TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Meta's new AI image generator was trained on 1.1B Instagram and FB photos

338 pointsby my12parsecsover 1 year ago

27 comments

cowboyscottover 1 year ago
Is training with user-generated content a way to launder copyrighted images? That is, if I upload an image of Ironman or whatever to my Facebook or Instagram page as a public post and Meta trains their model on that data, is there wording in my user agreement that says that I declare that I own the content, which then gives Meta plausible deniability when it comes to training with copyrighted material?<p>(apologies for the run-on sentence - it is early still)
评论 #38559437 未加载
评论 #38564319 未加载
评论 #38559806 未加载
评论 #38559367 未加载
评论 #38559666 未加载
评论 #38562848 未加载
评论 #38564336 未加载
评论 #38562954 未加载
评论 #38570177 未加载
评论 #38559984 未加载
评论 #38564549 未加载
评论 #38559340 未加载
juntoover 1 year ago
Before anyone tries it out from the EU, be warned that it will push to make a Meta account and merge any Facebook&#x2F; Instagram profiles together and once you’ve finally bitten that bullet, it will tell you that it isn’t available in your region.
评论 #38565205 未加载
评论 #38563251 未加载
评论 #38561389 未加载
xnxover 1 year ago
Link to tool: <a href="https:&#x2F;&#x2F;imagine.meta.com&#x2F;" rel="nofollow noreferrer">https:&#x2F;&#x2F;imagine.meta.com&#x2F;</a><p>One of many AI updates from Meta yesterday: <a href="https:&#x2F;&#x2F;about.fb.com&#x2F;news&#x2F;2023&#x2F;12&#x2F;meta-ai-updates&#x2F;#:~:text=Experience%20Imagine%20With%20Meta%20AI" rel="nofollow noreferrer">https:&#x2F;&#x2F;about.fb.com&#x2F;news&#x2F;2023&#x2F;12&#x2F;meta-ai-updates&#x2F;#:~:text=E...</a>
评论 #38564415 未加载
评论 #38559263 未加载
评论 #38559569 未加载
评论 #38559478 未加载
评论 #38559481 未加载
tiffanyhover 1 year ago
1.1B is tiny.<p>Given that FB &amp; IG combined have ~0.5B photos uploaded <i>daily</i>, this effectively translates to training data from just a few days of user generated content.<p><a href="https:&#x2F;&#x2F;www.brandwatch.com&#x2F;blog&#x2F;facebook-statistics&#x2F;#:~:text=Users%20generate%204%20million%20likes,350%20million%20photos%20per%20day" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.brandwatch.com&#x2F;blog&#x2F;facebook-statistics&#x2F;#:~:text...</a>.<p><a href="https:&#x2F;&#x2F;www.zippia.com&#x2F;advice&#x2F;instagram-statistics&#x2F;#:~:text=considerable%20400%25%20increase.-,At%20least%2095%20million%20photos%20and%20videos%20are%20posted%20on,day&#x27;s%20worth%20of%20Instagram%20content" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.zippia.com&#x2F;advice&#x2F;instagram-statistics&#x2F;#:~:text=...</a>.
评论 #38564236 未加载
评论 #38563788 未加载
评论 #38566919 未加载
评论 #38564147 未加载
dmazzoniover 1 year ago
If you ask it to generate an image of Taylor Swift, it refuses. But if you ask it to generate an image of a popular celebrity singer performing the song &quot;Blank Space&quot;, it generates an image that looks exactly like Taylor Swift some fraction of the time.
评论 #38564219 未加载
WendyTheWillowover 1 year ago
Because it’s trained on “real” people, will it be easier to generate ugly people? I have a hard time convincing DALL-E to give me ugly DnD character portraits.
评论 #38560634 未加载
评论 #38559875 未加载
评论 #38560544 未加载
评论 #38560712 未加载
TheCorehover 1 year ago
“Not available in your location<p>Imagine with Meta Al isn&#x27;t available in your location yet. You can learn more about Al at Meta in the meantime and try again soon.”<p>I wonder why it&#x27;s region-locked?
评论 #38559434 未加载
评论 #38559246 未加载
astrangeover 1 year ago
Doesn&#x27;t do &quot;hard prompts&quot; better than other systems I&#x27;ve tried. Looks pretty similar to them too.<p>eg: &quot;horse riding an astronaut&quot;, &quot;upside-down mini cooper&quot;, &quot;kanji alphabet soup&quot;.
评论 #38563956 未加载
brucethemoose2over 1 year ago
