Adam MosseriIn my job I spend a fair amount of time with both smart people who are incredibly skeptical of AI and its potentials and other equally intelligent people who are wildly optimistic. One way I square this circle is accepting that we all tend to overestimate what can happen in the short term and underestimate what can be accomplished over the long run.
The use cases I'm most interested in are around creativity. People often talk about ideas that look like a button that makes a finished video and shares it on your behalf, but that's not what this looks like. People have taste and emotional resonance that AI can't replicate. Instead, for creators, AI will show up as series tools of tools that make it easier for more people to be more creative.
Imagine a tool that translates your audio into another language and syncs your lips so you can grow in another country. Imagine an AI that gives you insight about the patterns of what's resonated and what's not about what you've shared in the past. Imagine one that reads over a brand deal contract and gives you an objective take on how favorable the terms are. AI doesn't need to disintermediate the creator. If we're in, we can find all sorts of ways for the technology to empower creatives.
ZippyVtuberđ
Igor Sovcik@mosseri I rather imagine tool that will protect my voice and face via some sort of adversarial patch against missuse if i am livestreaming, caling, etc. I also imagine tools that will filter or label Ai generated content on my social feeds since i don't want to interact with Ai content. Not that there can't be some good Ai generated content but at the moment we are witnessing mostly bad one if not malicious.
Adam@igisho @[email protected] I think a lot of people will probably also want to be able to filter AI generated content out of their feeds, but itâs not clear to me that anyone is going to be able to effectively catch all AI content as itâs getting harder and harder to detect as the models improve.
@mosseri @[email protected] i allways thaught that this is bigtechs competitive edge against OS federated networks. Same as Google is best at filtering your email. Imho: social nets of future has to be best of filtering unwanted trash from your social feeds. Also since EU Ai ACT and DSA you guys will be forced to label ai content anyway.
FifiSch@mosseri Tell this to the companies who create them and falsely advertise them, including your bosees.
(Yes, I know he can't read this, just needed to blow off steam)
@FifiSch @[email protected] Iâm here, but not sure I understand what youâre saying.
Aaron BurdettInteresting potential but I worry about âflatteningâ creativity. Look at YouTube: the structure, length, and even thumbnails across the platform all start looking the same because creators are financially incentivized to follow what the algorithm rewards. Iâm not sure I like a future where an AI directs the creative process because AI doesnât really innovate; it regurgitates.
Ivan VuÄica đłď¸@mosseri Generative ML is indeed too easily used and seen as "just type something and get the exact output", and too many folks will be using it that way. But gen-"AI" an excellent idea-starter and an excellent enhancer.
It would take a really long time for it to be a replacement, and people trying to use it that way (whether it's something as obvious as not having a human check the outputs, or less obvious as just reusing raw images/videos when telling a story or using it for bland voiceovers) are missing the point as much as full-out naysayers.
- Generative "AI" for code? Generates convincing but bad code that a human must validate. (It does sometimes summon a spark allowing it to come up with a good idea human would miss.)
- Generative "AI" for images? Generates bland images lacking ideas with obvious artifacts. (It can turn a sketch into a finished product, or it can put together a basic sketch of an idea that a human can re-do properly.)
- Generative "AI" for text? Generates convincing untrustworthy generic blather. (But it can be a starting point for research or implementation, as long as everything it outputs is treated as untrusted, verified multiple times -- carefully -- and, if the output is to be shared, completely rewritten.)
Generative "AI" is misused by both bad- and well-intentioned actors; however, it can be a nice tool (or at least a crutch), like the 'content-aware fill' and equivalents before it were.
I personally avoid consuming fully-ML-generated content, because if you can tell it's ML-generated, then it's because it's bland, non-transformative, and with an extremely high ethical cost attached to it. But, rejecting its use in day-to-day work and life is also leaving a valuable tool out.
Johanna, CanCon varietyLavender.
Itâs not âempoweringâ âcreativesâ.
Itâs further disempowering if not threatening the very existence of anyone but the elite few it was thoughtlessly built for.
@johannab @[email protected] itâs both. AI definitely threatens to crowd out content from creators. But it can also help in all sorts of ways. For platforms, the question is how do we maximize the benefits and minimize the costs. For creatives, the question is if and how the technology can help you do what you do.
For developers (&the enthusiastic platforms flogging their product) - how are you assuring from the start that you are mitigating the harm of inevitable abusers grabbing hold of tools when they are flung out there without oversight.
Taylor Swift & Gaza might both like to file stakeholder comments.
I swear, Adam, the LLM-fueled Facebook/Insta comment catfishers and clickbait frauds are a total trashfire and you(Meta, collectively) seem able to do NOTHING, as an enabler.
Every 3rd image in my FB feed is a realtor trying to sell me a vacation cabin that doesnât exist in a wilderness that somehow has an address not 3 blocks away from me.
