Why is AI not in your productivity statistics?
Because you lack imagination
As many people know, in 1987 Robert Solow quipped that “we see computers everywhere today but in the productivity statistics.” What few know is that the Nobel Laureate himself never was willing to use a personal computer. Even though he was still young in the 1980s and 1990s, he never used email. All of his correspondence was on paper, and he used a secretary to type his letters.
You may choose to disdain AI, just as Solow disdained the personal computer. But I strongly encourage you to do otherwise. The rest of us are using it more and more. I will give a couple of recent examples for me that might seem surprising.
The Job Search Counselor
I received a resume from a young man looking for a job, so that I could pass it along to someone I know who might have a position. I looked at the resume, and I was not willing to pass it along. I thought that it was vague and badly structured. I also intuited that there is a deeper problem, in that this young man is not really clear on what his career goals are and what next steps he should take.
I did not want to take on the project of coaching this fellow, so I gave the job to Claude. I wrote a two-sentence prompt to Claude asking for feedback on how to improve the resume and pasted the resume in. A split-second later, Claude came back with criticisms along the lines that I would make, as well as additional criticisms and details. The feedback took up more words than the resume itself.
My guess is the young man would benefit from a long, unstructured conversation with Claude about how to approach job search. I would bet that Claude would be a useful job search counselor.
The Social Worker
I have a friend who recently was given some devastating health news. As a result, he and his wife are confronted with many emotional, financial, and logistical problems.
In addition to receiving support from family and friends, this couple could benefit from meetings with a social worker. But I think it would be much easier, for many reasons, for the couple to just interact with a chatbot.
Again, I would recommend Claude. Claude would ask questions, so that they would discuss topics which the couple might otherwise not bring up. My guess is that Claude would have many helpful suggestions (as well as many clunkers), and this would prove to be valuable at this difficult time.
I can remember the late 1990s, when the Internet started to be really useful, but people did not know to use it. I think we are at the same stage with these chatbots.
Most of what you read consists of chatbots accomplishing specific tasks (or not), finding specific information (or not), and passing particular tests (or not). But many of us are finding them really handy for feedback in relatively less structured situations. They give us suggestions for questions to consider, ways to organize our thoughts, and ways that other people have approached such problems.
Christoph Schwaiger asks ChatGPT for insights based on his past usage of the chatbot.
To find out what blind spots ChatGPT thinks you have, this prompt is really useful:
"Based on the full history of our previous interactions, identify one insightful observation about me — something that might be consistent or telling, but that I may not be consciously aware of. In other words, what is one of my blind spots?
Pointer from Mark McNeilly
Lee Bressler thinks that in order to be used effectively, AI needs a different interface.
Picture a dashboard with buttons like “Brainstorm Ideas,” “Analyze Trends,” or “Draft Emails,” hinting at what it can do. Until we get that, AI’s stuck as a geek toy, not a mainstream tool.
The biggest barrier to using chatbots productively is lack of imagination. The only way to become adept at using them is to keep pushing your own envelope by trying to get help from them in novel contexts.
substacks referenced above: @
@
(1) I am not a software developer. Once upon a time I knew how to code a little, and with some effort, time, and stumbling through a few mistakes, I can make minor edits to some modern programs to adjust them the way I want. I can get by with MS Office "coding" in macros or visual basic, and my Excel-Fu is strong. At work I wanted to automate a task and just played with refining prompts on our newly-permitted, low-quality, and out-of-date chatbot to help me. We are not going to be allowed to use or train the out-in-the-wild cutting-edge systems, for reasons. Nevertheless, it was doing what I wanted -IN MINUTES- How can people not be labor-disruption alarmists after working with this stuff?
Now, do I tell my superiors? Are you crazy? Of course I don't. They hate this stuff, don't want to hear that I'm using it, because then that either poses a challenge, or they'd have to go up yet another learning curve, and they would prefer to retire in peace before having to do anything like that. So I sit there the same number of hours a day, and produce all that I'm asked. At their level of visibility and legibility (and value, after all, they are the ones paying me), there is no productivity gain. At my level of visibility, there is 1000% productivity gain.
(2) I've been testing whether my own subordinates, let alone new graduates, can keep up with what is a comparatively low-grade chatbot, for things like research and memoranda. They can't. It's not even CLOSE. This is not the Star Trek scifi future. I could have them all replaced TODAY and produce more, at higher quality, and with using less of my own time crafting precise instructions, reviewing, editing, and revising.
Often, when the chatbot did a better job, I just tell a subordinate, "good job" and throw their work product in the trash, and maybe polish the chatbot product a bit. I am getting very good at the skill of "humanizing" the uncanny-valley impact of chatbot writing style, and also tailoring that humanization to the preferences of the leadership. "Those new service-sector jobs!" My superiors apparently appreciate this, which is hilarious, and feels like being skilled at plagiarizing by editing just enough to pass. Do I tell anybody this? Are you crazy? Of course I don't.
(3) Unfortunately the tech bros where I work are never going to let me get access to any API to use these capabilities to build web apps and other tools. They are doing that for security or credentials reasons, but the consequence will prevent anyone like me anywhere in the organization with knowledge of the tasks and what could be automated from immediately creating major change. But, I know that with access to those capabilities I could, in AN AFTERNOON-, pretty much automate about 90% of what several hundred people are getting paid well to do now. Now I like those people - well, 'like' is strong, I sympathize with them, I want them to be able to pay their mortgages a little longer - and I have zero ability to personally benefit from any of that automation, so I am hardly going to push the issue, while I try to think about defending my own ability to earn an income against any other me's out there. But merely being made -AWARE- that even non-programmers like me can suddenly wield such incredible tools that could, at least theoretically, have an immediate devastating impact on the lives of hundreds or thousands, is a deeply shocking experience.
People who haven't experienced something like this are like Wile E. Coyotes who haven't looked down to realize that, while they weren't looking, the whole world has disappeared under their feet, and gravity is going to do its thing to them any second now.
ChatGPT is my new favorite way to shop. Love hearing about the pros, cons, and tradeoffs of various products I’m considering purchasing. And it will put everything into a table for comparison. What’s so glaring about this new tool is how bad the rest of the internet is for shopping research.
I also find it useful for coming up with questions for Socratic dialogue with my kids. Michael Strong recommended this and he was right.