I'm pretty sure Sam Altman made precisely the same point that asking great questions will be the distinguishing feature for those who use AI the most effectively. When I read pieces like this, I am even more disturbed by the disconnect between the narrowness of the conversation about AI in education compared to the level of intensity and excitement(?) around AI in the business world. I'm not sure teachers fully understand just how much this technology will transform the world in which students are entering and therefore only see AI through the lens of our current practice - the using AI to cheat conversation is a very real one at the moment and will only get more pronounced over time, but I am so glad to see how the Deep Research model also caused a fundamental shift for working professionals. I recently had a similar moment with respect to the mundane practice of student research papers and reached a similar conclusion - asking great research questions may replace the actual writing of the paper though that is going to be a tough sell to most educators. Thanks for this post. I'll be sure to be on the lookout for more.
The idea that AI can help us ask better questions rather than just providing answers is so powerful. It really shifts the value from simply collecting data to using it wisely. The concept of human networks becoming a competitive advantage makes so much sense especially as AI tools become more accessible.
What do you think will be the biggest challenge in integrating these Ai driven tools into human collaboration? Could it lead to more specialized fields or foster broader collaboration?
Some brief thoughts on the use of the “OpenAI deep research” (Odr) agent:
1. The use of meta-prompting shortens the time spent on the construction of the prompt for Odr. I give o3-mini a brief prompt and ask it to write a detailed prompt for Odr, which usually turns out to be at least a page long.
2. Meta-prompting also has the nice side effect that prompt construction relies on the knowledge of o3-mini rather than my own, which is often limited in areas for which colleagues ask me to engage Odr.
3. Curating the often voluminous reports can take an entire afternoon. I want to convince myself that things hang together. In rare cases I have found that results were lacking some important aspect, source materials were a bit outdated, or similar. This could usually be corrected by adjusting the prompt. By now, I use meta-prompting with “standard instructions”, so the generated reports are of high quality.
4. The biggest challenge for me is how to communicate the research results to co-workers. Many people are no longer willing to work through a detailed report for several hours, even if it is the highly technical report they asked for. I always ask for an Executive Summary at the beginning and tabular summaries of the main findings at the end of the report. As appendices I request a brief 5-slide summary for non-specialists and a longer set of slides for specialists. Hopefully these summaries can whet people’s appetites.
I am now using Odr for just about any interesting question that pops into my mind.
Original comic was XKCD and has been repurposed many times https://xkcd.com/2347
I'm pretty sure Sam Altman made precisely the same point that asking great questions will be the distinguishing feature for those who use AI the most effectively. When I read pieces like this, I am even more disturbed by the disconnect between the narrowness of the conversation about AI in education compared to the level of intensity and excitement(?) around AI in the business world. I'm not sure teachers fully understand just how much this technology will transform the world in which students are entering and therefore only see AI through the lens of our current practice - the using AI to cheat conversation is a very real one at the moment and will only get more pronounced over time, but I am so glad to see how the Deep Research model also caused a fundamental shift for working professionals. I recently had a similar moment with respect to the mundane practice of student research papers and reached a similar conclusion - asking great research questions may replace the actual writing of the paper though that is going to be a tough sell to most educators. Thanks for this post. I'll be sure to be on the lookout for more.
https://fitzyhistory.substack.com/p/the-20-page-research-paper-in-20
The idea that AI can help us ask better questions rather than just providing answers is so powerful. It really shifts the value from simply collecting data to using it wisely. The concept of human networks becoming a competitive advantage makes so much sense especially as AI tools become more accessible.
What do you think will be the biggest challenge in integrating these Ai driven tools into human collaboration? Could it lead to more specialized fields or foster broader collaboration?
Some brief thoughts on the use of the “OpenAI deep research” (Odr) agent:
1. The use of meta-prompting shortens the time spent on the construction of the prompt for Odr. I give o3-mini a brief prompt and ask it to write a detailed prompt for Odr, which usually turns out to be at least a page long.
2. Meta-prompting also has the nice side effect that prompt construction relies on the knowledge of o3-mini rather than my own, which is often limited in areas for which colleagues ask me to engage Odr.
3. Curating the often voluminous reports can take an entire afternoon. I want to convince myself that things hang together. In rare cases I have found that results were lacking some important aspect, source materials were a bit outdated, or similar. This could usually be corrected by adjusting the prompt. By now, I use meta-prompting with “standard instructions”, so the generated reports are of high quality.
4. The biggest challenge for me is how to communicate the research results to co-workers. Many people are no longer willing to work through a detailed report for several hours, even if it is the highly technical report they asked for. I always ask for an Executive Summary at the beginning and tabular summaries of the main findings at the end of the report. As appendices I request a brief 5-slide summary for non-specialists and a longer set of slides for specialists. Hopefully these summaries can whet people’s appetites.
I am now using Odr for just about any interesting question that pops into my mind.