I keep turning down hookups from Eddie Vedder on Insta and Threads, while heâs clearly having ChatGPT take over his publicity. đ
How does anyone who isnât Taylor Swift manage to validate authenticity? When a powerhouse like Meta cannot empower its own users to avoid or even just report these things?
Darius Kazemithis is such a cursed take in the specific examples even if I agree with the broad point of ML being useful in certain contexts
Me: let's use hammers to build houses instead of bash people's heads in
Meta guy: Some people are pessimistic about hammers. Some people are optimistic about hammers. I personally believe that we can use hammers to marginally perpetuate the shitty world we all live in
Word of Mouth đ 
@darius Yeah... the problem with hot takes like Meta Guy's is that the POV of the hot-taker is utterly ignorant of what it means to be a "creative". Even the term "creative" is MBA BS. No self-respecting artist or writer thinks of themselves as a "creative". That's more mealy-mouthed management jargon that reduces independent artists to a "job" at some shitty corporation. "Creatives" make the things their bosses tell them to make, because the bosses are all talentless "idea guys" who can't do fuck-all without paying someone else to do the actual work.
Kinda like "disintermediated". He's not talking to artists. He's talking to fellow middle-managers and MBAs. That's an economics term. And a decidedly transactional one.
Fuck this Meta Guy.
Kinda like "disintermediated". He's not talking to artists. He's talking to fellow middle-managers and MBAs. That's an economics term. And a decidedly transactional one.
Fuck this Meta Guy.
Paul-Gabriel Wiener@darius Imagine a tool that translates your audio and syncs your lips? You mean Star Trek's universal translator?
That tells you which of your posts resonated? You mean engagement analytics?
That reads a contract for you...? Wait. So the giant corporations that program the thing are going to replace your lawyer with an "unbiased" tool that will tell you if the contract terms are fair?
Only one of those seems potentially good, and it's from the 60s and has major pitfalls.
Jason Petersen (he)@darius I almost commented on this too and realized whatâs the point. But the guy seems like an absolute moron.
âBoth sides can be smartâ
âOh by the way actually one side is the one seeing things in reality the others are dumbâ
Truly idiotic writing.
@jason @darius Iâm not saying the other view is dumb at all. Iâm saying AI is going to be helpful and harmful to creatives, how it has both effects is going to change a lot over the next few years, not the next few months, and the effects will be relatively small in the short run and much more significant over the long run.
@mosseri you seem to be assuming this stuff is useful and arguing from there, which writes off an entire side that is arguing that it needs to prove itself more before we even move onto âis this helpful or harmfulâ
The long run thing isnât even a given. If boosters fuck this up enough it becomes radioactive to investors, weâre looking at another AI winter cycle before it gets more attention. The last one lasted roughly forty years.
filipa mv@darius this is an uncharitable read but:
an âobjectiveâ take on a law contract!
never having to learn about other cultures!
in a sense theyâre narrow minded fantasies.
âď¸ Information âď¸@darius Sadly we seem to have jumped from "wouldn't a natural semantic interface to useful features be nice" to "invest now apes we're going to the moon" and honestly the only solution I have is to give every person doing this a noogie.
Mohab@darius this all can just be distilled down to "imagine a world where AI can make you even more money"
It's all a question of money in their capitalist hellscape
dasebb@mosseri That would be cool!
MOVED to @@mosseri
Right now it looks like porn (but I suspect this didn't actually happen as reported).
https://mynorthwest.com/3956403/rantz-washingtons-lottery-ai-porn-user/
Eric MrozekThat can only really work if you casually ignore the tech CEOs who are saying that the aim is to replace labor.
If that's the case, our society is going to have some serious problems over the rest of this century.
If that's the case, our society is going to have some serious problems over the rest of this century.
@mosseri wow that went quickly from âI see both kinds of people on both sidesâ to âI am clearly in one side and strawmen on the other are wrongâ
Mike Liebhold@mosseri is it your kind of creative to ask an AI to compose a sad girl jazz version of boilerplate terms of service? https://twitter.com/goodside/status/1775713487529922702
Ryan FortneyMosseri i'm gonna be real with you here, if you think an LLM that plagiarizes and steals from living breathing artists is cool or good, and not one of the most evil inventions of the 21st century, I have no faith in Threads, or Meta
cm | Calgary | YYC | AlbertaIt's difficult for me to comprehend all of the necessary pieces of the puzzle when it comes to what's best for an app used by billions and keeping them happy and engaged + the needs of the business. Not a job I would want. It's easy for me to criticize and only think of myself, so I'm gonna stick to that. Can you add text formatting and links to feed posts on IG? And allow for more image ratio sizes in feed? Ty
Danny SkarkaYes. But this isnât a new feature. Itâs a potential atomic bomb.
Shoq@mosseri Well, perhaps. But they can also underestimate the threats in the short term, while overestimate our collective ability to mitigate them in the long.
HuddledMass@mosseri Tools depend on who wields them, and why. The people currently in charge of AI, shouldnât beâŚ
