Generative AI for SMEs with Food Processing Skills Canada

Kevin:

There’s certainly a lot of interest in the machine learning side of AI. A lot of the challenges employers are looking for are trying to get tangible productivity gains from AI, for example, to quantify fitting 8 more products in a box because of AI’s optimization, as opposed to saying my accountant has 4 more hours in the day. 

Our industry is very similar to how the Canadian economy is mostly comprised of SMEs. One of the surveys we’ve done found that ~50% of all food or beverage manufacturers don’t have a dedicated HR person or department. Even for things like hiring new staff, if AI can provide initial output to develop a training plan, it could then be modified to fit particular training needs. 

There’s some unease around the lack of knowledge among small business owners. Some operators are wearing 8 different hats trying to understand different tools. Figuring out how to get started is the biggest barrier. At a conference in March this year, we did a workshop on how to demystify and walk people through examples of how to use Gen AI. With a bit of help from prompt engineering, we could increase peoples’ interest in getting started. 

Mark:

The #1 barrier to GenAI adoption is how to get started and the unknown risks with IP and confidential information being put into some of the models. Many employees are already using AI on their own, but at the organizational level, providing a simple statement of what is or is not allowed, with examples, could help small businesses create a framework. 

There may be some fear from the lack of understanding. At the high-profile level, our sector is more familiar with the labour side. The media and the general public’s perception plays a part in the hesitance. For example, there was an announcement from the Government of Canada about increasing public sector use of AI. They highlighted the food safety sector as an example of the trust they wanted to build. That's our sector - we're dealing with both food safety and AI tools. 

We have to make sure we're not compromising one tech by using the other. Food and beverage manufacturers are often slower to adopt than other sectors, because Health Canada has to get involved, or CFI has to look at the food safety implications of switching anything. Some built-in hurdles in the system also affect the mentality of getting started with a new tool. In some locations, the lack of broadband connectivity limits engagement with new and powerful tools, for example, the location might be a cell phone dead zone in Eastern Ontario. 

Most conferences over the past 4 years have had positive stories to share about the impact of different technologies on the industry. Some manufacturers added technology that saved hours in calculations, reduced food waste, or streamlined measuring and identifying products to be shipped out -- these successes save millions. 

Courses, like the one sent out with this webinar, are how people can start interacting with AI. It'll help build base-level understanding, so they can investigate further. An introduction to prompt engineering and how it helps spark imagination and how to maximize productivity. 

On the other side, it's important to ensure the skills are there for people to handle it. We have a massive learning framework that we try to track skills. Cybersecurity was a newer addition to that list. There's a gamut of ways people can engage with it once they gain confidence and find the benefits and the risks they need to know. 

This kind of question came up recently when I was speaking to a ministerial group of people in government about teaching Excel skills or teaching Gen AI; those 2 aren’t separate anymore. You're starting to see Gen AI applications in any kind of data analysis tool, using it to create presentations or some amazing tools that are out there. It's giving you a headstart with the creative process to save time and you still add to it.

We're seeing a verticalization of applications for different areas and tools in certain industries and sectors that are built on top of those large language models. You're constraining it through retrieval augmented generation to pull from the box of information data. It can give answers because you constrained it. It's using its ability from being trained on the large language model. What typically happens is that it’s in those platforms, for example, where it's trained on the data, the information, the white papers, and the relevant body of knowledge, that food processors may want to access the legislation.

If you're a knowledge worker, a large organization, or even for yourself, there are a lot of low-risk activities to test Gen AI. Start getting familiar with it because that's how you learn about it - “Can it help me write this email faster?”, “What can I save time on, especially in low-risk activities?”. I recommend Perplexity as a start to using GenAI on a free LLM.

Driverless vehicles are already on the roads in North American cities and there's driverless freight coming. However, rural areas may not be as quick to adopt it because of connectivity issues related to the internet, etc. Some policies relating to packages signed off by drivers would have to be updated. Some areas in the US already have driverless taxis. They’re so used to it that they find cabs with drivers unfamiliar now. There are a lot of advances happening in spaces we don’t know about, especially with the exponential rate of change that AI grows. 

With the speed of change and the uncertainty around misinformation and disinformation, it’s important to stay updated on the disruption and the negative side of AI’s influence on our organizations and careers. It's going to be hard for governments to figure out how to put regulations and other things in place. But, It's going to be a very interesting time with the rate things are changing.