I'm not particularly interested in AI, and just wish the bubble would burst and go away, but every once in a while the problems it creates get so urgent I can't avoid writing about them. Two years ago it was the environmental footprint of all the fancy new data centres. Now, I'm afraid, it's everybody's safety. I started from the recent safety report that was covered in the media and addresses issues like deep fake videos, fraud, job losses, false health advice etc.
What I found to be missing from the report, however, is a lot more worrying than that, namely the fact that AI is already choosing targets in the wars in Gaza and Iran. So we allow it to kill humans. What could possibly go wrong . Well, it only gets more depressing from there (eg there's the AI2027 scenario which came out last year), so the latest feature is not for the faint-hearted:
Handing over to AI
Current Biology Volume 36, Issue 7, 6 April 2026, Pages R265-R267
Restricted access to full text and PDF download
(Unfortunately, this year's features will no longer become open access one year after publication - do contact me if you would like a PDF. Last year's features will still move to the open archives as this year advances.)
Magic link for free access
(first seven weeks only)
See also my new Mastodon thread where I will highlight all this year's CB features.
My mastodon posts are also mirrored on Bluesky.
Last year's thread is here .
Drones guided by AI are already a common feature of current conflicts. (Photo by Mass Communication Specialist 1st Class Victoria Granado/Wikimedia Commons.)
No comments:
Post a Comment