Thursday September 14, 2023; 12:14 PM EDT
- Here's a sentence recently written by ChatGPT:#
- With the meteoric rise of VisiCalc, the personal computer's fate was redefined.#
- 1. They say ChatGPT uses a massive database to decide which words are mostly likely in a particular context. This sentence is probably a good example of that. In commonplace sorts of writing with a flair for the cliché, the word rise may very well be preceded by the word meteoric. Or not. The decision would depend on what a person wants to say. If VisiCalc is important because of its rise, then by all means focus on that and make rise the main noun of the sentence. If VisiCalc is important because of the speed of its rise, then amplify the rise by saying something like meteoric rise, or use a fresher synonym for that idea. But if VisiCalc is important because it changed people's minds about the machines, focus on that and let the meteor shower recede from view. If, as they say, a predictive technology is doing the writing, that's fine, but somebody needs to decide what meaning is important for the particular piece of writing. Maybe someday, for non-tech pieces of writing such as this one, a machine will be able to make the right choice, for the occasion and audience at hand, between these different-meaning versions:#
- With the meteoric rise of VisiCalc . . .#
- With the rise of VisiCalc . . .#
- VisiCalc . . .#
- 2. Similarly, a predictive writing technology could predict a roll for fate in that sentence, but the same tool, or a human interloper on the project, might predict otherwise. Are we focusing on changing a machine's fate or changing how the machine's understood by the society? We include fate or remove it, depending on the answer. The meanings are different. If we're not focusing on fate, then we're focusing on redefinition, and we change the verb accordingly:#
- . . . the personal computer's fate was redefined.#
- . . . the personal computer was redefined.#
- Ordinarily, though, if a predictive tool has read enough writing manuals it realizes the power of active voice verbs in English-language prose, using not "was redefined" but "redefined" instead. That gives us the following version:#
- VisiCalc redefined the personal computer.#
- That's 5 tightly focused words instead of 12 loose, fuzzy-headed ones. I say fuzzy-headed because "meteoric rise" distracts from what is probably the best insight in the sentence, the fact that VisiCalc redefined this desktop tool. And "fate" also distracts from the same insight. Once a predictive tool like ChatGPT or a human being sees that similar wordings focus on and mean different things, we're set up for thinking about what we really mean. Can a predictive tool answer the question, "What do I really mean to say here?" Based on the sample sentence above, I'd say, on non-tech tasks, No, not yet.#
- 3. When a person hasn't done the work yet to think through the options in a field of possible words and meanings, hasn't done the work to pin down a precise answer to a question at hand, we might say that "So-and-so doesn't know their own mind."#
- So I conclude that for now ChatGPT knows its own database but doesn't know what it thinks about its database. ChatGPT doesn't know its own mind.#
- Which means that ChatGPT may, for the time being, on non-technical tasks, have no choice but to offer fuzzy-headed writing.#