Since generative AI was used to create the image that accompanies this post, I feel it’s prudent to mention that Jordan Acosta has a great list of free, ethically-sourced (i.e. non-AI) image resources Substack writers can use: https://www.jordanacosta.co/p/free-to-use-picture-resources
I’m not so sure. I’m a lecturer and exam author in accountancy and the current AI’s can’t add up anything correctly, make up references, can’t do double entry bookkeeping and are generally worse than the students. They will need to improve a great deal from here. From what I can see, actual experts in anything will tell you the same - it looks great if you only have a reasonable knowledge of whatever subject you’ve asked about, but for any more than that, it will need a lot of control. I use AI a lot, but basically as a very advanced search engine that gives me the information quickly that I can then go through to sift out the rubbish and correct all the errors.
Thanks for the thoughtful feedback, Aaron. All your points are valid - there are still plenty of bugs to work out before Skynet achieves self-awareness!
A factor I read about elsewhere is that it's all very well getting generative AI to do the leg work (draft the contract) and then get an experienced human lawyer to review and assure it; but that experienced human lawyer only got that experience by at some point being a junior lawyer who drafted contracts. Using AI for the lower grades cuts off the supply chain of experienced humans in the long run. So if we use generative AI for this stuff we will get to the point where there is no human competent to review and assure it. The counterargument I guess is that eventually the AI will be better than humans at all grades so why would you want a human anyway? - but in that case we will have arrived at a point of zero human agency and WE WILL NEVER KNOW if the AI is right because noone is experienced enough to judge it.
Thanks for the thoughtful feedback, Saxfrage! Yes, you make a valid point - it appears those entry-level, educational roles you speak of are already starting to evaporate and, as you point out, if that continues, we humans will just have to take our new AI overlords at their word in the coming decades.
Thanks, Trudi! I seem to remember you sending me an encouraging note three years ago encouraging me to persevere with this humble blog and I'm glad I took your advice. I'm OK, all things considered, hope you're doing well too.
Since generative AI was used to create the image that accompanies this post, I feel it’s prudent to mention that Jordan Acosta has a great list of free, ethically-sourced (i.e. non-AI) image resources Substack writers can use: https://www.jordanacosta.co/p/free-to-use-picture-resources
Thanks for the heads-up, Stephanie. Everyone please check out Jordan Acosta!
I’m not so sure. I’m a lecturer and exam author in accountancy and the current AI’s can’t add up anything correctly, make up references, can’t do double entry bookkeeping and are generally worse than the students. They will need to improve a great deal from here. From what I can see, actual experts in anything will tell you the same - it looks great if you only have a reasonable knowledge of whatever subject you’ve asked about, but for any more than that, it will need a lot of control. I use AI a lot, but basically as a very advanced search engine that gives me the information quickly that I can then go through to sift out the rubbish and correct all the errors.
Thanks for the thoughtful feedback, Aaron. All your points are valid - there are still plenty of bugs to work out before Skynet achieves self-awareness!
I’ll sell my ability to tell the truth 100% of the time. Every Ai I’ve tried seems incapable of doing that.
A factor I read about elsewhere is that it's all very well getting generative AI to do the leg work (draft the contract) and then get an experienced human lawyer to review and assure it; but that experienced human lawyer only got that experience by at some point being a junior lawyer who drafted contracts. Using AI for the lower grades cuts off the supply chain of experienced humans in the long run. So if we use generative AI for this stuff we will get to the point where there is no human competent to review and assure it. The counterargument I guess is that eventually the AI will be better than humans at all grades so why would you want a human anyway? - but in that case we will have arrived at a point of zero human agency and WE WILL NEVER KNOW if the AI is right because noone is experienced enough to judge it.
Thanks for the thoughtful feedback, Saxfrage! Yes, you make a valid point - it appears those entry-level, educational roles you speak of are already starting to evaporate and, as you point out, if that continues, we humans will just have to take our new AI overlords at their word in the coming decades.
PDF files full of prompts optimized for writing anti-AI blog posts. And highly fetishized pictures of my hands. My feet aren’t too easy on the eyes.
AI seems to be ushering in for developers at least what Nassim Taleb calls “extremistan” in his book the Black Swan. It’s a very worthwhile read!
Yep, I'm a big fan of Taleb, despite his tiresomely substantial ego. Definitely think we're headed into Extremistan in the near future.
Enjoyed this Nigel, hope you are well
Well maybe enjoyed is the wrong term but it was a great read
Thanks, Trudi! I seem to remember you sending me an encouraging note three years ago encouraging me to persevere with this humble blog and I'm glad I took your advice. I'm OK, all things considered, hope you're doing well too.