“ They cannot distinguish "unknown" from "false". Missing information is silently assumed away.
They are not consistent. The same scenario, rephrased, can produce a different answer.
They hallucinate citations. Article numbers that sound right but don't exist“
I don’t agree with all of these reasons why LLMs would fail at tax law. Maybe 2 yrs ago this held true but not now
You can determine and store missing data. create a structured output and ground it in actual tax law and use and cite only that knowledge base
I wouldn't use AI to write even if it could match my tone, but it currently doesn't do a good job of writing like me.
I tried with Opus 4.5 a few months ago to have it read my monthly retrospectives and then write a new one based on my weekly updates for that month. It was similar to the example I showed for James Mickens[0] where I see the similarities to my writing, but it feels more like someone parodying me than actually writing like me.
There is also a general mindset of worklife balance and enjoyment from life.
as someone who spends a lot of time in Spain but lives in the US, the Spanish prioritize social interaction much more than the US (sweeping statement I know) - you go to many towns and cities in Spain and locals are socializing multiple nights per week in vibrant bars and cafes an having so much fun. London has a bit of this with pub culture but less family friendly.
The US on the other hand, the focus is on work and friends rarely get together and we study why people are socializing less (bowling alone etc. ).
I don’t disagree with you, but this has been going on for a while…
Google monetized the the by indexing it and monetized what you wanted to find.
Facebook monetized the eyeballs from the pictures and posts you added.
Now LLMs will monetize all web content.
To play devil’s advocate - LLMs do give something back. Those with ideas and no coding experience can now build entire businesses for little to zero cost. This seems different
It does feel like the collaborative, free open nature of the web has gone and the optimism that brought… it feels like no one would build Foursquare today.
But then I wonder if I’m just old an jaded and to the younger generation creating content, for them the web is open and expressive- just in a different way
I can relate to this so much! IMHO Foursquare genuinely did gve the better recommendations for food and drink and I still think this recommendation problem is far from solved.
The natural world was not meaningfully abundant… Way before the industrial revolution land which was once used for opening hunting was closed off by the ruling class. Even before the Industrial Revolution you had a new class of merchant and factory owners who earned riches to buy land and keep the poor from hunting on it. Much of the natural resources out of reach for the majority and only accessible by those with deep pockets
What’s interesting is this feature (Spotify DJ) really excels when you give it qualitative input “workout music that pairs well with a sunrise” - and can deliver stronger results that hunting for a playlists
I’m fairly certain Spotify’s core meta data adheres to the US music industry largely set / reinforby Nielsen.
I’m curious why the author would want to happen with the feature if not move from 1 artist to another
For chart reporting? Am unfamiliar with the Nielsen standard, but given the state of musical metadata more broadly it's probably not very sophisticated.
Would expect any provider like Spotify to just export the reports Nielsen requires, not design their core systems around it.
I wouldnt have inagined AI overviews would impact tech news site heavily.
I can say personally I have gone from being a daily reader of sites like The Verge to virtually never visiting, the experience and quality of the content fell off a cliff and I try to seek out more meaningful content. Maybe this is a story more about people having alternatives and the rise of Substack and seeking out quality over dross?
Yep same. The reason I don't visit most of them anymore has nothing to do with AI and everything to do with ads and quality of content (which I suppose may have to do with AI) i.e. enshitification
I don’t agree with all of these reasons why LLMs would fail at tax law. Maybe 2 yrs ago this held true but not now
You can determine and store missing data. create a structured output and ground it in actual tax law and use and cite only that knowledge base
reply