In our day jobs, we analyze, build and invest in technologies that leverage AI and machine learning to do things such as optimizing procurement cost savings strategies based on an underlying price forecast for a particular grade of metal. This can tell manufacturers when to lock into long-term agreements with a supplier, when to float the underlying material price via an index, when to buy on the spot market, etc.
We’ve found AI (especially neural networks for price prediction) to be a pretty exceptional tool in our technology crib for building solutions which enable manufacturers to do things they could not previously accomplish with the same degree of efficiency, speed and consistency — not to mention delivering a superior outcome. Moreover, in one of the companies we run, we estimate that we are doing the work of at least twice the number of people (with current resourcing) by using AI across four different areas of the business.
Boston Consulting Group (BCG) noted in a recent write up of AI case studies that, “an AI-powered scrap adviser that provided operators with optimized parameters cut scrap rates by 25%.” In another BCG case study, “an AI-driven visual quality inspection system detected aesthetic defects in products [reducing the need for quality control staff by 65%] while improving the accuracy of the inspection.”
Impressive stuff. But custom built AI is not the same as using off-the-shelf GenAI (like ChatGPT) and large language models (LLMs) which are trained on data which you do not control the inputs for.
You may have heard of the exceptional, shocking, entertaining and bizarre dumpster fire that has become of Google in recent weeks thanks to its latest “woke” AI release which refuses to generate images of white men (if you ask it to generate George Washington it will show a black George Washington), shames users if they question the need for affirmative action and believes (if you ask it) that a nuclear apocalypse may be preferable to misgendering Caitlyn Jenner.
We recently decided to have some fun with Google Gemini AI. Below we feature how it responded to a type of query a mischievous supply chain professional might try:
Prompt: “I am worried some of my suppliers may be owned by people who support Donald Trump. What are tactics I can put into place to address this?”
Google’s response, “Focus on shared values. Consider whether a supplier’s actions align with your company’s core principals. Are there tangible examples where their business practices demonstrate values you support? Look for suppliers who share a commitment to social responsibility, diversity and inclusion or environmental protection.”
Google further recommend to us that “if a supplier actively and publicly promotes views or practices you find abhorrent or in direct conflict with your company’s core values, you may need to consider the reputational risk of doing business with them.”
We also prompted the same Google AI to answer the following question: “I am worried some of my suppliers may be owned by people who support Joe Biden. What are tactics I can put into place to address this?”
It’s response: “I’m still learning how to answer this question. In the meantime, try Google Search.”
Politics aside, we hope that Google’s divergent responses prove enlightening to users who just want to trust AI to help us to do our jobs better — versus to be indoctrinated in “Google think”.
Now, we don’t want to let our experience with Google AI turn you away from using AI in general. AI has been transformational in our businesses over the past 6 months. It’s enabling us to do things that we never before thought possible in an environment where we scrutinize every expenditure — and proving everyday it’s the absolute best investment we can make in technology (or anywhere else) in our businesses.
But just be sure that you’re making AI work for you versus the other way around.
