What does the last sentence even mean? Let’s just ignore that.
That being said, why do you think the vibe of “doing taxes is a chore” and the meme “they should teach taxes in school” come up regularly?
I can think of multiple ways in which “AI” could help in doing taxes. LLMs could rephrase a form request in multiple ways or easier language to help people understand what is being asked of them. Language models could provide examples and cross references. You could have image models scan and recognize your receipts. A model trained on tax datasets could validate inputs beyond simple syntax / value checking and e. g. ask a person if they really meant to enter that one weirdly high value.
Naturally, the final result needs to be checked by the person submitting it and the program can’t be held liable, but please let’s not ignore the fact that related technology could be employed in a useful manner, I’m tired of discussions where perfect is the enemy of good.
And let’s not pretend doing taxes is incredibly easy for everyone. If you organize all your receipts perfectly all year round and always know what to put into every field of every form, good for you. Maybe there is someone else out there now suddenly having multiple jobs, or a single parent not knowing how and if to file some social benefits, both struggling with their taxes though. Maybe there are multiple countries with wildly varying tax processes, too, of which neither you or I may know anything.
Finally, AI does not always mean a process where a generative model hallucinates data.
What does the last sentence even mean? Let’s just ignore that.
That being said, why do you think the vibe of “doing taxes is a chore” and the meme “they should teach taxes in school” come up regularly?
I can think of multiple ways in which “AI” could help in doing taxes. LLMs could rephrase a form request in multiple ways or easier language to help people understand what is being asked of them. Language models could provide examples and cross references. You could have image models scan and recognize your receipts. A model trained on tax datasets could validate inputs beyond simple syntax / value checking and e. g. ask a person if they really meant to enter that one weirdly high value.
Naturally, the final result needs to be checked by the person submitting it and the program can’t be held liable, but please let’s not ignore the fact that related technology could be employed in a useful manner, I’m tired of discussions where perfect is the enemy of good.
And let’s not pretend doing taxes is incredibly easy for everyone. If you organize all your receipts perfectly all year round and always know what to put into every field of every form, good for you. Maybe there is someone else out there now suddenly having multiple jobs, or a single parent not knowing how and if to file some social benefits, both struggling with their taxes though. Maybe there are multiple countries with wildly varying tax processes, too, of which neither you or I may know anything.
Finally, AI does not always mean a process where a generative model hallucinates data.