I understand that this overview of Microsoft’s use of AI is intended to allay concerns from government agencies, which are a significant source of revenue for Microsoft, but amidst the practical explanation of how and why AI can be valuable to government agencies, and indeed any business, there are some interesting points to note about what AI is and its limitations.
The general logic here is sound – AI is not the only solution, and human oversight is still required to truly maximize the value of these tools. This is something that anyone who has used an AI system knows, but it’s worth emphasizing amid all the hoopla and fear that AI will transform work as we know it.
But what I find most interesting are statements like “AI is just math” – and you're right – current AI tools aren't really what you'd define as artificial “intelligence” at all, as there is no internal “thinking” process at work – it's simply machine calculations that take this data point and that data point and push out the most likely outcome.
They're much better at this task than previous systems, and the output is more human-like. But these systems don't “think” and don't take into account the depth of the question to give a nuanced response. As Microsoft points out, it's just math, and that math is relative to the data inputs a particular system has access to.
This is contextually important to understanding how these tools work, as are notes on “grounding” and training AI models for specific applications.
In essence, AI tools are valuable, but the hype surrounding them can lead to overestimating the impact they will have on your work.
This is one element of this overview, but other notes are noteworthy from both a government and general business perspective.
