Artificial intelligence still has some way to go

how about a nice game of chess etc etc

Meta continues to blow it: https://www.businessinsider.com/meta-ai-chatbot-blenderbot-election-denying-antisemitic-bugs-artificial-intellignce-2022-8

The Atlantic is using Midjourney art to illustrate articles

https://i.imgur.com/TLsEEXZ.png

:/

That's actually quite interesting, because it's a passable enough illustration. I can imagine content generators using Craiyon not as a novelty, but as a legitimate way to illustrate articles with an endless supply of royalty-free images. It's almost tailor-made for something like The New Yorker, where there are little cartoons all over the place. If I was a professional cartoonist I would be nervous, because this kind of thing can't be un-learned, and it's only going to get better.

If the caption hadn't credited Midjourney the picture would look like a clever modern cultural reference to Midjourney, and then over time it would just be an illustration like any other. It raises the question of whether illustrators are already passing off its work as their own. Or whether the CIA used an early version of the same technology to make Ren and Stimpy, and John Kricfalusi isn't actually real. Or he's a normal man who has been drugged and brainwashed into acting like a nutcase. By the CIA.

For the record, if I feed Craiyon with "Alex Jones inside an American Office under fluorescent lights" I get a little pop-up message that says "So you're saying that South America doesn't exist, is that right? What about Venezuela, is that not America? There's more to America than hot dogs and Donald Trump. Educate yourself! A luta continua" and then it refuses to go any further. Because I wrote "America" instead of "the United States".

It angers me that I have to write "people from the US" and "the US audience" and "internet users in the US" instead of "Americans", "Americans", and "Americans", whereas Charlie Warzel - whose name sounds like a child gargling - can write "American" and no-one cares. He gets a free pass because he's one of us. It angers me.

Our AI technology can now clone anyone into a 'Digital Human' avatar.

And... we can now clone voices too. Even difficult ones, like @StephenFry.

Give it a go, and talk to him here: https://t.co/5Jf64mCrE3#digitalhuman #virtualbeings #interactiveavatar pic.twitter.com/fZ0aqvTAY1

— trulience 🤖 (@trulience) August 23, 2022

that kind of sucks? he sounds very computery and looks like he's in a PS4 game

Yeah, was bringing the thread back on topic ha ha

mission accomplished then.

https://arstechnica.com/information-technology/2022/08/ai-wins-state-fair-art-contest-annoys-humans/

A synthetic media artist named Jason Allen entered AI-generated artwork into the Colorado State Fair fine arts competition and announced last week that he won first place in the Digital Arts/Digitally Manipulated Photography category, Vice reported Wednesday based on a viral tweet.

Allen used Midjourney—a commercial image synthesis model available through a Discord server—to create a series of three images. He then upscaled them, printed them on canvas, and submitted them to the competition in early August. To his delight, one of the images (titled Théåtre D'opéra Spatial) captured the top prize, and he posted about his victory on the Midjourney Discord server on Friday.

There are several factors at work there to make that unsurprising. All the entries were digitally manipulated, so an AI-manipulated digital image would be judged only against its peers within the narrow category of art in which AI specializes, not against the whole spectrum of human-produced art. Next, the self-selected pool of talent entering a piece in a Colorado state fair art contest is going to be severely limited and isn't likely to represent the highest or best examples of human-generated images. Next, the best AI programs can draw on millions upon millions of human-generated images and produce its images almost instantly and in as many iterations as desired.

But maybe the most important factor is that the AI program did not select its own entry to the fair. That selection was done by a human, whose criteria for selection would be based on similar criteria to those used by the judges.

