Artificial intelligence is still experimental, with a need to explore questions about safeguards, testing, infrastructure, and project priorities.

As I see it, artificial intelligence (AI) is a serious plague that presents a host of risks and challenges largely ignored by an industry hungry for low-cost labor, higher profits, and happier shareholders. I believe safeguards, power consumption, and a lack of thorough testing are real-time problems for something that is still largely experimental.

Now, this isn't going to be one of those sci-fi-inspired paranoid rants about AI taking over, but it is a word of caution in relying too heavily on the technology as some sort of savior or panacea for solving the world's problems. The one thing few seem to understand fully is that AI is experimental, although dozens of IT companies have fully embraced AI rather than doing extensive testing. In my opinion, not enough time and research have been invested on implementing guardrails or safety triggers. Companies have gone "all in" and integrated AI into their flagship products without thinking of worst case scenarios. Again, this is my observation from the timelines of products released pre-AI and those released post-AI.

I personally doubt that a few months' worth of testing is adequate to test a list of obvious what-if scenarios, much less those that are a bit less probable, but still possible. The last thing anyone in a business environment needs is some runaway AI bot going through its algorithms and deciding that years of data need to be purged or that those same years of data need to be encrypted with no human intervention or knowledge of the decryption code.

I have a perhaps unhealthy level of paranoia about everything, but I'm especially paranoid about the unmitigated inclusion of AI into everyday objects, my trusted software applications, and my vehicle. I prefer to reserve a significant level of control for myself. Now, don't assume that I'm anti-AI. I'm not. In fact, I was an early adopter of ChatGPT and other AI applications, and I earned a few AI "badges" from my employer through training and testing. If you recall one of my earlier Welcome columns, I have always been interested in AI to the point of learning AI programming languages back in the dark ages of IT.

I also think AI poses a risk to the power infrastructure, in that it consumes far more power than is received as a benefit from its consumption. Some technologies are under investigation to remedy the power issue, such as small modular reactors (SMRs; i.e., small nuclear reactors) that would power data centers without tapping the public grid. Solar power is another good option, as well as building data centers underground for a cooler and more stable environment that doesn't require so much energy to maintain a healthy operating temperature. To offset some of the power consumption, they could implement priorities to AI-related tasks. For example, generating memes and graphics would be classified as very low priority and tasks such as designing new cancer-fighting drugs would take the highest priority.

Like AI, my attitude and opinions of AI-enabled products and services is ever-evolving. On any day, I can have opinions that range from acceptance to rejection. Each of my opinions of an AI application is formed case-by-case. I enjoy using AI-enabled apps for certain things, but I don't want or need it to drive my car, write my articles, or take my job. I can live with AI, but I know I can also live without it and have for many years. AI should be an enhancement, an add-on, or an option, but not a default or required element of any application, tool, vehicle, or appliance. I need the choice to opt in and opt out for each.

We must protect ourselves from becoming too comfortable with complacency, automation, and accepting the default – whatever that might be. We can't accept cheaper or lazier as the new normal, nor buy into the promise that less is more.

Ken Hess * Senior ADMIN Magazine Editor