What I'm wary of is that optimistic technologists are over extrapolating again: I see chatgpt (& similar) as this open-ended canvas that I can make things with. Other people can use this open-ended canvas to make things! Other people _will_ use this to make things!
Yes, there may be something different this time because the UI is just raw text. But so is a google search box.
Tim Wu's "The Master Switch" talks about how this isn't even limited to code. Same pattern of "open democratic exuberance leading to consolidation and oligopoly" happened for printing press, telegram, radio, television, the internet.
1. The warning here is exactly what you point out—this technology is *not* open or democratic and absolutely will lead to consolidation and oligopoly given the economic incentives and loose regulatory environment around it. Even if people *can* make things, the fact that its not permissionless will have a chilling effect on what people actually do make, since a centralized authority could pull the plug at any time.
2. An open-source permissionless alternative to ChatGPT/AlphaCode/etc would plausibly produce a different outcome. It still may be the case that most people *won't* make things with this, but the fact that they *can* and at least some folk on the long-tail *will* is all that's necessary for potentially huge shifts in innovation and other outcomes. Google is a search box, but not a creation tool. This is an open-ended creation tool with a familiar UI (thanks Google) and therefore *might* overcome the initial friction that has dampened the momentum of other tools in the no-code/end-user programming/design for emergence paradigm.
Agree on 1, which is why I thought StableDiffusion the coolest entrant. The linux to Windows.
re: 2. I hope you're right! After building and observing users of tech products for a while, I don't think tech is the limiting factor. It's attention: Writing and blogging is a tool that literally anyone has access to. How many journal?
Attention is certainly a limiting factor, but those must be contextualized against (legible) payoff. Writing and blogging has a very illegible and delayed payoff with a relatively high cost (it took me several hours to write this post!), whereas entering a prompt and getting a little robot that solves a problem for me in a few seconds has quite a different cost/payoff relationship and on a much shorter timescale. I don't expect a sea change in the near term (the robot just isn't good enough yet) but as far as I can tell it's clearly headed in that direction.
What I'm wary of is that optimistic technologists are over extrapolating again: I see chatgpt (& similar) as this open-ended canvas that I can make things with. Other people can use this open-ended canvas to make things! Other people _will_ use this to make things!
Yes, there may be something different this time because the UI is just raw text. But so is a google search box.
Tim Wu's "The Master Switch" talks about how this isn't even limited to code. Same pattern of "open democratic exuberance leading to consolidation and oligopoly" happened for printing press, telegram, radio, television, the internet.
Great points, two things:
1. The warning here is exactly what you point out—this technology is *not* open or democratic and absolutely will lead to consolidation and oligopoly given the economic incentives and loose regulatory environment around it. Even if people *can* make things, the fact that its not permissionless will have a chilling effect on what people actually do make, since a centralized authority could pull the plug at any time.
2. An open-source permissionless alternative to ChatGPT/AlphaCode/etc would plausibly produce a different outcome. It still may be the case that most people *won't* make things with this, but the fact that they *can* and at least some folk on the long-tail *will* is all that's necessary for potentially huge shifts in innovation and other outcomes. Google is a search box, but not a creation tool. This is an open-ended creation tool with a familiar UI (thanks Google) and therefore *might* overcome the initial friction that has dampened the momentum of other tools in the no-code/end-user programming/design for emergence paradigm.
Agree on 1, which is why I thought StableDiffusion the coolest entrant. The linux to Windows.
re: 2. I hope you're right! After building and observing users of tech products for a while, I don't think tech is the limiting factor. It's attention: Writing and blogging is a tool that literally anyone has access to. How many journal?
Attention is certainly a limiting factor, but those must be contextualized against (legible) payoff. Writing and blogging has a very illegible and delayed payoff with a relatively high cost (it took me several hours to write this post!), whereas entering a prompt and getting a little robot that solves a problem for me in a few seconds has quite a different cost/payoff relationship and on a much shorter timescale. I don't expect a sea change in the near term (the robot just isn't good enough yet) but as far as I can tell it's clearly headed in that direction.