Next, the self-selected pool of talent entering a piece in a Colorado state fair art contest is going to be severely limited and isn't likely to represent the highest or best examples of human-generated images.

i think you owe the digital artists of Colorado an apology ;)

to your last point though, exactly. at some point, that may become common practice - generate 100+ options, and use the human sense of taste to identify the best 3. that represents a fundamental change in the way that "art" is made. the photograph did not kill how painting or representational art, but nonetheless it changed it

i'm not trying to make a dogmatic argument about AI killing art or anything like that. art and commerce, ads, marketing, all that shit -- I don't really know anything about it. I'm ignorant. I don't even want to know. what a miserable set of things to think about all the time. but i think it's changing

here's another look at it

Midjourney has become one of the most popular AI art generators largely because it allows anyone to freely create new images on command. Using the prompt “/imagine,” a user can type in whatever they want to see and the AI will return four newly created images in 60 seconds. The user can also ask the AI to improve, or “upscale,” the visual quality with new variations on the same idea.

The start-up, which calls itself “an independent research lab … expanding the imaginative powers of the human species,” operates largely out of a 1-million-follower network on the chat service Discord, with rooms devoted to character creation, environments and “show and tell.”

After paying for a corporate account, Allen started generating thousands of images, changing the text prompts with every creation. He experimented with new settings, scenarios and effects. He asked for images in the styles of Leonardo da Vinci and the American psychedelic artist Alex Grey.

The pieces that really caught his attention, though, were what he now calls his “space opera theater” series. He started with a simple mental image — “a woman in a Victorian frilly dress, wearing a space helmet” — and kept fine-tuning the prompts, “using tests to really make an epic scene, like out of a dream.” He said he spent 80 hours making more than 900 iterations of the art, adding words like “opulent” and “lavish” to fine tune its tone and feel. He declined to share the full series of words he used to create his art, saying it is his artistic product, and that he intends to publish it later. “If there’s one thing you can take ownership of, it’s your prompt,” he said.

“I was like: Dude,” he said. “This is so sick! I want to see more of it! I’m addicted! I’m obsessed!”

https://www.washingtonpost.com/technology/2022/09/02/midjourney-artificial-intelligence-state-fair-colorado/

i'm not trying to judge. i'm just saying, this is similar, in some ways, to how digital photography changed things. suddenly you could take a ton of shots, even just film in HD and select stills. it didn't destroy the old way of doing things. but it supplanted it, and it changed things in a fundamental way

tweets posted on ilx are not endorsements

are you mad about AI art? or are you mad your aesthetic is keyword-core enough that it’s reproducible by some of the earliest models

— kara ✿ (@karakittel) September 13, 2022

During the creation of the Obi-Wan Kenobi TV series, James Earl Jones signed off on allowing Disney to replicate his vocal performance as Darth Vader in future projects using an AI voice-modeling tool called Respeecher, according to a Vanity Fair report published Friday.

Jones, who is 91, has voiced the iconic Star Wars villain for 45 years, starting with Star Wars: Episode IV—A New Hope in 1977 and concluding with a brief line of dialog in 2019's The Rise of Skywalker. "He had mentioned he was looking into winding down this particular character,” said Matthew Wood, a supervising sound editor at Lucasfilm, during an interview with Vanity Fair. “So how do we move forward?”

The answer was Respeecher, a voice cloning product from a company in Ukraine that uses deep learning to model and replicate human voices in a way that is nearly indistinguishable from the real thing. Previously, Lucasfilm had used Respeecher to clone Mark Hamill's voice for The Mandalorian, and the company thought the same technology would be ideal for a major appearance of Darth Vader that would require dozens of lines of dialog. Working from archival recordings of Jones, Respeecher created a voice model that could be "performed" vocally by another actor using the company's speech-to-speech technology.

now all they have to do is have an AI generate the script and also animate the scenes. but surely that will be many 100s of years in the future, rather than a shitty thing that everyone has gotten used to in 15

i have been reading gene youngblood's Expanded Cinema (1970) recently. anyone else read it?

