucking Dive

While the technology is replying to 2,000 customer quote requests a day, it opens the door to automating other transactions shippers and carriers choose to do by email. The large language model (LLM) the technology uses can be trained to identify an email about a load tender, a pickup appointment or a shipment tracking update. For spot quotes, C.H. Robinson has already trained the model to differentiate between a quote request for truckload, less-than-truckload (LTL), intermodal or air freight.

Supply Chain Information Highway

Point of Sale

Walmart used large language models to create or improve more than 850 million pieces of data across its product catalog. Without generative AI, the process would have required 100 times the head count to complete in the same amount of time, executives said.

How Amazon continues to improve the customer reviews experience with generative AI

Customer Experience

What to consider when procuring GenAI capabilities

Platform Approach vs Point Solution

Foundation Models: Open vs Closed

Maxime Labonne on X: "I made the closed-source vs. open-weight models figure for this moment."

[Open] Llama: The open-source AI model you can fine-tune, distill and deploy anywhere. The latest instruction-tuned model is available in 8B, 70B and 405B versions.
[Closed] OpenAI releases o1, its first model with 'reasoning' abilities

Model Capability: Quality x Speed x Price

Techniques & Architectural Patterns

There are four architectural patterns to consider when building a large language model–based (LLM) solution, including prompt engineering, retrieval augmented generation (RAG), fine-tuning and pretraining. Databricks is the only provider that enables all four generative AI architectural patterns, ensuring you have the most options and can evolve as your business requirements change.

Repeatability, Governance, and Security

Databricks AI Security Framework (DASF)
- Map risks to common AI security frameworks
- Get actionable recommendations on 59 controls that apply to any data and AI platform
Winning at GenAI: Building the right processes for the data intelligence future

Avoid data poisoning and other key risks for sensitive GenAI applications, similar to using pre-nuclear age metals for sensitive equipment. Be cautious of training on data generated by LLMs that is now spread across the Internet.

Low-background metal: Pure, unadulterated treasure
- The radiation isn’t strong enough to pose a health risk, but it does interfere with some sensitive scientific and medical equipment. It’s hard for a geiger counter to accurately measure radiation if the metal it’s made from is, itself, radioactive.

Where is GenAI headed next in Supply Chain Management?

Slow, manual supply chain planning processes can be a thing of the past, with AI Agents taking on repetitive tasks that aren’t a good use of human capacity.

AI agents rationalize a plan and execute it with tools

What are compound AI system and AI agents? | Databricks

AI/BI: AI-powered chatbots and dashboards accept natural language prompts to perform analysis on a businesses’ data, drawing insights from the full lifecycle of their data. AI/BI agents parse requests, decide which data sources to, and how to communicate findings. AI/BI agents can improve over time through human feedback, offering tools to verify and refine its outputs.
Customer service: AI-powered chatbots, such as those used by customer service platforms, interact with users, understand natural language, and provide relevant responses or perform tasks. Companies use AI chatbots for customer service by answering queries, providing product information, and assisting with troubleshooting.
Manufacturing predictive maintenance: AI agents can go beyond simply predicting equipment failures, autonomously acting on them by ordering replacements, or scheduling maintenance to reduce downtime and increase productivity.

Exponential Industry

Discussion about this post