Anthropic's Claude confidently and incorrectly diagnoses brain cancer based on an MRI.
this post was submitted on 18 Mar 2024
1 points (100.0% liked)
TechTakes
1432 readers
16 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
"But look how convincing [it] sounds!"
.... how did we get to the point where the ai bros are un-ironically telling us, as a selling point, that their shiny toy literally gives false yet convincing-sounding medical diagnoses ?!?!?!
If I were working on Claude and wanted to hype it up, I would not talk about this experiment online or in public. If I were working on Claude and wanted to be responsible towards "the public", I would use this example as a cautionary warning, not to further hype up the tool.
This feels like the slight period at the beginning of the NFT craze when I wasn't yet comfortable dismissing out of hand anyone excited about them, because surely there was a least some useful application that wasn't for scamming people, and surely this many people couldn't all be so deluded about the same idea.
oh lol just made this a post too
“Oh, some patient data. Let me quickly casually scan this into the sv datacorp. What’s that…privacy concerns? Naaaaah I changed the filename”
A friend that wants you to have an aggressive brain tumour to make an AI look good is no friend at all
There are far too many people in this world who learned both wrong things from the “Pray tell, Mr Babbage” anecdote
The path to hell is paved with liberal markets.
also a very big opinion with the rationalists and EAs
Someone on discord linked me to this AI generated horror-show of a "childrens" video: https://www.youtube.com/watch?v=nywBUSaHFgY (I hesitate to give them any more views, Internet Archive says youtube vids take a few days to process though)
Which appears to be a channel of entirely AI generated video paired to children's songs. And it is gosh darn creepy and gross and exploitive and unhealthy.
I hope youtube cracks down on shit like this.
Behind the Bastards had a few episodes on how AI children's books are on offer at Amazon - also creepy and unhealthy.
I think there's been variants of this for almost a decade and YT's fine with it. I learned about Elsagate (https://en.wikipedia.org/wiki/Elsagate) originally from Dan Olson (https://www.youtube.com/watch?v=LKp2gikIkD8) but it was earlier than that.
🙅 - youtube applies helpful filters and actively works to tamp down things actively gaming metrics 👉 - get dem tasty ad-clicks boiiiiiiiis
it's truly a place with one of the longest-standing incentives problem configurations too, I think? I recall people both on- and off-site there complaining about this form of it going back at least 6~8y by now
just realised that Musk's angry face in the Don Lemon interview is the same face as an angry Skibidi Toilet
if my style isn't ridiculous-ulous-ulous-ulous
a lotta yall still dont get it
AI equity holders can use multiple ZIRP juices on a single AI
so if you have 1 OpenAI and 3 ZIRP juices you can create 3 new AI startups
How many victory points is that
going to choose to believe that ZIRP juices are a previously unreleased Juicero product
If I came to believe the entire tech scene at this point was all ZIRP grift would I be right?
at this point it's the standard analysis, so yes
view more: next ›