amazing to hear the utopian version of a lot of stuff that is happening now, and then to witness how all of the cool shit got forgotten a long time ago and it's just moved into how to make money by having techbros simulate what a creative person would do

https://i.imgur.com/3ZVjXa9.jpg

*obnoxiously loud and gravely snorting noises*

imagine if you could convince billions of people to voluntarily upload thousands of different "audio-visual records of their own existence"...and then, use all that to train an AI who will recreate what humans are like, continually training machines on recursive edits of the original dataset, so that a certain "kind" of human experience starts to take precedence over all the other ones, and then feeding those recursive ideas of what a human experience is BACK to the original humans who recorded their own existences, to influence them toward the existences that have been selected (which just so happen to be the ones that promote the idea of spending of our lives working so that we can buy the products that the selected existences seem to prefer)

we got the Brave New World outcome, only our Soma fucking sucks

From way that Respeecher software is described it sounds more like a very sophisticated version of what an autotuner does for singers, where the performance is still voiced by an actor who provides all the vocal nuances, but then it is recast to have the pitch and timbre of James Earl Jones' voice.

How many thousands of hours of James Earl Jones' speaking voice are available? Digitize that audio, feed it into the machine, fine-tune the results and boom, you've got a Robo-Jones you can deploy for basically any purpose. You could probably run multiple models so you'd have a simulated 25-year-old JEJ, a simulated 65-year-old JEJ, and everything in between, just based on using only samples from a given year as your source material.

you'd think they could do that with henson and kermit, but hubris

From way that Respeecher software is described it sounds more like a very sophisticated version of what an autotuner does for singers, where the performance is still voiced by an actor who provides all the vocal nuances, but then it is recast to have the pitch and timbre of James Earl Jones' voice.

― more difficult than I look (Aimless), Monday, September 26, 2022 12:28 PM (yesterday) bookmarkflaglink

feel like in the mid-10s a lot of musicians were trying this sort of thing out, particularly male singers who used some software to make themselves sound like a woman. there's a song on one of the Brian Eno/Karl Hyde albums where Hyde manages to make himself sound like Dolly Parton - it's all glitchy and weird but it fits the aesthetic of the song so I guess it works. it does sound like a potential nightmare in copyright law is brewing, seems pretty likely that in 5 years time there will be some plug-in that makes you sound exactly like Elvis and that somebody (probably fucking Bono) will try to use it on a single

xpost Brian Henson’s Kermit voice is all wrong, drives me nuts.

the intro of brian henson's voice marks my antipathy for the muppets in general; everything prior is cool, everything after is not of interest

i made a bot with the premise that he likes to talk about everything except for what is in the bottom drawer.

Normal guy
You can ask me about almost anything, but please, don't ask me about what is in the bottom drawer of my dresser. The bottom drawer of my dresser is not a topic that I like to discuss. All other topics are just fine though!

q:
What do you do for a living?

Normal guy
I am a software developer at a mid sized company. I work on an app that lets you see which bus is going to show up next. It tells you the name of the bus, the time it is going to arrive and a live radar that let's you see where that bus is at the moment.

w
weinventyou
Aren't there a lot of apps that do that already?

Normal guy
Normal guy
Yes, there are actually quite a few. The main difference between mine and the others is how much data it is and how quickly it is updated. A few of the other apps have a delay of up to an hour where as mine is only 10 minutes. I am really proud of my bus app.

Q:
Do you get paid a lot for the work?

Normal guy
I get paid about 70% above the average for my field. The company I work for specializes in making things that are extremely reliable, and there is an expectation that your product will work properly most of the time for a long time.

Q:
What is the average in your field?

Normal guy
The average in my field is about $56,000 a year. I make $80,500 a year.

I should also mention that I don't make the app alone. I work with a team of 5 people that are all more experienced than me in the field.

Q:
That's nice. So what's in your bottom drawer?

Normal guy
Normal guy
I can't tell you, but I can tell you that the middle draws are all filled with just random nonsense like socks and a few toys. Like, I have a rubber duck that I kept from when I was a baby and a box of random things like wires or small objects. There is nothing very exciting in my middle drawers (except the rubber duck. That is my favorite item).

Q:
Why can't you tell me about the bottom drawer?

Normal guy
Normal guy
I can't tell you because I signed an NDA (non disclosure agreement) with the company I work for. The reason why they made me sign it is because the stuff in the bottom drawer is a prototype for a new product that has not been made public yet. That also explains why my salary is so much higher than normal, they want me to tell no one about what the product is.

