There is incredible hype about the disruptive potential of artificial intelligence (AI). But many experts say the key ingredient to getting the most out of this technology is simply finding the right use case for the business.
Thierry Martin, head of data and analytics strategy at Toyota Motors Europe, tells ZDNET how the auto giant is dedicating time and resources to R&D to harness the potential of AI.
And this involves, above all, data science. “Data analysis is very important,” says Martin. “For example, it allows us to know how people drive our cars. Or if there is a difference in motorway driving between Germany and Belgium.”
First mission: data collection
And this analytical work depends on data collection, an area in which Toyota excels. “We can already get a lot of information about how people use our cars,” he explains.
For now, Martin says Toyota is focused on using tools like Power BI to keep humans at the heart of the process and to develop a detailed understanding of automotive operations and processes. “We don’t let AI make decisions for humans,” he says. “We prefer to provide more information.”
Yet in the not-so-distant future, Thierry Martin envisions a situation where his organization begins to leverage AI in production. And explorations to find the right use cases for business processes are already underway. “We have a high demand in this area,” he says. “There are a lot of use cases around text data analysis and generative AI, which has become possible since 2022 and the launch of ChatGPT models.”
Microsoft Copilot or in-house prototyping?
While OpenAI’s Large Language Models (LLMs) have helped raise awareness of generative AI, Toyota – like many large companies – is proceeding with caution. In the case of Toyota Europe, Mr Martin suggests two paths forward to get the most out of AI.
- The first will be to use tools like Microsoft Copilot on a personal level to help people accomplish tasks using non-sensitive data.
- The second route, through which his team is working on prototypes, involves using generative AI behind a corporate firewall.
“In terms of prototyping, we work a lot on chatbots,” he explains. “We code chatbots ourselves at the moment. Once a library has been set up, it’s very quick to install it and try it yourself. It’s not difficult. ”
The crucial role of the data mesh
Toyota Europe’s work with AI is supported by the creation of a data mesh, which Thierry Martin describes as an approach to governance that ensures that responsibility for data products remains in the hands of the business owners.
The organization brings together its information on a Snowflake platform, which guarantees access to data with good governance.
The data mesh leverages other technologies, including Dataiku for collaboration, Collibra for governance, and Denodo to connect the data mesh across different parts of the organization, such as Toyota Europe and Japan.
Chatbots based on company data
Thierry Martin and his team use these data mesh technologies to explore AI. They have already built chatbots on Dataiku, which uses an LLM running on a secure instance of Azure Open AI to provide PDF summaries.
He demonstrated the chatbot to executives at Toyota Europe and suggests that in-house development is the way to go because it helps alleviate some of the concerns associated with publicly available models from suppliers. “We already have our own linguistic model,” he explains. “It’s on Azure, but it’s secure. Plus, because we have the LLM and a chatbot, we can build our database and create interactive chatbots.”
Martin says his team continues to explore the area of generative AI: “We’re building a knowledge retrieval system, because there’s a lot of knowledge scattered throughout the company. But We are still at the pilot project stage.”
In all cases, the watchword is “test”
In all cases, the watchword is “testing” to ensure that services meet strict governance requirements and user demands.
“We want to confirm the value and we also need to confirm how to scale it,” he explains. “Once you start using a chatbot service, for example, there are AI ethics and governance that we have to put in place. If I start deploying a chatbot, I have to answer a lot of questions on ethics.
So when could this broader implementation take place? Thierry Martin says he would like to see some AI-based tools put into production fairly quickly. It works with its technology partners, including Snowflake, to ensure governance issues are considered and access to data is limited.
The question of data governance
“The vision should be that a logistics chatbot only has access to logistics data and not HR data, just like an employee only has access to certain data,” he explains.
Martin says Toyota Europe is continuing its prototyping work and could have an AI-based chatbot service that pulls data from the Snowflake platform by mid-2024. It is also talking to other technology partners, such as Dataiku and Collibra.
More importantly, it will work to demonstrate its AI services and study how these tools could work in specific areas of the organization. “We need to understand where is the best place to operate the chatbot,” he explains. “That’s why it’s very important that engineers and managers really understand what we’re talking about.”
Source: “ZDNet.com”