George FreemanThought Leader | Author | Architect| Florida, United States
Although I greatly desire to post a non-AI topic, I find myself compelled to offer up contrarian thought on this subject as the overarching narratives surrounding AI lack (in my opinion) veracity, not necessarily due to merit but because lifecycle impacts and considerations lack appropriate challenge-based-review in the public square.
Note: I’m not against technological advancements in the field of study we call artificial intelligence, but I am opposed to the sensationalized, obfuscated self-aggrandizement that proliferates the industry.
I say this as one whose software design portfolio includes many products over the decades that can reference AI through lenient subfield technobabble. So, although I am a project professional, my background in highly abstracted frameworks, rule-driven metadata, fuzzy logic, adaptive object models, NLP, and the like provides me with a degree of insight beyond project management.
With that intro out of the way, the topic is “Transparency, Facts, Truth, Trust, and Accountability » Reevancy to AI.” To that end, I have noted in previous posts a progression of linked points that I believe are common sense and logical. They are:
[1] - Transparency » Leads to facts (and enables accountability). [2] - Facts » Lead to the discovery of truth. [3] - Truth » Authorizes trust.
QUESTION: Do you believe this progression of points is correct (for life in general), and if you do, how does that impact your thoughts and views toward knowledge provided by AI-based frameworks? Recognizing that transparency, the first and most pivotal point in the above progression, is an unlikely deliverable from generative AI-based service providers. Saving Changes...
I'd support the progression you provided - the greater the level of transparency, the higher the level of trust. I usually use the analogy of an open kitchen restaurant - as a diner, you are likely to feel more comfortable and trust that what is being plated hasn't fallen on the floor!
While AI tools are just the latest example, there are many non-AI tools which similarly obfuscate what is being done to provide us with results. A simple example is the MS Excel pivot table capability. Unless one is willing to manually verify how a pivot table has been generated and whether its results are accurate or not, we are taking a leap of faith that Microsoft has worked out all the glitches and thought through all the plausible scenarios for its usage.
Where we have the biggest challenge is where a decision process that is currently quite transparent becomes opaque as a result of using AI tools. In most cases, these substitutions are happening in situations where the original state was not transparent to begin with so we are merely replacing one form of obscurity with another.
Kiron
...
1 reply by George Freeman
Feb 05, 2024 7:17 PM
George Freeman
...
A common issue in enterprises occurs when an executive in one division asks a “burning business question” from their enterprise data hub, and another executive in a different division makes the same interrogation. Guess what? They get different answers and start the divisional blame game, feeling assured of their distinct results.
So, let’s forecast this out: [1] - After the president jumps in to diffuse the feud, the two executives blame IT (who would have guessed) for their “faux pas” and convince the president that they need to bypass IT and leverage a bleeding-edge AI-based business analytics service to solve the problem.
[2] - So, they engage the service with board-level accolades given to these two executives. But on queue, a new “burning business question” presents itself, and our two executives engage their “natural language prompts” to answer this new and highly complex question—and the divisional blame game is afoot again.
The moral of the story: Context is the sovereign ruler of all things—business. If you are off by one degree, you can have material impacts, hence the need for “structure.”
We should ask ourselves, are we heading down a “structure path” with AI, wherein we can grant “trust” to its outputs? Or is AI taking us down an unstructured path where “accountability” is a narrative and little more?
So, I agree with your summary, we are replacing one form of obscurity with another!
Saving Changes...
George FreemanThought Leader | Author | Architect| Florida, United States
Feb 05, 2024 7:25 AM
Replying to Kiron Bondale
...
George -
I'd support the progression you provided - the greater the level of transparency, the higher the level of trust. I usually use the analogy of an open kitchen restaurant - as a diner, you are likely to feel more comfortable and trust that what is being plated hasn't fallen on the floor!
While AI tools are just the latest example, there are many non-AI tools which similarly obfuscate what is being done to provide us with results. A simple example is the MS Excel pivot table capability. Unless one is willing to manually verify how a pivot table has been generated and whether its results are accurate or not, we are taking a leap of faith that Microsoft has worked out all the glitches and thought through all the plausible scenarios for its usage.
Where we have the biggest challenge is where a decision process that is currently quite transparent becomes opaque as a result of using AI tools. In most cases, these substitutions are happening in situations where the original state was not transparent to begin with so we are merely replacing one form of obscurity with another.
Kiron
A common issue in enterprises occurs when an executive in one division asks a “burning business question” from their enterprise data hub, and another executive in a different division makes the same interrogation. Guess what? They get different answers and start the divisional blame game, feeling assured of their distinct results.
So, let’s forecast this out: [1] - After the president jumps in to diffuse the feud, the two executives blame IT (who would have guessed) for their “faux pas” and convince the president that they need to bypass IT and leverage a bleeding-edge AI-based business analytics service to solve the problem.
[2] - So, they engage the service with board-level accolades given to these two executives. But on queue, a new “burning business question” presents itself, and our two executives engage their “natural language prompts” to answer this new and highly complex question—and the divisional blame game is afoot again.
The moral of the story: Context is the sovereign ruler of all things—business. If you are off by one degree, you can have material impacts, hence the need for “structure.”
We should ask ourselves, are we heading down a “structure path” with AI, wherein we can grant “trust” to its outputs? Or is AI taking us down an unstructured path where “accountability” is a narrative and little more?
So, I agree with your summary, we are replacing one form of obscurity with another! Saving Changes...
Unfortunately, when decision makers choose to act without considering the pros & cons of their actions, there are likely to be unintended and unexpected consequences. The scenario you've portrayed is the latest example, but not too long ago a similar situation would have emerged with end-user computing initiatives which were launched without any guidance or governance from IT.