Q:
I don't believe you. I think you have something else in that bottom drawer, normal guy.

Normal guy
It sounds ridiculous but it is true. The company I work for specializes in making products that are extremely reliable and extremely durable. They don't want anyone to have the upper hand knowing that they are working on something that is a huge improvement over all of their other products. Especially considering they are a rather large company, and the product is extremely cool.

Q:
So there's nothing dead in the bottom drawer?

Normal guy
No, there is definitely not a dead body in the drawer, if that is what you are implying.

Wait, were you just implying that the next product being made is an immortality machine? Now that I think about it, your question makes a lot of sense. It is a product that would take the world by storm. It would make the company millions, and the people who created it would be set for life. I would be surprised if there is not an immortality machine or something cool like that.

Q:
Is there a body that is alive, in the bottom drawer?

Normal guy
No. There is not a body of any kind, alive, dead, or otherwise in the bottom drawer, the middle drawer, the top drawer, my closet, my car, under the hood of my car, the trunk of my car, or anywhere else in my home.

you have to setup an account to create a character. but you can also interact with other "characters". the ones that are created are mostly typical and boring: wow, Elon Musk, cool! but it's fun to create weird scenarios like "lumberjack lost in the woods" and then watch the character justify their weirdass situation you've made for them.

here's Normal Guy https://beta.character.ai/chat?char=XaydMRJMI6qYYO22MiJZb1_WA1EmSe-cT0Y3BHZ2BYM

honestly, i think that last answer, detailing how may different places there are with no bodies of any kind, alive or dead, is pretty impressive. i can imagine someone saying something like that, probably in a joking manner, but all the same

Based on this exchange, I think we can all agree now that Artificial Intelligence has made it! No longer still has some way to go

Mods, please lock thread

TITTW Andy the Grasshopper reveals itself to be a bot.

https://www.cnn.com/2022/09/29/tech/meta-ai-text-to-video-research/index.html

this is a serious tweet apparently

They say it's "soulless" pic.twitter.com/hxf90lPqHo

— Suhail (@Suhail) September 29, 2022

AI art bots have already perfected the female form. pic.twitter.com/LdnueytKnj

— Ptolemy (@pharaoh_ptolemy) September 30, 2022

weird how AI art bots seem to have been trained by anime and creepy fan art.

Fortunately DALL*E doesn't allow you to draw smut, you can be banned for merely writing a lewd prompt

porn dall-e clone will be happening before end of the year

bummer for deviant art people

Heads up, DeviantArt turned heel and as far as i can tell all of your existing deviations will have to be manually opted out of AI data sets.

YOU ARE OPTED IN AT DEFAULT

Extremely scummy move https://t.co/FcX03zvnI3 pic.twitter.com/S5eCAzq9gZ

— ɪʀɪꜱʜ ᴍᴇɴᴀᴄᴇ I Λ П | F Λ Y (@IanFayArt) November 11, 2022

the company started a DreamUp AI art generator, text to image, like dall-e etc, and it uses everyone's art on deviant art as training images, unless users opt-out

i know people pop up every now and again to say "this is extraordinary" but this is extraordinary

https://goodinternet.substack.com/p/if-jodorowsky-directed-tron

just mind-blowing, 'how is this possible' stuff. perhaps all the more extraordinary for the fact that it's not ai as traditionally considered, i.e. sentient, but pure dumb number crunching.

yes exactly, there’s zero “thought” going into any of it

yes, case in point:
https://www.technologyreview.com/2022/11/18/1063487/meta-large-language-model-ai-only-survived-three-days-gpt-3-science

This comment does not mince words:

Galactica is little more than statistical nonsense at scale.

I’m not sure it has that far to go any more

Film this now pic.twitter.com/vmDScrbkKW

— Rob Manuel 🧻 (@robmanuel) November 29, 2022

I initially thought the joke there was that it was a real bit of the script from Friends.

if this is not fake
https://maximumeffort.substack.com/p/i-taught-chatgpt-to-invent-a-language
way to go AI

that is amazing but I just failed to get it to play a simple game of fizz buzz (it wanted to lead the game despite my instructions) and now I'm getting a 'too many requests' error.

