Americans don’t know what to do if their work hours contract. If Gen AI speeds productivity, we just do more. Gallop data show that America has high engagement workers compared to much of the world. For knowledge workers, where Gen AI is most useful, the achievement orientation and professional identity suggest that we will try to get a competitive advantage using AI. A recent paper in Nature showed that in the physical sciences, researchers using AI in their workflow publish 3x as many papers (not Gen AI: this goes back decades). It maybe part selection effect, but it may also reflect how AI speeds up research which is not used to work more leisurely. Instead, team sizes are lower and researchers become lab leaders faster. The knowledge economy is competitive, and we use efficiency gains to strive for success instead of taking our foot off the gas. I predict that this will be especially true for “makers” compared to “managers.” Gen AI speeds “making.” Creators, coders, writers, and researchers are all “makers” for whom Gen AI is a force multiplier. It would take a lot of discipline and intentionality to convert efficiency into lower workplace stress and better work-life balance.
" I predict that this will be especially true for “makers” compared to “managers.” Gen AI speeds “making.” Creators, coders, writers, and researchers are all “makers” for whom Gen AI is a force multiplier."
This is a great insight. I also saw another paper about researchers using GenAI, but they found that they were getting more papers out, but also being about to divide up findings into the minimums needed for pubs better. A large study might yield several related papers and conference presentations, and those using GenAI were able to get more papers and presentations out of the same amount of work. At least that was one of the ideas, not sure there was direct causation, but it makes sense for me.
Context switching is the big thing I’m trying to think through.
If people are expected to do 40 hrs of work in 8hrs, then you would have to keep track of 40 hrs of work in your head and try to maintain quality control for all that work. Seems like it would be much more stressful than just a regular 8 hrs of work.
The big thing I am hoping to see is salary increases as productivity increases and margins increase, but am doubtful everyone will see the rewards.
Yeah, I think we are entering another race to the bottom, more work and less pay. In the early 2000s when I was teaching an MBA MIS class, a common assignment that I gave was to interview someone who had been working for 20 year plus and ask how technology had changed their work. Students were surprised to find that the Internet destroyed a lot of profit. There was a lot of friction for consumers that needed to find alternatives or competitors. Tens of billions of dollars of profit went out of the insurance industry and others. Even Microsoft lost a lot. Linux would not have happened if the internet didn't exist and people could not collaborate easily. It would have been a small project among bbs nerds, and microsoft would have dominated the LAN server market with 70-90 percent margins.
Scott, my impression after reading both is that they were written for completely different audiences. I don't know if that was the intention. In looking at the prompt, I noticed you mentioned you want to send it out to your Substack, but I don't know if Claude has background information already that helps it know about the audience for your Substack or the Columbia Business Times in more detail. From what I've heard you say in presentations about your method, I suspect maybe it does, but it makes the prompt harder to evaluate for your readers.
Did it expand on the things that you felt most needed expansion? Were there other points where a different expansion would be more useful?
If it did this well with citations, would it have been better to use it to help with research from the beginning? If Gemini would have been better with YouTube, why not use it specifically for that purpose? I wonder if it might have produced more eye-catching diagrams? If you are going to use AI in this way, why not pick and choose the best ones for different tasks. Of course that would slow you down a bit.
That leads me to another question. When it comes down to speed (or productivity) vs quality, where do you draw the line? What is the balance? That may not be relevant for an exercise like this, but it would be very important in the workplace.
Going back to the research, and perhaps even to the original concept for the CBT article, I am struck that this is based on research at a tech company. A lot of the research seems to be into tech companies, programming, etc. I know there are some other studies, I think I saw a reference to one on law offices this week. But as Luiza Jarovsky noted yesterday, coding seems to stand in for all other tasks when discussing AI and human skills. She called it a "false equivalence."
I did not find one version really better than the other. They were just different. Do you feel using the AI in this particular way enhanced or detracted from your thought process? To me, that is the critical question.
" When it comes down to speed (or productivity) vs quality, where do you draw the line? What is the balance? "
I think this is a critical question and a difficult one to answer, and as these systems improve, the answer might change. The time spent increasing the quality of an output is an opportunity costs on another project or output. I think it is still very much early days, and we are getting it sorted. That is one reasons I wanted to try a simple prompt, and simple workflow. I don't think it is good enough, but I did complete some other work recently and spent a lot of time using one AI to do a meta analysis of a body of work and then used that output for a writing project, and it was better than I would have produced. It took me about 4-5 hours, but I think it would have been 80+ hours (since I am not that good at research, writing academic papers, etc.).
