Where should your AI efforts begin? | CQI | IRCA Skip to main content

Where should your AI efforts begin?

Progress indicator

Published: 16 Oct 2024

Technology consultant Dr Jeffrey Funk considers the rise of generative artificial intelligence and how it can benefit the quality profession.

We are currently in the third ‘wave’ of artificial intelligence (AI) advancement, the first having taken place in the 1950s and the second in the 1980s. The third wave began with neural network AI, which built upon ‘big data’ and algorithms widely used by Google, Meta and Amazon during the 2010s, and included search engines, social media, and the placement of ads and news in search results and social feeds.

In the mid-2010s, consulting companies began making more explicit forecasts for AI, some expecting the industry to be worth $16tn by 2030 because of its impact on the virtual world and on physical products and services. Bestselling books predicted the end of work, and business professors argued that strategy had already become all about AI.

Overall, however, private investment in AI has fallen from US$130bn in 2021 to US$96bn in 2023, despite an eightfold rise in investment in cloud computing infrastructure for generative.

What should we learn from this? AI is hard to implement; it is particularly hard to replace workers and, so, augmentation is still the better way to go.

Innovation on farms emphasised augmentation of human capabilities for centuries, with new tools used in combination with horses and other animals, and later with tractors. This accelerated in the mid-1850s, and it took 90 years for farming jobs to drop from 90% of all jobs in 1850 to 20% in 1940. It only took 55 years for manufacturing jobs to fall from 26% of all jobs in 1960 to 10% in 2015, but mostly because government policy encouraged foreign imports.

We must find niches that really value AI, particularly generative AI, beginning with those that are the most economically favourable.

Early niches should not require high accuracies because hallucinations, in which generative AI products make stuff up, are common. This is where quality professionals can shine. They understand statistical process control and the challenges of obtaining high yields and accuracies.

"We must find niches that really value AI, particularly generative AI, beginning with those that are the most economically favourable."

Dr Jeffrey Funk, technology consultant

One way to bypass these requirements is through hidden workers, another form of augmentation. Rodney Brooks, co-founder of iRobot, co-founder and chief technology officer of RobustAI, and former director of computer and AI labs at Massachusetts Institute of Technology, argues that almost every successful deployment of AI has had one of two expedients: a person somewhere in the loop; or a very low cost of failure, should the system blunder.

Applications in search, advertising, social media, cybersecurity, and e-commerce are common, but the expansion to physical applications has been slow, except in logistics, such as warehouses. Even some virtual applications are experiencing slow diffusion.

For generative AI, the first successful cases are likely to be in video advertisements and short TikTok and YouTube videos. It will take time, however, for this success to move from entertainment videos to educational ones, because of problems with hallucinations.

The bigger sector is Hollywood’s creation of TV dramas and movies (and videogames), and partial success is likely because it is one of the few industries in which the output is images and words. Bigger impacts will come when AI can help fit together multiple scenes that are recorded separately but then require fillers, which can be AI-generated.

Generative text will have a smaller impact, despite being widely used by students and others for generating mediocre papers, reports and email. They are good at generating text, but often do not perform the prescribed objective. As Andrew Orlowski wrote in Britain’s Daily Telegraph: “ChatGPT is a lazy bureaucrat’s dream – but a nightmare for the rest of us.”

Chatbots, order-taking in fast-food restaurants, and call centres will probably see the biggest application of generative text in the next five years, but there are still too many hallucinations, and the benefits are not that large.

One analysis of call centres in the Philippines found that large language models only increased productivity by 14% on average and the improvements were smaller for the best performers. For order taking, McDonald’s ended its work with IBM, while, reportedly, Wendy’s and White Castle found that mistakes occurred on 14% and 10% of the orders respectively.

Software development and maintenance are likely to be the second largest beneficiary of generative AI, but only after the processes are documented and guard rails are implemented, to prevent individuals from gaming the system.

Identifying processes and how AI can enable improvements is another area in which quality professionals can shine. They know it is a big mistake to extrapolate from superficial improvements in support tasks to improvements in processes at the corporate level, a huge leap of faith when we know that many tasks are not needed. For the ones that are needed, it is difficult to characterise a high-quality output – the first step in using AI.

Accuracy is often the most important variable when you are trying to satisfy a customer, and until the frequency of hallucinations decreases rapidly, quality professionals should use AI products that are currently available.

A recent article in the Wall Street Journal noted that companies are increasingly favouring smaller and mid-size generative AI models over the large, flashy models that made waves in the AI boom’s early days. The models are trained on less data and often designed for specific tasks, making them cheaper and potentially more accurate.

Andrew Ng, co-founder and head of Google Brain and now an entrepreneur, has made this argument for years for AI in general, and perhaps it will be the way forward for generative AI, too.

About the author: Jeffrey Funk is a retired professor. He received the NTT DoCoMo Mobile Science Award for his work on mobile communication and has published five books. His sixth book, Unicorns, Hype, and Bubbles: A Guide to Spotting, Avoiding and Exploiting Investment Bubbles in Tech, will be published by Harriman House in October 2024.

Join our Digital Transformation special interest group

The newly formed Digital Transformation SIG aims to advance the understanding and development of practices of Quality Management in the digital age.

Quality World

Get the latest news, interviews and features on quality in our industry leading magazine.

Access the CQI Quality Learning Hub

CQI Quality Learning Hub .png

Supporting your professional development

Join the SIG

Develop your knowledge and represent your industry.