At OpenSauce 2025, Second Hall for "Could AI Make This Panel?"

• Writing phrases the some of the speakers responded with. There are several people on the panel, but I don't recognize anyone other than Primeagen.
• Started off by asking whether or not the speakers would use AI and if it would make them more productive.
  • Generally against using generative AI to do creative work
  • AI should be used to automate the tedious work, like lipsyncing and rotoscoping in animation.
  • It's a mistake to try and replace an entire role with AI.
  • You should own and be responsible for all of your work, even if AI did that work.
  • If I'm writing a message, I'll ask for sentiment analysis and then take its feedback into account to edit the message.
• How do you decide whether or not the AI makes decisions for you or if the AI carries out decisions you've made?
  • If the AI does the work, it doesn't feel like I've done the work.
  • We've had rotoscoping assistance before AI.
  • It's really helpful to have generative images to communicate to other humans.
  • We've done some testing on AI writing scripts for videos, but it's not good enough yet. The AI does a better job when you try to provide guidance, but it's still bad.
  • Primeagen: Tried to ask ChatGPT if something they wrote is funny, but ChatGPT is not funny.
  • This could be an outcome of the output being made more safe
• Found a use for AI in code review to catch small mistakes, like when other humans are not available due to time zone differences.
  • I often use it for review, daily. I don't use the output of the review, but I take their suggestions into account.
  • AI code review used to not have enough short-term memory, and it's starting to get good at that.
  • We will have jobs indefinitely because they cannot carry out the entire work end to end.
  • AI seems good at assisting after the work is done, like spell check.
• Has anyone has surprising experience with AI recently?
  • Video generation has been spooky. It's starting to get difficult to tell whether or not a video is genuine.
  • You can follow the process and decisions a human does to deconstruct their work and understand and criticise it, but I cannot do that with AI.
  • The barrier of entry for producing fake pictures and videos is so much lower now, resulting in much more misinformation.
  • Live in-person events will start to get popular again because the value of virtual experiences is lowered.
• People are asking for summaries or reviews for things, HostFresh found that it causes people to go for bad products and services.

My thoughts:

• I found the responses of some on the panel cursed, particularly when it came to using AI to assist in communicating with other humans.
• No one mentioned the ironies of automation paper or any related concepts, which is too bad.