&gt; It can handle complex prompts better than Stable Diffusion XL, but perhaps not as well as DALL-E 3<p>This is a interesting statement, as Stable Diffusion XL implementations vary from &quot;worse than SD 1.5&quot; to &quot;Competitive with DALL-E 3.&quot;
评论 #38561261 未加载
neilvover 1 year ago
Interesting. Unlike some other popular image generation training, is there a <i>chance</i> that Meta <i>technically</i> got copyright permission for many&#x2F;most of the images that were posted to its properties?<p>I&#x27;m thinking: When the user who uploaded the image was also the copyright holder, that might&#x27;ve been covered by an agreement that technically permitted this use by Meta.<p>(Copyright isn&#x27;t the only legal issue, though. For example, a person in a photo that someone else uploaded doesn&#x27;t necessarily lose right to their likeness being used for every purpose to which a generative AI service might be put.)
评论 #38565110 未加载
Havocover 1 year ago
Meta is asking me to log in with my facebook account. Then after authenticating with my FB account meta says I don&#x27;t have a meta account.<p>Is this all some sort of scam to get me to click accept on whatever godforsaken ToS comes with a meta account? If the FB account is good enough to freakin AUTHENTICATE me then just use that ffs.
RegWover 1 year ago
I wonder what other purposes FB have used those 1.1B+ publicly visible photos to train models for?
评论 #38565009 未加载
andsoitisover 1 year ago
The images of ourselves have now been absorbed into an AI.<p>An intelligence that knows a shit ton about a very very large number of people.
zoklet-enjoyerover 1 year ago
I&#x27;m not sure if they cut me off for generating too many images or because of the content of my images. Everything is now giving the response &quot;This image can&#x27;t be generated. Please try something else.&quot;<p>This only started after I put in the prompt manbearpig did 9&#x2F;11. It was ok with some really weird stuff though
miguelazoover 1 year ago
Wow, another reason to delete my accounts.
评论 #38560503 未加载
neomover 1 year ago
Really struggles with fingers, probably worse than any AI image generator I&#x27;ve seen so far. Maybe there aren&#x27;t a lot of finger-showing images on IG and FB!
al_be_backover 1 year ago
to me these innovations seem akin to Concept Cars in the Motor industry; there&#x27;s some utility, until some executive takes it center-stage, and pisses-off most of the core users.<p>the biggest value in these networks is real User-generated content, you can&#x27;t beat billions of real users capturing real content and sharing habitually.<p>even if wording in the Terms permit certain research&#x2F;usage, you&#x27;ve got market and political climates to consider.
nextworddevover 1 year ago
I tried this and was floored how good this was
andrewstuartover 1 year ago
And weirdly, every image it generates is sort of a combination of your grandma and an influencer on a beach on a tropical island.
jafitcover 1 year ago
All I can say is it’s really fast
lumostover 1 year ago
This is almost certainly going to be used to generate actual pictures of real people in the nude etc.
评论 #38559752 未加载
评论 #38559691 未加载
评论 #38559750 未加载
评论 #38560273 未加载
评论 #38561131 未加载
评论 #38559534 未加载
评论 #38559579 未加载
seydorover 1 year ago
So it&#x27;s just faces?
squigglydonutover 1 year ago
Artists...leave.
FpUserover 1 year ago
Canada. It asked me to create Meta account only to tell me that it is &quot;not available in your region&quot;.<p>Fuck you Meta and fuck you Zukerberg.
miked85over 1 year ago
Yet another reason to steer far clear of anything Meta.
__loamover 1 year ago
This is extremely shitty to a lot of users.
nothrowawaysover 1 year ago
The title is misleading. It uses publicly available photos, which means it uses the same image as other AI models like GPT, midjiurney ...
评论 #38565905 未加载