You will all remember Hansel and Gretel. Abandoned by their stepmother, they left a trail of breadcrumbs to find their way out of the forest. Birds ate the crumbs, and Hansel and Gretel got lost, and eventually fell into the clutches of an evil, cannibalistic witch. Imagine, with me if you will, that AI was around at the time. Had AI trained on those breadcrumbs before they were eaten, it could have produced a map to lead Hansel and Gretel to safety.
Space does not allow me to consider the completely legitimate question of why these stories always feature an evil stepmother, and never an evil stepfather. Instead, I propose to use my LinkedIn character allotment to ask another burning question, and that is: “What would have happened if instead of dropping breadcrumbers in a linear fashion, Hansel and Gretel had thrown them into the air to be carried by the winds to all corners of the forest? Of course, the birds would still have eaten them. But could AI have still saved their wretched lives?
The answer, my friend, is that the breadcrumbs blowing in the wind would not have been clean data for AI to map. Hansel and Gretel would still be witch food, just like they were in the original tale.
The tech people tell me that AI works on pattern recognition. No recognizable pattern of data means no pattern recognition, which means no decent results.
So, what does all this mean for our law offices? I suggest that it means that we must clean up our data so that AI can do its magic.
Let me conjure up an example for you. Suppose that I want AI to draft a shareholder’s agreement for the wicked witch’s bread and breakfast resort, so I turn it loose to train on my database of precedents. In my database, there are masterpieces drafted by Murray Magnificent and Marvellous Maureen, as well as some bumf from Boris and some drivel from Donald. The output generated by AI is not going to be nearly as good as it would be if it trained only on Murray’s and Maureen’s masterpieces.
I am sure that you can come up with more examples on your own. Use AI for marketing – will AI send material to extinct addresses or departed employees? Have AI generate retainer letters – just how many precedent variations are being used in the firm? Instruct it to generate accounts – how many formats are being used and is a single format being used consistently for each client?
Back around 1969, someone (I asked AI, but it could not tell me exactly who) is reported to have said, “to err is human… to really fowl things up you need a computer.” Whatever disaster prompted that remark will undoubtedly pale in comparison to the fire hose full of screw ups that will spew forward and drown those who turn AI loose without first getting their cottage in order.
Did you need a project? Now you have one! Get your house in order before you bring in AI.
Read what Appara has to say about technology, trust, and legal process here: https://bit.ly/4ukujnA