If you’re my age or older, you most likely remember the first rudimentary spreadsheet application you used. You may not remember the actual moment you discovered the sort functionality, but you probably remember the feeling you got when you hit that button for the first time. In one magical moment, a jumbled list of items - randomly entered in the nonsensical order they popped into your brain - was immediately transformed into a perfectly alphabetized list worthy of your municipal library’s card catalog. Amazing! Revolutionary! Life changing?!
Today we live in the world of Generative AI. It is amazing, revolutionary, and probably life changing. This technology has already changed, and will continue to change, how we use and interact with computers in business, and in our personal lives. I use it daily. I use it in solutions for my clients. It takes tedious tasks of text transformation and turns them into simple push button experiences. Just like the sort function does.
Of course, GenAI does a lot more than order or sort text. It translates. It rewrites. It summarizes. It clarifies. It extracts. It interpolates. It expounds upon. It combines THIS and THAT into one thing. It re-imagines THIS, as if it were actually THAT. It recontextualizes THIS as if THAT didn’t exist. Its capabilities go well beyond simply sorting a list. In fact, it often gets simple sorting wrong, because it doesn’t use a hard-coded algorithm to produce it’s text transformation. Instead, GenAI uses vast troves of written example language to determine the probability of the next word it should write, and then the next, and the next... until it reaches it’s maximum output or it runs out of space.
The results of this seemingly simple exercise are impressive. It means that any writer with a rough idea of what they want to write, no longer needs to stare at a blank page (or screen), they simply ask a question or propose a concept and the technology generates a draft. Any reader who doesn’t want to read a 400 page transcript of a court proceeding, can get a summary that guides them directly to the “important” parts of the text. And any person who needs to transform any text from one form, or language, or perspective, to another can get a draft version of that transformation very easily and quickly.
However, there is a downside to Generative AI. It so easily manipulates and transforms text, that it often gives the appearance of being intelligent. This is an illusion. We tend to identify people who easily manipulate and transform text as intelligent people, which gives rise to the fallacy that a machine that does the same is an intelligent machine. We talk about GenAI “passing the bar exam”, as if that means it understands the law, and then we further extrapolate that we can use GenAI to replace lawyers. As tempting as that proposition might be to those of us who are not lawyers but work with them regularly, it’s not going to happen any time soon, if ever.
But, you might say, “I can ask GenAI to reason and it often gets the right answer. And its reasoning ability gets better with every new version. So eventually, it will be a fully reasoning machine intelligence.” Maybe. But probably not.
Any “intelligence” that GenAI exhibits is the intelligence that relies strictly on language processing. I’m definitely not a neuroscientist, but I am positive that the language processing that we humans do is not based on the probability of words as they appeared in historical texts we’ve read or heard in the past. And while I suspect a substantial part of our reasoning ability is directly tied to our language processing ability, I don’t believe that language in any way constitutes the entirety of our intelligence. Though it may, in fact, be the only way we can express our intelligence to others.
This leaves us with a machine “intelligence” that is very capable of communicating its ideas as confidently and convincingly as the most intelligent people in the world, but which may not actually have any underlying “intelligence” at all. Therein lies the actual danger of GenAI. I’m not worried about GenAI (or any other AI) becoming malevolent and taking over the world or wiping out humanity.I am actually worried about GenAI convincing people to do stupid things, some of which could potentially wipe out humanity.
A lesser, but more immediate threat may be the current marketing and hype environment. The companies who create GenAI, and especially those who buy the companies who create GenAI, are quite capable of convincing their customers and clients that GenAI needs to be at the center of everything. Technology companies who are not otherwise AI focused, should not restructure their entire tech stack or product offering to put GenAI at the center. Equally, law firms should not be looking to center their practice around GenAI technology. I predict that the first law firm to successfully restructure to take “full advantage of Generative AI”, will be the next firm to fail miserably.
Re-imagining the future of your company with the hottest and hypie-est technology at the center (when your business isn’t that technology) is the equivalent of restructuring the spreadsheet application around the sorting function. Obviously, that is ridiculous. And yet, the sort functionality has a place. In fact, it is now available in any and every application where sorting text might be useful. It’s not relegated to the spreadsheet, but companies aren’t built around it and firms don’t rethink their business models to accommodate it. It’s a feature. It’s a function. It’s a single component in a panoply of tools that can be combined to create great value.
GenAI is the same. It’s a component. One that should be included and used anywhere and everywhere that it makes sense, but it’s real value is as a single tool in the toolbox. A useful tool. An important tool. A “game changing” tool, maybe. But a tool that is best used for extremely limited purposes as part of a larger process to solve problems that are otherwise extremely complicated and time consuming, if not entirely impossible, to solve without AI. GenAI is not THE tool that will do everything you can possibly want, as long as you ask it nicely. But it is A tool that your vendors, developers, prompt engineers, knowledge managers, and innovation team can incorporate into a larger suite of applications to solve real business problems for your company or your practice. And if they do it right, you won’t know or care how it works, because you’ll just need to select the text you want to transform and hit a button to get the result you want. Just like clicking the sort button.
Want to print your doc? This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
