GregCocks · @GregCocks
718 followers · 569 posts · Server techhub.social
Kevin Dominik Korte · @kdkorte
7 followers · 239 posts · Server fosstodon.org

The battle between authors and AI companies is interesting. With Book3, there is a significant push to wipe the training data off the internet. Given that Meta and OpenAI can buy all the copyrighted material they need, the losers might be open-source projects and the knowledge of the training set. Both would be catastrophic for access and transparency. Let's hope fair use doctrines prevail. Please have a look at WIRED for a deep dive into the topic.
wired.com/story/battle-over-bo

#ai #trainingdata

Last updated 1 year ago

IT News · @itnewsbot
3643 followers · 270857 posts · Server schleuss.online
IT News · @itnewsbot
3613 followers · 270119 posts · Server schleuss.online
Cory Doctorow's linkblog · @pluralistic
44869 followers · 42996 posts · Server mamot.fr

But other political scientists sharply disagreed. Last year, @henryfarrell, and published a thoroughgoing rebuttal to Harari in **:

foreignaffairs.com/world/spira

They argued that - like everyone who gets excited about AI, only to have their hopes dashed - dictators seeking to use AI to understand the public mood would run into serious bias problems.

4/

#jeremywallace #abrahamnewman #foreignaffairs #trainingdata

Last updated 1 year ago

Nicole Hennig · @nic221
307 followers · 1514 posts · Server techhub.social

Exclusive: AP strikes news-sharing and tech deal with OpenAI axios.com/2023/07/13/ap-openai

#AI #trainingdata

Last updated 1 year ago

Blort · @Blort
926 followers · 9746 posts · Server social.tchncs.de

@lupyuen What I'd like to know is how / plan to legally defend scraping the entire public internet for their .

Guesses,
@sflc / @eff ?

#alphabet #google #llm #trainingdata #law #legal #ai #generativeai

Last updated 1 year ago

peteo · @peteo
59 followers · 407 posts · Server mastodon.nz

It worries me that tech commentators are talking about how generated content will future AI As if somehow pulling data from the internet could be considered clean and appropriate for training

(A reminder that training included content from

#ai #poison #trainingdata #openai #4chan

Last updated 1 year ago

Nicole Hennig · @nic221
290 followers · 1304 posts · Server techhub.social

Israel Ministry of Justice Issues Opinion Supporting the Use of Copyrighted Works for Machine Learning - Disruptive Competition Project project-disco.org/intellectual (same as Japan did)

#AI #copyright #trainingdata

Last updated 1 year ago

Mia · @mia
1467 followers · 1797 posts · Server hcommons.social

Great discusion at the 'Latent Space: From Datasets to Digital Heritage' event at Birkbeck today bbk.ac.uk/events/remote_event_

The eeriest part was uploading my profile pic from work to haveibeentrained.com/ and finding a tribe of middle-aged women with short blonde hair. I guess they're all real? Apart from the potential for haircut inspiration, imagine what we could achieve en masse!

#ai #data #machinelearning #trainingdata

Last updated 1 year ago

Nicole Hennig · @nic221
270 followers · 1189 posts · Server techhub.social

Japan Goes All In: Copyright Doesn't Apply To AI Training bit.ly/43fVqCn

#AI #copyright #trainingdata

Last updated 1 year ago

Nicole Hennig · @nic221
270 followers · 1188 posts · Server techhub.social

Japan's Copyright Exception for AI Training Data | Shelly Palmer bit.ly/43DGTjG

#AI #copyright #trainingdata

Last updated 1 year ago

Nicole Hennig · @nic221
270 followers · 1181 posts · Server techhub.social
Jason Pester (GameDev) · @jay
301 followers · 360 posts · Server mastodon.gamedev.place
Jason Pester (GameDev) · @jay
300 followers · 356 posts · Server mastodon.gamedev.place

Hmmm... Windows 11 / 10? File Explorer will include built-in Zip, 7-Zip, TAR, RAR, and possibly other archival format compression / decompression in a future update

Mouse Without Borders in latest PowerToys allows you to use one mouse+keyboard across multiple devices
(github.com/microsoft/PowerToys)

#microsoft #build #build2023 #microsoftbuild #cloud #mesh #chatgpt #gpt #openai #edge #bing #github #azure #ai #ml #ar #vr #mr #xr #windows #accessibility #inclusion #trainingdata #copyright #infringement

Last updated 1 year ago

Jason Pester (GameDev) · @jay
300 followers · 354 posts · Server mastodon.gamedev.place
Jason Pester (GameDev) · @jay
299 followers · 351 posts · Server mastodon.gamedev.place
peteo · @peteo
51 followers · 293 posts · Server mastodon.nz

@itnewsbot
In all our understanding of , there are still articles like this that are all sun and roses

Where is the discussion on project , on , of resources on quality of ?

Where is the statement?

Absolutely, businesses should be thinking about, and and usually preparing for future use of , but we are definitely approaching peak . Let's at least be a little professional in our discussions

#ai #risk #privacy #cost #trainingdata #conflictofinterest #hype

Last updated 2 years ago

George Kalyvas · @george
7 followers · 5 posts · Server me.dm

@coachtony
Giving writers the option sounds reasonable. Instead of opting everyone in without their consent or knowledge, let's have it as a setting writers can toggle on. Promote it so writers know about it and understand the rewards and tradeoffs of licensing their writing for LLMs. Also could it be anonymized somehow, in case someone doesn't want to be able to write a story in their style?

#chatgpt #medium #ai #llm #trainingdata #aicompanies

Last updated 2 years ago

Massimo · @kobaltauge
96 followers · 2045 posts · Server social.tchncs.de

@coachtony @cewcj


Training data for AI is not measured in quality or popularity only by amount. So every post is valuable for Ai therefore every post should get a part of the money.

#aicompanies #trainingdata #medium

Last updated 2 years ago