Conversation

Notices

1. Embed this notice
   Charlie Stross (cstross@wandering.shop)'s status on Thursday, 19-Dec-2024 20:36:40 JST Charlie Stross

   Current mood: angry at the mere existence of the UK government's public consultation on allowing large AI firms to train on copyrighted data (it's a whitewash job, thrown up just before Christmas), yet too tired to read the full document and then write a response. (At least, today.)

   • Embed this notice
     Ben Tasker (ben@mastodon.bentasker.co.uk)'s status on Thursday, 19-Dec-2024 20:50:38 JST Ben Tasker

     in reply to

     @cstross I started responding yesterday but got fed up and had to hit Save and come back.

     As you might expect, a lot of the questions (at least as far in as I've got) tend to assume that you agree with the proposal.

     Charlie Stross repeated this.
   • Embed this notice
     Joanna McKenzie (jobob@mastodon.me.uk)'s status on Thursday, 19-Dec-2024 22:01:50 JST Joanna McKenzie

     in reply to

     @cstross I'm a data scientist and it's not even a good idea for /AI/. Sure that subset of applications that need Really Massive datasets will be easier, but only a handful of companies can access the compute power to train them anyway. Meantime the smaller, more targeted AI models that are, you know, processing x-ray images or making prescription data digital, don't see any benefit at all.

   • Embed this notice
     Charlie Stross (cstross@wandering.shop)'s status on Thursday, 19-Dec-2024 22:46:37 JST Charlie Stross

     in reply to

     • Joanna McKenzie
     • Benjohn

     @benjohn @Jobob Nope, and this isn't an accident—the big LLM corporations don't want artisanal competitors to get a nose in the door.

   • Embed this notice
     Benjohn (benjohn@todon.nl)'s status on Thursday, 19-Dec-2024 22:46:38 JST Benjohn

     in reply to

     • Joanna McKenzie

     @Jobob @cstross is there good terminology to distinguish smaller artisan / boutique ML models, like the example you give, from the “EAT ALL THE DATAS” kind?

   • Embed this notice
     Ben Tasker (ben@mastodon.bentasker.co.uk)'s status on Saturday, 21-Dec-2024 21:12:51 JST Ben Tasker

     in reply to

     @cstross I've finally finished and submitted it.

     Days of reading, answering, coming back and re-reading all to answer questions that aren't much better than "why do you think our preferred route is beneficial"

     Towards the end, there's a particular delight which moots dropping protection for computer generated works because there's no evidence that that protection has encouraged increased use of AI in the UK.