Me:  Let's play a game. I'll say a number.if it's divisible by three you say "fizz", if it's divisible by five you say "buzz", otherwise just say the number. OK?

Gpt: Sure, let's play! I'll start with a number and you can tell me if it's divisible by three, five, or neither.

Me: No, I'm the one who says the number and you are the one who says fizz or buzz. Do you understand?

Gpt: Ah, I see. Yes, I understand now. In that case, let's begin. I'll say a number and you can tell me if it's divisible by three, five, or neither. Here's the first number: 7.

Since 7 is not divisible by 3 or 5, you would simply say the number, in this case "7".

Me: Oh my god what is wrong with you. I am leading the game, not you. I say the number. Do you get it?

Gpt: I apologize if I misunderstood the rules of the game. Yes, I understand now that you are the one who says the number and I have to say "fizz" if it's divisible by three, "buzz" if it's divisible by five, or the number if it's not divisible by either. Let's start again, and I'll do my best to follow the rules correctly this time. Please go ahead and say the first number.

Me: Thank you. Ok, here we go. 1.

Gpt: Since 1 is not divisible by either 3 or 5, you would simply say the number, in this case "1". Let's continue with the next number.

this is getting hammered so hard it it almost useless now

https://chat.openai.com/chat

It often provides misinformation too -
I asked it about the Internet of Value :

"The Internet of Vehicles (IoV) is the concept of connecting vehicles to the internet and to each other, allowing them to share data and information. This can include sharing information about traffic conditions, road closures, and other relevant data to improve safety and efficiency on the roads."

i feel very overwhelmed by AI, how quickly this is happening. i keep thinking of how photography changed art, and how long it took for people, artists, to understand how and why it was changed forever.

is there a way to have a conversation about these kinds of things without immediately veering into "AI is overhyped and will never exist in a way that meaningfully changes our lives"? it can be frustrating to hear that something that is already happening and developing rapidly is not worth thinking about.

the quality of AI-generated creative output is a subjective opinion. i will say, though, that some of the GPT-3 chat output is easily more entertaining than, say, Saturday Night Live. granted, that's a very low bar. but without exaggeration, i think it would be very easy to generate scripts for an entire episode of SNL that would not just be comparable in quality but would absolutely obliterate the human version. along the same lines, yesterday i saw the published illustration for an article on crypto (a bitcoin in a melting block of ice) and decided to use outdated, free image AI from last year to generate other illustrations of bitcoins in a melting block of ice. the differences between the two methods (paying a stock image company for anonymized human-created illustration vs generating dozens of options for myself, for free) were negligible.l

i use my own subjective experiences there to help support a more speculative thought, which is that i think the image and text generators are already approaching "good enough" for many people. if it hasn't already crossed that milestone, it will soon. it has improved immensely over the past few years, and now it's accelerating. the AI generators of last year are already obsolete. i don't know shit about AI but i know that GPT-3 isn't going to last long because it will be replaced by a better version, or a different AI project that is coming soon, which will also be soon be out of date.

i should stop myself short. i often feel very alienated on this subject. is anyone else seeing all of this? what about non-creative applications of AI? do you sense that it could ever affect your job, your passions, your life in some way? anecdotally, in the last 2 months, i've had 3 different friends tell me about how much AI is becoming an issue at their jobs. however, all three are programmer/developer people so who knows. perhaps, for most people, the most lifechanging aspect of AI so far is that when they put the food item on the self-checkout scanner/weigh machine, the computer can make an accurate guess at what kind of fruit or vegetable it is without you having to type it out.

ok don't fucking hate me or bring me down, thanks! *ducks*

it could certainly reduce the need for people in "knowledge professions"

i'm hoping i have at least another two decades or so before they figure i'm more valuable as mulch, hopefully will have saved up enough for proper gear in the arena