AI Risks & Risks from AI

There are 2 broad streams of risk around AI. "AI risk" and "Risk from AI". So what's the difference?

AI risk is the inherent risk from AI itself. Thus far with IT, choice of technology, architecture, partner or startup, computing, data, automation in fact everything you know about IT is about repeatability, reliability and once tested will give you the same result every single time. Once tested and proven you could trust IT (More or less). If AI is an IT discipline, it now introduces variability and probability of a magnitude between 0 and 1. Especially with Deep Learning, GenAI etc. With the introduction if any form of AI, you introduce hallucination, blue ocean thinking vs red, green world vs red world debate, opinions and everything is right and correct - it only depends on your perspective.

All the above before you even get to "Risk from AI". What are you planning? Where is your data? Where is the rest of the data coming from? Who has it? Who owns it? What are vested interests in the data? Which industry? What application? How much are you going to rely on the output of your inputs depending on the AI? Will AI use discretion in the output given to you? Are the inputs in text, images, audio or video media? Are the outputs inn the same media as the inputs? If converted from one media to another the anecdote that comes first in mind is "Lost in Translation". Can you afford the loss in Translation? Can you ever quantify and qualitatively decide on anything that comes out with an AI layer on top of it?

One aspect of all the hype is the the future of Technology, IT and ICT (Information & Communications Technology) is AI. Any and all legacy software, ERP's, SCMs, CRM's and all the other 3 letter acronyms will shift toward AI. So you will be hearing a lot of the same hype and click-baits, however with minimal substance. 99 out 100 startups, especially in AI are going to fail. Today's news won't even exist a year later.

Now let's take, nature and natural selection. Nature over millions of years has shown biases. Arguable biases can include; survival of the fittest, winner takes all, the weak are sacrificed for the benefit of the group, flowers are weeds with a large advertising budget, a crying baby gets more milk for humans means the tallest trees, the brighter flowers, the louder, shriller mating calls all have "bias" implications. Won't all these biases have to be a part of the AI data-set? Now think up a list of all the risks each of these will introduce into anything you do with anything you claim to have AI on it.

My advice to you, don't put "AI" on any of your labels, till you really understand what "AI risk" and "Risk from AI" can do to your business, your customers, your vendors, your colleagues, your communities and your products. By the way; forget Governments and Regulators, they will be hard pressed to keep up with anything like the juggernaut that AI has released - they can't, won't and should not be saving you anyway. It's like stating that the long-term solution for people on earth is to move to another planet. Yes it will happen - but not in our life time.