Comment by refulgentis
3 hours ago
Get this: the byline, Keyana Sapp? A Palantir employee in AI strategy. https://news.ycombinator.com/item?id=47777209, flagged off main page)
Don't let them get away with this, they're using a topic that we all appreciate specifically to divide our reactions into "if it's AI, it's good! What's the problem?" and god knows what the actual endgame is. But it's certainly not Palantir maintaining a consumer rights blog.
FWIW fact check is great, their RAG stuff works fine. But the unsourced "anonymous anecdotes" are made up, can't find backing for any of them and they're sort of entry-level rage-bait. (ex. DC shoes snowboard boots now designed in Florida by people that never designed snowboard boot)
I'm going to admit I'm emotionally invested in this blog. I really enjoyed this write up.
I'm troubled by your statement because I can't tell if you are saying it is incorrect and AI made it and therefore BAD. Or if you are just saying AI made it and therefore BAD.
Writing is at such a precipice. Every time I compose an email, gmail underlines every single sentence to say there is a better and more concise way of saying it. I feel stupid so I generally accept it. Isn't this AI writing as well? But, the thoughts and intentions are mine.
How is this different here? The author is pointing out really relevant information that I know anecdotally to be true. If that story is 85% written by AI, or 15% written by AI, I still see the human behind it.
I'm troubled by AI writing, don't get me wrong. But, it deserves further thought. (And, I also have strong negative bias towards palantir...)
> the unsourced "anonymous anecdotes" are made up
This would be enough to make me distrust the whole thing.
If the byline said "written with AI assistance, anecdotes verified," I'd have no issue. The problem isn't the tool, it's the undisclosed institutional author using a sympathetic topic as a vehicle.
Editing suggestions are qualitatively different from writing the thing wholesale.
The anecdotes appear fabricated. I couldn't source the DC Shoes / Florida designer claim.
Generally, I'm sympathetic towards this worldview, and it would be disruptive to its impact to have obviously AI-written articles being passed around as authoritative grounding.
Even if 100% correct and verifiable, "obviously AI-written" and "100% correct" is enough for people who aren't as sympathetic as us to the overall point to dismiss it.
I'm sympathetic to their case, the phrase I'm trying to make happen is "AI DDOS'ing" - we can't people new to the material to read and verify endless reams of words, they approach infinity in the limit.
Really good points, thanks so much.
This feels like the new frontier for "journalism": can we easily verify it?
That feels like a step in the right direction, regardless of how it happens.
People like Steve Bannon know this, and have called what you describe "flooding the zone" and it works. Our brains can't process excess information so we search for simple answers. AI will make this exponentially more effective as a tactic for spreading narratives.
Can we have something like this blog that's not AI slop? Please!
This is an illustration of why AI is terrible: it just destroys trust. Is the blog good or bad? It's really hard to tell without putting in more work than it'd take to write a similar article.
At least in the pre-AI days, if you saw polished writing, it meant someone at least put some effort into it.
gross. I want to unread it now. annoying because this kind of brand degradation is something important to me. but that's not enough to make me read a Palantir AI dude.
Weirdly, the author has his mediocre GRE scores on his LinkedIn page.
I was working on a grant application a few days back, tl;dr: color science x LACMA, big art museum in LA. Opus 4.7 told me to include my AP exam scores from _19 years ago_. I took 13 in one year, which is highly rare, but certainly not CV/grant application material at 37 years old, and it didn't need to scrape the bottom of the barrel to find something to say.