dodeqaaa's profile picture.

dodeqaa

@dodeqaaa

dodeqaa reposted

This uninterrupted streak has to end

reshetz's tweet image. This uninterrupted streak has to end

dodeqaa reposted

“NPR will no longer post fresh content to its 52 official Twitter feeds, becoming the first major news organization to go silent on the social media platform.” It feels the “government-funded media” label is inaccurate and misleading. npr.org/2023/04/12/116…


dodeqaa reposted

It’s progress right? ^^

This post is unavailable.

dodeqaa reposted

Perhaps some of you kind souls who follow your favorite "ethicists" would let them know their asses are busted? Virtual kisses if you call out their hypocrisy on this, as they are WELL aware! (They blocked me some time ago.) 🙏


dodeqaa reposted

I asked several other people in the Rationalist community about this, and they told me that they weren't familiar with it, BUT that they think it's 100-percent plausible, given what the community is like and how it operates. This is hardly the tip of the iceberg.


dodeqaa reposted

One with intimate knowledge of the community since it's founding told me that it's become a "full grown apocalypse cult," and hence the life insurance thing wouldn't be surprising.


dodeqaa reposted

The dataset you're talking about for GPT-3, which is hosted somewhere else for legal reasons, is referenced on HuggingFace by their Employees. They know it's illegal and infringing, but they make sure HF makes it easy to download it.

This post is unavailable.

dodeqaa reposted

HuggingFace is the most "legit" site for ML/AI research and datasets. Usually datasets are linked to a COMPANY or User in the URL like so: /dataset/company/name But here it's almost a "root directory". They are facilitating this, and dodging liability. huggingface.co/datasets/the_p…

stealcase's tweet image. HuggingFace is the most "legit" site for ML/AI research and datasets.
Usually datasets are linked to a COMPANY or User in the URL like so:
/dataset/company/name

But here it's almost a "root directory".

They are facilitating this, and dodging liability.
huggingface.co/datasets/the_p…
stealcase's tweet image. HuggingFace is the most "legit" site for ML/AI research and datasets.
Usually datasets are linked to a COMPANY or User in the URL like so:
/dataset/company/name

But here it's almost a "root directory".

They are facilitating this, and dodging liability.
huggingface.co/datasets/the_p…

dodeqaa reposted

My hope is that we'll use AI to lift up the human spirit, not to crush it. The people who are so eager to talk about how humans will soon be obsolete make me sad. Humans are the source of the ground truth of the models being trained, and their only audience. AI is our creation.


dodeqaa reposted

Sure. AI is our creation. So is capitalism, colonialism, slavery, cigarettes, & nuclear missiles. Technology isn't neutral or "pure". Tech utopianism/dystopianism shouldn't just focus on the model itself: the world we have right now is NOT ripe for human flourishing with AI.


dodeqaa reposted

Yo...#VoiceActors ...Why the hell are you auditioning for shit that is going to be used to take our jobs away. I don't care how much they are paying. 50 AUDITIONS for this on Voices...ARE YOU KIDDING ME?!!??

kat_loveland's tweet image. Yo...#VoiceActors ...Why the hell are you auditioning for shit that is going to be used to take our jobs away. I don't care how much they are paying. 50 AUDITIONS for this on Voices...ARE YOU KIDDING ME?!!??
kat_loveland's tweet image. Yo...#VoiceActors ...Why the hell are you auditioning for shit that is going to be used to take our jobs away. I don't care how much they are paying. 50 AUDITIONS for this on Voices...ARE YOU KIDDING ME?!!??

dodeqaa reposted

Some people relate and care with the harms of automated exploitation. The people who can extract their cut... conveniently find ways to not have to care.

And following the money. As in why there is so much less to go round. As in the lies & propoganda for austerity, wage freezes or caps. On Australian news, not since the 1960s here, has 93% of economic gains gone to the wealthiest 10%. Leaving 7% for the 90% of civilians. PyramidS



dodeqaa reposted

The rate at which procgen is misunderstood as "AI" or machine learning is increasingly annoying


dodeqaa reposted

"This is a kind of fraud that imitates TV evangelists: if you don’t send us money, you will go to hell." That's basically all of these orgs at the moment. If you don't give us the money, China will do the devil AGI and end the world, but white bros will bring AGI utopia.


dodeqaa reposted

The piece from @datasociety’s @jennaburrel is fire! techpolicy.press/chatgpt-and-co… But strong as it is I think it *understates* AI’s copyright problem at 2 points: 1st, the training data is not just “old digitized books” it includes new pirated ebooks as well….


dodeqaa reposted

The whole article is wild but this part discussing Eleuther’s (a group comprised of the usual suspects) scope of data gathering for a GPT-3 model, using anonymous entities to avoid legal risks (potential data laundering?) and the “copyright takedown policy” is mind blowing .

kortizart's tweet image. The whole article is wild but this part discussing Eleuther’s (a group comprised of the usual suspects)  scope of data gathering for a GPT-3 model, using anonymous entities to avoid legal risks (potential data laundering?) and the “copyright takedown policy” is mind blowing .
kortizart's tweet image. The whole article is wild but this part discussing Eleuther’s (a group comprised of the usual suspects)  scope of data gathering for a GPT-3 model, using anonymous entities to avoid legal risks (potential data laundering?) and the “copyright takedown policy” is mind blowing .

dodeqaa reposted

Four factors against AI scraping falling under fair use.

When it comes to the use of copyrighted works in training generative artificial intelligence (AI) systems, does the unauthorized use qualify as #fairuse? Learn more from our first blog in a two part series on #AI and #copyright by Cala Coffman: bit.ly/3Ut7OuT

Unite4Copyright's tweet image. When it comes to the use of copyrighted works in training generative artificial intelligence (AI) systems, does the unauthorized use qualify as #fairuse? Learn more from our first blog in a two part series on #AI and #copyright by Cala Coffman:
bit.ly/3Ut7OuT


Loading...

Something went wrong.


Something went wrong.