Americans don’t know what to do if their work hours contract. If Gen AI speeds productivity, we just do more. Gallop data show that America has high engagement workers compared to much of the world. For knowledge workers, where Gen AI is most useful, the achievement orientation and professional identity suggest that we will try to get a competitive advantage using AI. A recent paper in Nature showed that in the physical sciences, researchers using AI in their workflow publish 3x as many papers (not Gen AI: this goes back decades). It maybe part selection effect, but it may also reflect how AI speeds up research which is not used to work more leisurely. Instead, team sizes are lower and researchers become lab leaders faster. The knowledge economy is competitive, and we use efficiency gains to strive for success instead of taking our foot off the gas. I predict that this will be especially true for “makers” compared to “managers.” Gen AI speeds “making.” Creators, coders, writers, and researchers are all “makers” for whom Gen AI is a force multiplier. It would take a lot of discipline and intentionality to convert efficiency into lower workplace stress and better work-life balance.
" I predict that this will be especially true for “makers” compared to “managers.” Gen AI speeds “making.” Creators, coders, writers, and researchers are all “makers” for whom Gen AI is a force multiplier."
This is a great insight. I also saw another paper about researchers using GenAI, but they found that they were getting more papers out, but also being about to divide up findings into the minimums needed for pubs better. A large study might yield several related papers and conference presentations, and those using GenAI were able to get more papers and presentations out of the same amount of work. At least that was one of the ideas, not sure there was direct causation, but it makes sense for me.
Context switching is the big thing I’m trying to think through.
If people are expected to do 40 hrs of work in 8hrs, then you would have to keep track of 40 hrs of work in your head and try to maintain quality control for all that work. Seems like it would be much more stressful than just a regular 8 hrs of work.
The big thing I am hoping to see is salary increases as productivity increases and margins increase, but am doubtful everyone will see the rewards.
Yeah, I think we are entering another race to the bottom, more work and less pay. In the early 2000s when I was teaching an MBA MIS class, a common assignment that I gave was to interview someone who had been working for 20 year plus and ask how technology had changed their work. Students were surprised to find that the Internet destroyed a lot of profit. There was a lot of friction for consumers that needed to find alternatives or competitors. Tens of billions of dollars of profit went out of the insurance industry and others. Even Microsoft lost a lot. Linux would not have happened if the internet didn't exist and people could not collaborate easily. It would have been a small project among bbs nerds, and microsoft would have dominated the LAN server market with 70-90 percent margins.
Scott, my impression after reading both is that they were written for completely different audiences. I don't know if that was the intention. In looking at the prompt, I noticed you mentioned you want to send it out to your Substack, but I don't know if Claude has background information already that helps it know about the audience for your Substack or the Columbia Business Times in more detail. From what I've heard you say in presentations about your method, I suspect maybe it does, but it makes the prompt harder to evaluate for your readers.
Did it expand on the things that you felt most needed expansion? Were there other points where a different expansion would be more useful?
If it did this well with citations, would it have been better to use it to help with research from the beginning? If Gemini would have been better with YouTube, why not use it specifically for that purpose? I wonder if it might have produced more eye-catching diagrams? If you are going to use AI in this way, why not pick and choose the best ones for different tasks. Of course that would slow you down a bit.
That leads me to another question. When it comes down to speed (or productivity) vs quality, where do you draw the line? What is the balance? That may not be relevant for an exercise like this, but it would be very important in the workplace.
Going back to the research, and perhaps even to the original concept for the CBT article, I am struck that this is based on research at a tech company. A lot of the research seems to be into tech companies, programming, etc. I know there are some other studies, I think I saw a reference to one on law offices this week. But as Luiza Jarovsky noted yesterday, coding seems to stand in for all other tasks when discussing AI and human skills. She called it a "false equivalence."
I did not find one version really better than the other. They were just different. Do you feel using the AI in this particular way enhanced or detracted from your thought process? To me, that is the critical question.
" When it comes down to speed (or productivity) vs quality, where do you draw the line? What is the balance? "
I think this is a critical question and a difficult one to answer, and as these systems improve, the answer might change. The time spent increasing the quality of an output is an opportunity costs on another project or output. I think it is still very much early days, and we are getting it sorted. That is one reasons I wanted to try a simple prompt, and simple workflow. I don't think it is good enough, but I did complete some other work recently and spent a lot of time using one AI to do a meta analysis of a body of work and then used that output for a writing project, and it was better than I would have produced. It took me about 4-5 hours, but I think it would have been 80+ hours (since I am not that good at research, writing academic papers, etc.).
BTW, I think this post generated the most "unsubscribes" of any lately, so perhaps leaving the rewriting to Claude it not a great idea! 🤣