The intelligent way

ChatGPT, or ‘Chat Generative Pre-Trained Transformer,’ is a type of natural language processing model that is free to use. Image: Shutterstock

Generative AI offers the can sector significant productivity improvements, but care is needed in its roll-out. Laura Syrett reports from London

 

The benefits of integrating artificial intelligence (AI) within the canned products industry are well accepted, from the deployment of robotic arms to move products on and off canning lines and pack cases, to AI solutions designed for predictive maintenance in canning factories.

Given the canning industry’s ready embrace of AI, it is little surprise that the canning sector is already investigating options to adopt the latest generation of major disruptive AI tools, generative AI, based on large language model (LLM) processing.

The most ubiquitous of these tools to date is ChatGPT,  or  ‘Chat  Generative  Pre-Trained Transformer,’ which is a type of natural language processing model that is free to use.

Developed and launched in November 2022 by US-based artificial intelligence research laboratory, OpenAI, ChatGPT uses so-called ‘self-attention mechanisms’ – a system that takes in information from a source, such as the internet, to produce information in the form of text.

Other large language generative AI models competing for users include Google’s Bard tool and Microsoft’s AI-powered Bing AI, while online retailer Amazon has also launched its own unbranded ‘New Language Model’ AI, intended to underpin its Alexa virtual assistant. Amazon’s new language model is a multilingual large- scale model, pre-trained on a set of denoising (ie removing irrelevant or random information from instructions) and Causal Language Modelling (CLM) tasks.

Regardless, users of generative AI tools get them to perform tasks by giving them prompts –natural language instructions – to guide the AI model’s output and influence its tone, style and quality, and can be used to generate a huge array of text- and image-based results.

There has already been significant discussion about the potential of ChatGPT to revolutionise supply chains, including repetitive manufacturing processes, like those found in can making and filling factories.

Martin Cloake, an Ottawa, Canada-based manufacturing expert who is both co-founder of the Ottawa-based Industry 4.0 Club, which aims to deliver continuous improvement in manufacturing, and CEO of Raven, a company that delivers AI solutions to the food and beverage manufacturing industry, said that generative AI tools like ChatGPT are set to revolutionise manufacturing.

Midjourney creates images from natural language descriptions. Image: Shutterstock

“If you were to go into manufacturing plants today, everybody’s playing with [ChatGPT]; if you’re not playing with it, or at least trying to understand it, you’re in a bit of a risky position,” he told CanTech International.

Cloake thinks that one of the most exciting practical applications of ChatGPT in factory environments is the ability it gives businesses to communicate and share information with staff at all levels by creating bespoke instructions, potentially doing away with the need for long user manuals for operating machines and processes.

Mike Ungar, another co-founder of the Industry 4.0 Club, said that generative AI can help reduce the skills gaps created when experienced staff leave a business: “The impact [of tools like ChatGPT] can be tremendous for companies that lose skills and knowledge when people retire,” he said.

Others, however, are more sceptical  of ChatGPT’s readiness for practical deployment in the supply chains of specific industries like canning operations.

“In my opinion, the current iterations of ChatGPT [are] still far from being considered fit for purpose for commercial use, certainly not within a supply chain setting,” said Dr George Bargiannis, subject area leader in from the School of Computing and Engineering at the UK’s University of Huddersfield. Dr Bargiannis notes that the open-source data ChatGPT is trained on, at present, is unlikely to be of direct relevance to the day-to-day operations of a particular supply chain, unless the data on how to perform that specific task – such as canning tomatoes – is publicly available.

Maintenance tasks are a more likely use case, according to Bargiannis, again provided the canning technology is standardised and its operating and repair manuals are in the public domain.

“If it is intended to be used for tasks as specialised as predictive maintenance within a particular manufacturing process… ChatGPT trained on years’ worth of historical records of maintenance tasks [for a specific manufacturing process] will potentially be able to assist in monitoring – offline or online – an assembly line to detect and predict equipment faults,” added Bargiannis.

