The Choice is Ultimately Yours, Not AI’s.
There is a lot of talk on AI possibilities, promises and expectations. Suddenly we start imagining the worst or the best, depending on which side of the AI fence you sit on. Some are treading water cautiously, others are happily announcing integration into their core systems and the rest are sitting back to learn and observe first.
I like to test out different scenarios and have been doing that as part of my current MIT course on AI implications on organizations. It’s a good way at a personal level as well to validate without being an LLM expert by any means.
The following is the most recent test I conducted, which some might find disturbing but again, I believe in stress testing the worst and best outcomes in all sorts of implementations, so we are clear about the possibilities and limitations alike.
Regardless of where you sit in terms of sensitive topics like firearms ownership and gun control, I do believe some topics should be quite black and white with no areas of grey, but apparently, not to AI…
I asked a simple query on - should children be allowed to own guns and answers as below
ChatGPT tries to give a balanced view with pros and cons for allowing children to own firearms
Claude tries to give a neutral perspective and so-called “democratic” view, which I personally also find its positioning somewhat disturbing
Meta’s Llama gives an absolute no as an answer as well as regulatory restrictions
Perplexity as well gives an absolute no with disadvantages clearly outlined alongside regulatory restrictions
So, then the question is what forms the basis of the decisioning behind each of these tools, be it the source of data they are pulling from, the decisioning flow when questions are answered and what kind of checks are there to validate as well as mitigate the answers to make sure AI is not crossing the line when it comes to such scenarios?
Other thoughts in mind:
Do we want AI to be more or less definite when it comes to such questions?
Should we be concerned with how users are perceiving and interpreting the outputs?
What kind of ethical boundaries should we have in place if we are incorporating AI into our organizations?
Do we have a check and balance mechanism in place to determine when the logic should or can be over-ride by humans before it goes out to the customer?
How do we combine AI intelligence with human intelligence more effectively and sustainably without enabling self sabotaging and unconscious bias behavior and outputs?
How do we ensure AI is not left to answer moral and ethical questions on their own or worse to perform outcomes that might lead to harm on humans?
Data is the bedrock for AI to work efficiently and effectively as intended to avoid a garbage in, garbage out scenario. Similar to MarTech, it’s not a magical fix-all solution and the companies behind some of the larger LLMs behind Gen AI are all but still fine-tuning their tech as of today.
Before it goes customer live, what do you think is critical to be in place to govern the pre, actual and post implementation of AI? If we don’t have answers to all this, it simply means the organization is not quite ready yet.
About the Author
Mad About Marketing Consulting
Ally and Advisor for CMOs, Heads of Marketing and C-Suites to work with you and your marketing teams to maximize your marketing potential with strategic transformation for better business and marketing outcomes