- It is like compression… but backwards ^^ - And lossy! 
- now that you mention it… - the real question is how long before we just have automated agents sending corporate emails to one another without any human in the loop 🤣 - Its 100% already happening. 
- Minus 10 years. 
 
 
- Most corporate communications are unnecessarily fluffy to begin with because it makes it look like more work was done. Most of the time I don’t even understand why I’m explaining something and it feels like the only requirement is to have words on a page. - Sometimes the only requirement IS to have words on a page. Think about a disaster recovery plan, for example. Now, you probably don’t want an LLM to write your disaster recovery plan, but it’s a perfect example of something where the main value is that you wrote it down, and now you can be certified that you have one. - I just asked GPT to create a disaster recovery plan for a ransomware attack, and actually the information it gave wasn’t wrong or bad. But it’s also very generic, and it will rarely/never tell you correctly the specifics to your applications or where to click. 
 
- there’s a whole book on the subject of bullshit jobs incidentally https://en.wikipedia.org/wiki/Bullshit_Jobs 
 
- This is a legitimate use case for LLM, though. - Not everyone can communicate clearly. Not everyone can summarize well. So the panel on the right is great for the people on the other end, who must read your poorly-communicated thoughts. - At the same time, some things must look like you put careful thought and time into your words. Hence, the panel on the left. - And if people on both sides are using the tool to do this, who’s really hurt by that? - Yes, but there is a real risk here that either the expansion added false details or the summary is wrong, especially the summary. 
 
- beware! soon, it will be able to turn that long email into a meeting! 
- The AI arms race has begun! - Isn’t this kinda thing happening already in the recruitment industry? - pretty sure stuff like resume screening is done using machine learning nowadays 
 