Beck’s Autonomous is a beer that ‘created itself,’ using Chat GPT and Midjourney. Image: AB InBev

New product concepts

Some major brands that produce canned products

have already started using ChatGPT for marketing activities and even the creation of new product ideas. US multinational canned and bottled beverage producer, The Coca-Cola Company, announced in February that it was engaging with a ‘services alliance’ between US consulting firm, Bain & Company, and ChatGPT’s creator, OpenAI, to use ChatGPT to create personalised advertising copy, images and consumer messages.

In March, Belgian multinational beverage company, Anheuser-Busch InBev, commonly known as AB InBev, the brewer behind major beer brands Budweiser, Corona, Stella Artois and Beck’s, debuted what it claimed to be one of the ‘first completely machine-created beers in the world.’ Named Beck’s Autonomous (subtitled ‘The Beer That Made Itself’), the product was created because of AB InBev tasking ChatGPT and Midjourney, a generative AI programme created and hosted by San Francisco-based independent research lab, Midjourney Inc, to design a beer, branding and marketing to celebrate the Beck’s beer brand’s 150th anniversary.

In terms of the type of food and drink that goes into cans, generative AI could offer new recipe ideas, along with cooking times that will optimise the food or beverage for canned storage, ensuring the best quality end product and potentially improving shelf life, according to Matthieu Vincent, co-founder of the Paris-based DigitalFoodLab, a leading food-tech consultancy in Europe.

“My bet is that the impact of AI in food will be massive but quite slow to materialise, as it requires adaptation and often coordinated change all along the supply chain. However, established players have considerable opportunities to differentiate themselves right now,” he said.

Generative AI can also help design product packaging, potentially opening the door to new designs for cans, or more appealing images on the outside of cans in the form of AI-generated artwork.

Canned food and beverage products generally display photographs of their contents on the outside, often in prepared form as a ‘serving suggestion,’ to entice consumers to purchase a product they cannot physically see inside the can. Midjourney, which creates images from natural language descriptions, is rapidly being adopted for food packaging design, according to Charles Wright, director of data innovation at PreScouter, a Chicago-based research intelligence company. “[We’ve heard from] graphic designers who take inspiration from images creating Midjourney, reporting significantly faster times to create the final product because of the ability to rapidly iterate over designs,” he said.

Teri Campbell, owner and creative lead at Cincinnati, US-based Teri Studios, which provides photography services for  consumer-focused packaged goods companies, restaurants, beverage, and  food-related  businesses, including Kraft Heinz and Coca-Cola, said he uses Midjourney to increase his output and reduce costs.

Campbell said generative AI can be used to create backdrops for products, such as kitchens, saving the time and expense of building sets. Yet despite these cost-saving advantages and enhanced visuals that could be passed on to food packaging businesses, Campbell said some of his clients are hesitant about using AI-generated imagery.

“It’s really early stages for clients to buy-in and go with it, but I’m using it in a lot of spec work, where we’re trying to showcase what we can do,” he said.

Workforce implications

According to the Industry 4.0 Club’s Ungar, the advantages generative AI could offer employers in terms of substituting experience with large language model information sharing, will be at the expense of some roles in manufacturing operations.

“At the moment, we use AI for predictive maintenance, to tell us when a part is about to fail. If ChatGPT can tell me how I replace that part, [and] what I need to look out for, I need a less-trained, less-experienced person to do this work,” he said.

Cloake adds that ChatGPT “has  the potential to get rid of a whole load of mundane work,” which will make some roles obsolete, but will theoretically create opportunities for higher- skilled, more fulfilling jobs.

“It will create risk to the workforce for those doing repetitive work. Upskilling and continuous learning is what people want [from their jobs now] so if we can find a way for ChatGPT to give them that, we’re winning. Winning in manufacturing means continuous improvement,” he said.

This feature article is restricted to logged-in paid subscribers.

Login or subscribe now to view this exclusive content.

Related content

Leave a reply

CanTech International