AI - to Trust or Not to Trust

I recently attended a leadership summit with many professionals in the manufacturing world. Trust me when I say AI (Artificial Intelligence) was a hot topic. Three questions repeatedly came up during discussions:

• Will AI replace my job?
• How can we use AI in our day to day?
• and naturally, What is AI?

No, I do not believe AI will replace your job. You will adapt to the new jobs AI inherently creates including building datacenter warehouses, managing these giant datacenter warehouses, designing power grids to support AI, managing the power grids, governing the use of AI, and so on.

AI has many use cases. Think of all the annoying, repetitive, human-error prone tasks we do as part of our jobs. AI can perform these mundane tasks way quicker than you can, and AI won?t complain!

AI in plain terms is just a super quick way for computers to search all the interconnected bits of information all over the Internet, process all that information (using math and statistics), turn it into language we can understand and continuously train. Techies will refer to AI as a sophisticated Large Language Model - LLM, because we all needed another acronym.

What are the rules of engagement?

I noticed, though, few questions came up surrounding trust, governance, and privacy. Being who I am, the lack alarmed me. Here are my questions to consider before we accept AI into our businesses:

Can we trust the supporting infrastructure?

AI takes massive amounts of natural resources and depends on our critical infrastructure - think land, water, precious metals, labor, and energy. Can AI datacenters stay powered on all the time? What about our hotter climate? Will we have the power to cool our AI datacenters? Do we have enough gold, tin, copper, and cobalt?

Can we trust the AI training models?

At the core of AI is what information we feed it. What happens if we feed our model ?junk food? and our AI model becomes unhealthy? How will we undo these bad habits and behaviors?

Can we trust the AI model with our private information?

Is our information even private? How will we reclaim our sensitive protected information after a data leak into the model? Where are the boundaries and guardrails? Who owns the information? Who is accountable?

Can we trust ourselves to be good custodians and not bad actors?

Will AI recognize malicious intent and sabotage? Much like social media platforms have evolved into recipients of non-human bot posts and targeted advertising, will AI evolve into the same? Who will ultimately be in control? Who is responsible for the governance ? AI itself?

We need good governance

With good governance, clear rules, privacy, model-scope and an understanding of our risks, we can trust AI to improve the quality of our jobs and lives. If we do not bake governance into every phase of the AI model development, then we are doomed.