#gpt3

[aiweirdness]

Horti's new plant advice chatbot is based on gpt-3 and things are going well

Don’t ask an AI for plant advice • Tradescantia Hub

Plant seller Horti recently added an AI "plant doctor" to their website. Within hours of testing, it was offering potentially fatal advice.

🔥 writeup by tradescantia hub

[cfiesler]

Elon Musk did not create an AI trained on your fanfiction.

Hi, AI ethicist + fanfiction expert here. (This is one of those times where I feel uniquely qualified to comment on something...)

I’m seeing this weird game of telephone about the Sudowrite AI that I think started out pretty accurate, but now has become “Elon Musk created an AI that is stealing your fanfiction” (which frankly gives him far too much credit). I can probably say more about this, but here are a few things that I want to clarify for folks, which can be boiled down to “Elon Musk has nothing to do with this” and “this is nothing new”: Elon Musk is not involved in any way with Sudowrite, as far as I can tell. Sudowrite does, however, use GPT-3, the widely-used large language model created by OpenAI, which Elon Musk co-founded. He resigned in 2018, citing a conflict of interest due to Tesla’s AI development. It wasn’t until after he left that OpenAI went from being a non-profit to a capped for-profit. Elon Musk doesn’t have anything to do with OpenAI currently (and in fact just cut off their access to Twitter data), though I can’t find anything that confirms whether or not he might have shares in the company. I would also be shocked if Elon actually contributed anything but money to the development of GPT-3.

Based on Sudowrite’s description on their FAQ, they are not collecting any training data themselves - they’re just using GPT-3 paired with their own proprietary narrative model.  And GPT-3 is trained on datasets like common crawl and webtext, which can simplistically be described as “scraping the whole internet.” Same as their DALL-E art generator. So it’s not surprising that AO3 would be in that dataset, along with everything else (e.g., Tumblr posts, blogs, news articles, all the words people write online) that doesn’t use technical means to prohibit scraping. 

OpenAI does make money now, including from companies like Sudowrite paying for access to GPT-3. And Sudowrite itself is a paid service. So yes, someone is profiting from its use (though OpenAI is capped at no more than 100% return on investment) and I think that the conversations about art (whether visual or text) being used to train these models without consent of the artist are important conversations to be having.

I think it’s possible that what OpenAI is doing is legal (i.e., not copyright infringement) for some of the same reasons that fanfiction is legal (or perhaps more accurately, for reasons that many for-profit remixes are found to be fair use), but I think whether it’s ETHICAL is a completely different question, and I’ve seen a huge amount of disagreement on this.

But the last thing I will say is that this is nothing new. GPT-3 has been around for years and it’s not even the first OpenAI product to have used content scraped from the web.

[ohnoitstbskyen]

Hey TBskyen, what's your opinion on AI taking over, or at least really hurting the creative field. Like say taking over writing, art etc etc.

Sorry if this was asked before, AI just makes me wary of my own aspirations as a writer cause if an ai can just do what I do a lot better and faster is there a point in trying to publish that work?

What AI art and writing is a threat to is your professional career first and foremost. It is automation, and the function of automation is to drive down labor costs and outcompete artisans by sheer volume. I can't promise you that you'll be able to earn a living from writing in the future, this technology could genuinely destroy the commercial market, but if your worry is that the AI is going to simply be better than you, then let me put your fears to rest.

The AI cannot and will never be better than you.

These machine learning algorithms (which are not actually even AI, I should say), can only ever do one thing, which is reproduce the data that is fed into them. They can mix and remix that data in a hundred billion different configurations according to whatever parameters are specified, but they can't actually create anything.

Algorithms have nothing to add, they don't invent anything. They have no experience, they have no perspective, they have no intent. Algorithms will never write a story to express anything, they'll only reassemble parts of other stories to fit a desired output.

This is not to say that algorithm art won't pass the Turing test, that's a fairly low bar, just that fundamentally, algorithms will never, ever generate something that is of higher quality than what's fed into them. And they will never invent anything new, or add anything to the conversation.

Something which is true and will remain true forever is that somewhere out there, there is someone who needs the thing that you create. And they need that thing from you, in your voice, from your perspective and informed by your experiences. This isn't poetic fancy, this is observed experience. Humans tell stories and create art because we fucking need to. And we need these things to connect with one another.

That is always going to matter.

You might only have an audience of one. You might never make any money doing it. You might not even be alive when the person who needs your work finally finds it, or if shit goes really bad, it might be lost to time and they never find it.

But it fucking matters that you tried. Algorithm art is the mechanisms of capital trying to suck the soul out of one of the few areas of human existence that they haven't managed to drain completely yet, and to keep writing and creating while under this assault is a form of resistance that we sorely need.

[destinationtoast]

"Entertainment" and modern AI-fandom interactions

There's a 1985* sci-fi short story I once read by M.A. Foster called “Entertainment” that predicts what will happen in an increasingly AI-generated art world.  I can't find any excerpts or summaries of it online, but what I recall is:

In the future, humans can prompt machines to create any art -- e.g., "What would a collaboration between early-era Peter Gabriel and late-era Beethoven look like, with a music video directed by Werner Herzog?"  (That's a made up example, but someone originally gave me the story because Peter Gabriel and early Genesis are actually referred to in the text, and I was at the height of my fandom. XD )  An AI then comes up with a bunch of different examples, and the human who gave the prompt chooses the one(s) they like best.  They then release the creation to the broader world, and people make micropayments to stream it.  Everyone competes for attention, hoping to go viral or at least make a decent living.

