I've been wondering what 'job' ChatGPT does - either supplementing / changing existing roles, or creating new ones. Many of the use cases shared so far feel like flavors of 'assistant', what we often call junior staff, or, in older domains, apprentice. I don't see concrete applications for that in spaces like mine right now (leading a design / strategy function for a Big Tech business unit). There's other vectors for finding a path to productization, of course, but there's also ample history and data for this one across industry and academia broadly.
Difficult to share your private information to it, but I have been astonished by how useful it is at critiquing and generating new concepts in some (but not all) of our domains.
I've shared with a friend in a different discipline who is an academic (with several PhD students) and she was pretty surprised by the quality of its riffing.
I've got quite granular wih some of my questions but, admittedly, not as granular as you might need to in an actual business role.
Riffing on that, assuming you could work within corporate information boundaries - e.g. "keep things in the firewall..." - I can see two useful jobs for an Assistant who can:
1) take lots of (mostly words, but including some numbers) material in,
2) synthesize / distill meaning in relation to recurring questions or topics of focus, and then
3) summarize and publish qualitative and quantitative insights back out.
The first is status report / recurring readout authoring, which eats managerial time and attention like nobody's business at many levels. The second is initial parsing / coding of raw inputs around voice of customer, product discovery, research, etc.
There may be more.
It feels like the key aspects of these are: offload repetitive summarization (weekly / monthly / quarterly) that's not accompllishing much in the way of real human sense making or generating new thinking; and digesting a large spike of material more quickly / in parallel than with a single human perspective.
There is a better ChatGPT called Claude, built by a load of OpenAI alums who jumped ship after GPT3, (Possibly earlier) harder to Jail break, doesn't make so many obvious mistakes, and admits to them when it does.
I'd be interested to chat to ChatGPT, if it could actually chat, i did try to get it to write stuff, but it just parroted stuff back at me for my use case. Run the male and female Hero's Journey back to back, etc. Though I did find somebody who was working out how to give it a blank slate and prime it for what you want. Not been able to get a connection to it since. Kobold AI isn't bad if you install it locally. That has a range of modes, Game, Story, Chat, etc. And different models you can run. Might be worth a look for the technically competent. It's a Python install that provides a Web interface on a local port. Just like Standard diffusion.
For those who recall, search engines were a paid service. That window was very short. We now take that for granted through our browsers. I'm betting we see a similar situation with chat bots.
Well, exactly speaking the search engines are still not free, you pay at least with your data. A trend I hope to itself repeat with chat bots. Though I guess over time cost will reduce drastically because of economy of scales.
So after all of the downtime I was a little despondent as it seemed unusable. However my note taking program MEM.ai recently integrated it into its premium monthly package and now, as you say Azeem it’s an indispensable part of my workflow. I had to buy more credits via MEM which still has teething issues but having it built into where I work had changed the game. I was also thinking about how it will likely show up in apps like Grammarly, and my email program Polymail mentioned that they may integrate it also at some point.
Interesting, I have struggled with Mem.ai after using it for 6-7 months. Fell in love and then out of love quite quickly. I just wasn't sure where I was looking. Glad to hear you are having a good experience.
I've been wondering what 'job' ChatGPT does - either supplementing / changing existing roles, or creating new ones. Many of the use cases shared so far feel like flavors of 'assistant', what we often call junior staff, or, in older domains, apprentice. I don't see concrete applications for that in spaces like mine right now (leading a design / strategy function for a Big Tech business unit). There's other vectors for finding a path to productization, of course, but there's also ample history and data for this one across industry and academia broadly.
Difficult to share your private information to it, but I have been astonished by how useful it is at critiquing and generating new concepts in some (but not all) of our domains.
I've shared with a friend in a different discipline who is an academic (with several PhD students) and she was pretty surprised by the quality of its riffing.
I've got quite granular wih some of my questions but, admittedly, not as granular as you might need to in an actual business role.
Riffing on that, assuming you could work within corporate information boundaries - e.g. "keep things in the firewall..." - I can see two useful jobs for an Assistant who can:
1) take lots of (mostly words, but including some numbers) material in,
2) synthesize / distill meaning in relation to recurring questions or topics of focus, and then
3) summarize and publish qualitative and quantitative insights back out.
The first is status report / recurring readout authoring, which eats managerial time and attention like nobody's business at many levels. The second is initial parsing / coding of raw inputs around voice of customer, product discovery, research, etc.
There may be more.
It feels like the key aspects of these are: offload repetitive summarization (weekly / monthly / quarterly) that's not accompllishing much in the way of real human sense making or generating new thinking; and digesting a large spike of material more quickly / in parallel than with a single human perspective.
i think you could, in a private environment, get it to synthesise solutions to optimisation or productivity problems as well.
There is a better ChatGPT called Claude, built by a load of OpenAI alums who jumped ship after GPT3, (Possibly earlier) harder to Jail break, doesn't make so many obvious mistakes, and admits to them when it does.
I'd be interested to chat to ChatGPT, if it could actually chat, i did try to get it to write stuff, but it just parroted stuff back at me for my use case. Run the male and female Hero's Journey back to back, etc. Though I did find somebody who was working out how to give it a blank slate and prime it for what you want. Not been able to get a connection to it since. Kobold AI isn't bad if you install it locally. That has a range of modes, Game, Story, Chat, etc. And different models you can run. Might be worth a look for the technically competent. It's a Python install that provides a Web interface on a local port. Just like Standard diffusion.
For those who recall, search engines were a paid service. That window was very short. We now take that for granted through our browsers. I'm betting we see a similar situation with chat bots.
Well, exactly speaking the search engines are still not free, you pay at least with your data. A trend I hope to itself repeat with chat bots. Though I guess over time cost will reduce drastically because of economy of scales.
So after all of the downtime I was a little despondent as it seemed unusable. However my note taking program MEM.ai recently integrated it into its premium monthly package and now, as you say Azeem it’s an indispensable part of my workflow. I had to buy more credits via MEM which still has teething issues but having it built into where I work had changed the game. I was also thinking about how it will likely show up in apps like Grammarly, and my email program Polymail mentioned that they may integrate it also at some point.
Interesting, I have struggled with Mem.ai after using it for 6-7 months. Fell in love and then out of love quite quickly. I just wasn't sure where I was looking. Glad to hear you are having a good experience.