Post

Carl T. Bergstrom
@carlbergstrom.com

Ignore for a moment the question about whether Anthropic trains on user sessions. I am I right in reading this as saying that Anthropic's "never-before-discovered" reverse transcriptase was *patented* three years ago?

Carl Zimmer@carlzimmer.com

Four days ago, a scientist was surprised by the news that Anthropic discovered new virus genes for making DNA. He says he’s been studying them for years—and feeding his data to Anthropic’s AI as part of his own research. Coincidence? Here’s my story. nyti.ms/3VSWayc

nyti.msDid Anthropic’s A.I. Really Make a Scientific Discovery on Its Own? (Gift Article)An expert at the University of Copenhagen said his team had been sharing its research with the company’s A.I. model, Claude, and that its new finding matched their work.
Craig Kaplan
@triggerloop.bsky.social

The RT was annotated in the sequence previously for sure, but what was discovered was the associated clustered repeats that are expressed and appear to associate with the RT, i.e a potential anti-viral system of unknown function but potentially programmable in some way

4 reposts49 likes

Craig Kaplan
@triggerloop.bsky.social

There are now claims that the RNA and repeat aspect were derived from user prompts or potentially patent filings as Claude users had made this discovery earlier but had not published.

checkerbored.bsky.social
@checkerbored.bsky.social

Can't help but notice that every time one of these models "discovers" something it turns out that a human person did it earlier. Each specific case may be ambiguous as to whether the model came to it independently, but in aggregate it tells a very clear story...

Peter Ellis
@pjie2.bsky.social

Indeed. Not a new discovery - only question is whether Claude independently found them, it stole the info from supposedly private chats, or there was already some other public domain leakage of the info.