(There's a dystopian aspect, where if you don't make enough money and your balance drops below zero, you disappear back into the human factory to get remade.  Also, people don't have sex in person -- they pay each other for the rights to their likeness, and they have sex with simulated versions of one another.  All of which is rather interesting, but not as directly relevant to the point I'm making here.)

M.A. Foster did an impressive job foreseeing a bunch of aspects of modern AI and online culture (especially keeping in mind that there was no Web or social media or digital streaming or online micropayments at the time this was written).   And it’s becoming easy to imagine that we may reach a point where many of the story’s predictions about art come true, as well.  

Currently, you can give increasingly complex prompts and get AIs to respond with something that makes sense and seems like a valid reply. Newer AIs often create text and images that are both exciting and terrifying due to what feels like a sudden potential to blend in with or replace human output.  Fandom, along with everyone else, is unnerved.  After all, will we still need fan creators in a world where we can prompt AIs to do this?

(source)

Or this?

(source)

Or when AI can even take the prompt “Professor Charles Xavier and Erik Lehnsherr kissing next to a conflagration” and output this?

(source)

Okay, there’s definitely still some funkiness going on there with some body parts; cherik fanartists can still obviously do better, for the moment.  And more generally, AI output is still frequently goofy, full of embellishments and fabrications, literally tasteless, and/or flat-out wrong – but we can see so many promises of how good it will get.  And we can imagine how AI output will increasingly be incorporated into transformative fandom and shipping culture.  

But all of the above examples illustrate another thing that M.A. Foster got right: If you want AI to produce something really interesting and compelling, it's important to have a human come up with a good prompt and then select the best output. 

This isn’t new; for years, we have been living in an increasingly curatorial world.  For instance, with stock photos of nearly everything, and digitized versions available of much of the world's art and photography, and endless hours of new YouTube & TikTok videos uploaded to the web every passing minute, it's much easier these days to create new images or videos or other visual works without being an artist.  But building a compelling visual work based on others' images -- a mood board, edit, collage, fanvid, etc. -- still benefits enormously from being driven by a human with a particular sense of style and particular goals in mind.  And the more that any human wants to see something that is different from the most common or most popular images that already exist, the more likely they are either going to have to create it themselves – or at least push the AI really hard in that direction via increasingly specific prompts and feedback.  (None of these roles are unique to online culture, either – art commissioners have historically prompted things, and art collectors and museums have curated them -- but these days we all have access to a much wider world of online works, and we all curate our own tumblrs and pinterest boards and so forth, even if we don't explicitly create curatorial works for fandom.)

The thing I found most unrealistic about "Entertainment" was that people weren’t tempted to try their own hand at creating art; it was a purely remix + curation culture.  In reality, even if AIs get really excellent at creation, so good that their fic and art are as good as your favorite fan creators’ work, I don't think they're ever going to suppress our own creative urges.  We live in a world where there are already 313 Dean/Castiel high school AU hurt/comfort fics – and yet people were still inspired to write/update two more this week.  People are not going to stop creating new fanworks just because the AIs are increasingly able to join in and create more.  And, for some time yet, humans are still going to lead the way in creating new canons with compelling stories and characters, which machines will then learn from and remix.  (That is another thing that human artists have also always done -- drawing inspiration from and remixing one another's art -- and something that fandom in particular is pretty great at.)

TL;DR human contributions to fandom will still be very important for fandom for the foreseeable future.  Even if the internet -- and now AI -- have helped us shift from spending more time as solo creators to also having increasingly active roles as prompters and curators.  

---

This post was partly inspired by @fansplaining 's latest discussion of AI & recent fandom panic, "Artificial Fandom Intelligence", as well as @cfiesler 's post, "Elon Musk did not create an AI trained on your fanfiction." They also addressed other issues that fans are worried about, like the idea of AIs and their creators getting credit and/or monetary reward for new fanworks trained on existing human-generated fanworks. If you've read other good meta about any aspect of fandom & AI, I'd love pointers -- please feel free to share in the notes!

*At least, it was collected in a 1985 anthology called Owl Time; I’m not sure when/where it was originally published.

[ranidspace]

"ai generated greentext this ai generated greentext that"

Objectively funnier, asking the AI smash or pass:

[wtf-fun-factss]

GPT-3 managed to outperform UCLA students when it comes to reasoning skills. Click to read the full fact.

[pink-ellephant]

Holy shit holy shit holy shit GPT3 is fucking blowing my mind. This AI is better at coding than I am. I just asked a stackoverflow style question about wrangling some data I’m currently analyzing for a paper, and it gave me a beautiful, elegant, ANNOTATED, well-explained tidyverse solution in R.

This is absolutely incredible. I am completely floored. Do you know how much this is going to democratize software development?! Anyone who has taken an intro class in data structures and processing and can speak the language of online help forums can get this thing to write code that reliably does what they want it to do. This makes evaluation in computer science courses harder, but think of the intelligent, creative people who don’t have the resources or opportunities to take those courses but can watch some lectures on YouTube and learn to use ChatGPT. This is going to change the world. I really hope we can make sure it’s for the better.

[cephalopistol]

trainboy's epic adventure