Back before time began (in the 1990s) I used to maintain an open source software package. I was being paid to do it, so I had a lot of freedom to decide how to approach it. Once its popularity took off a bit, I started to get a lot of requests for help and bug reports on the mailing list. At first I found this hugely problematic—it was really time consuming. But after a while I started to notice patterns in the questions and bug reports, and realized that rather than being a problem, they were a trove of information.
So I started to just figure out what to do next based on the patterns I was seeing on the mailing list. I was doing rapid releases, so the feedback loop was fairly tight, and what I found was that after a while, the traffic started to really decline. I'd still get the occasional thing, but nowhere near as much.
This required a bit of humility—sometimes the user would do something that seemed completely stupid to me, and I needed to make a change because of that, and I _wanted_ to argue with them that they were being silly. But in fact there was usually a way to genuinely improve the software as a result of their question, even if it felt like a dumb question.
The software has since been EOL'd, so this is ancient history, but it honestly surprises me how infrequently I see customer-facing organizations do this sort of thing. Rather than simply addressing the customer report, the organization doubles down and tries to do nothing, because the customer is mistaken.
It's really sad that this reaction is so prevalent. Imagine if instead of trying to find new ways to ignore the customer, they actually fixed the customers' problems!
I fear that humility and/or transparency (2 facets of the same thing, I'd suspect) are individual qualities that humanity just can't seem to bring from the individual level into the broader strokes of civilization for some reason, and the best we've been able to do to date is the enforced transparency of a functioning democracy. It certainly hasn't made its way into the corporate world to any observable degree.
That's an understandable position to take, but it hasn't always been this way. There are knobs to turn. The problem is they were turned the wrong way so long ago that we tend to think of them as givens rather than variables. But even in the last ten years this has obviously changed for the worse, which means better is not impossible.
Grownups; perhaps, yes: as measured in dog-years' time. Ehm.. I've assumed you were referencing the 'grownups' at The Economist that wrote the article...
Similarly the last part of the talk where OpenAI explain the extraordinary coordinated multi agent attack that they accidentally unleashed on hugging-face, is all about how we need more and more ai powered automated cyber defence to protect us from the new paradigm of ai powered automated cyber attack.
Last November, my water bill was 10x its normal amount. I had a leaky appliance (that I had fixed). I used Chat GPT to navigate the appeals process and got my water bill appealed. My hope is that rather then a cycle of escalation, AI reveals all the things that are intentionally opaque, and then we just make better bureaucratic systems. But I am probably naive
This is the exact sell for hiring systems. Drowning in the flood of AI generated applications and resumes? You need an AI driven applicant tracking system to automatically handle the sourcing, filtering, and interviewing. AIs all the way down.
Everyone doing this should still be reading the material and correspondence. These agents are irreducibly sloppy and this is serious stuff in a person’s own name.
"More AI" doesn't inspire confidence as a solution to AI. Remagined processes and "better AI" might be. The example of automatically assessing whether a design complies with building code is compelling. This kind of thing is already reducing bottlenecks.
Policies that implicitly rely on friction will need to be redesigned: either apply AI to move faster or stipulate limits on response time.
Excellent callback to a fantastic publication
Back before time began (in the 1990s) I used to maintain an open source software package. I was being paid to do it, so I had a lot of freedom to decide how to approach it. Once its popularity took off a bit, I started to get a lot of requests for help and bug reports on the mailing list. At first I found this hugely problematic—it was really time consuming. But after a while I started to notice patterns in the questions and bug reports, and realized that rather than being a problem, they were a trove of information.
So I started to just figure out what to do next based on the patterns I was seeing on the mailing list. I was doing rapid releases, so the feedback loop was fairly tight, and what I found was that after a while, the traffic started to really decline. I'd still get the occasional thing, but nowhere near as much.
This required a bit of humility—sometimes the user would do something that seemed completely stupid to me, and I needed to make a change because of that, and I _wanted_ to argue with them that they were being silly. But in fact there was usually a way to genuinely improve the software as a result of their question, even if it felt like a dumb question.
The software has since been EOL'd, so this is ancient history, but it honestly surprises me how infrequently I see customer-facing organizations do this sort of thing. Rather than simply addressing the customer report, the organization doubles down and tries to do nothing, because the customer is mistaken.
It's really sad that this reaction is so prevalent. Imagine if instead of trying to find new ways to ignore the customer, they actually fixed the customers' problems!
One of the main hallmarks of this era, rather sadly, is a distinct lack of humility.
True, but it doesn't hurt to point out the alternative! :)
I fear that humility and/or transparency (2 facets of the same thing, I'd suspect) are individual qualities that humanity just can't seem to bring from the individual level into the broader strokes of civilization for some reason, and the best we've been able to do to date is the enforced transparency of a functioning democracy. It certainly hasn't made its way into the corporate world to any observable degree.
That's an understandable position to take, but it hasn't always been this way. There are knobs to turn. The problem is they were turned the wrong way so long ago that we tend to think of them as givens rather than variables. But even in the last ten years this has obviously changed for the worse, which means better is not impossible.
Eleven-year-old me can't believe that Abelard Snazz is being referenced by actual grownups
Grownups; perhaps, yes: as measured in dog-years' time. Ehm.. I've assumed you were referencing the 'grownups' at The Economist that wrote the article...
Tim Long, Just Up the Hill from Lock 15
Similarly the last part of the talk where OpenAI explain the extraordinary coordinated multi agent attack that they accidentally unleashed on hugging-face, is all about how we need more and more ai powered automated cyber defence to protect us from the new paradigm of ai powered automated cyber attack.
https://youtu.be/87DyyMV0kCY?is=HrOsaXqfAQOvv_gW
Last November, my water bill was 10x its normal amount. I had a leaky appliance (that I had fixed). I used Chat GPT to navigate the appeals process and got my water bill appealed. My hope is that rather then a cycle of escalation, AI reveals all the things that are intentionally opaque, and then we just make better bureaucratic systems. But I am probably naive
Thank you. What I found interesting is how ‘solutionism’ has now become only an AI problem... I wonder why?
This is the exact sell for hiring systems. Drowning in the flood of AI generated applications and resumes? You need an AI driven applicant tracking system to automatically handle the sourcing, filtering, and interviewing. AIs all the way down.
Everyone doing this should still be reading the material and correspondence. These agents are irreducibly sloppy and this is serious stuff in a person’s own name.
"More AI" doesn't inspire confidence as a solution to AI. Remagined processes and "better AI" might be. The example of automatically assessing whether a design complies with building code is compelling. This kind of thing is already reducing bottlenecks.
Policies that implicitly rely on friction will need to be redesigned: either apply AI to move faster or stipulate limits on response time.