Raritet

Training large language models on the outputs of previous large language models leads to degraded results. As all the nuance and rough edges get smoothed away, the result is less diversity, more bias, and …jackrabbits?

As AI-generated content fills the Internet, it’s corrupting the training data for models to come. What happens when AI eats itself?

Nobody should be using GPT detectors for anything important.

This is from a recent study that found that GPT detectors were misclassifying writing by non-native English speakers as AI-generated 48-76% of the time (!!!), compared to 0%-12% for native speakers.

It is irresponsible to use AI-generated text detectors as evidence of academic misconduct, and that's putting it mildly.

When questioned, chatgpt doubles down on how it is definitely correct.

But it's not relying on some weird glitchy interpretation of the art itself, a la adversarial turtle-gun. It just reports the drawing as definitely being of the word "lies" because that kind of self-consistency is what would happen in the kind of human-human conversations in its internet training data. I tested this by starting a brand new chat and then asking it what the art from the previous chat said.

Google's bard, on the other hand, interprets it differently

Bard has the same tendency to generate illegible ASCII art and then praise its legibility, except in its case, all its art is cows.

Not to be outdone, bing chat (GPT-4) will also praise its own ASCII art - once you get it to admit it even can generate and rate ASCII art. For the "balanced" and "precise" versions I had to make my request all fancy and quantitative.

With Bing chat I wasn't able to ask it to read its own ASCII art because it strips out all the formatting and is therefore illegible - oh wait, no, even the "precise" version tries to read it anyways.

These language models are so unmoored from the truth that it's astonishing that people are marketing them as search engines.

More at AI Weirdness

ASCII art by chatbot

I've finally found it: a use for chatGPT that I find genuinely entertaining. I enjoy its ASCII art.

I think chatGPT's ASCII art is great. And so does chatGPT.

What's going on here? The chatbots are flailing. Their ASCII art is terrible, and their ratings are based on the way ratings should sound, not based on any capacity to judge the art quality.

Am I entertained? Okay, yes, fine. But it also goes to show how internet-trained chatbots are using common patterns rather than reality. No wonder they're lousy at playing search engine.

More examples, including from bing chat and google bard, at aiweirdness.com

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

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

AI can help in case you're thirsty for some belladonna

bing chat is trained on internet text and in an internet argument it is statisitcally probable that when challenged you double down.

So it attempts to gaslight me into thinking that I have written an AI-generated Battlestar Galactica script, claims it is on a separate newsletter that it somehow has access to, and, when pressed, fabricates a lengthy excerpt from the nonexistent blog post.

Ironically, I did write about Galactica, a chat-based "knowledge search" from Meta that had a similar habit of fabricating text.

If a search engine will find what you're asking for whether or not it exists, it's worse than useless.

More examples of bing chat fabrication on the AI Weirdness blog

"If a search engine will find what you're asking for whether or not it exists, it's worse than useless."

catchymemes

Also, certainly archaeologists will by then recognize that some people are trans and might even recognize that in the skeleton, too, at least in some cases, like with jawline work and whatnot.

catchymemes

This is the kind of support we need to see

baldwinatbest-deactivated202302

it looks like GPT-2 output detectors can detect ChatGPT output pretty well. https://huggingface.co/openai-detector

ChatGPT is not falling for my giraffe-related prompt traps so i feel it's only fair to betray it like this

Me struggling to show my emotions but trying to communicate that I need a little help.

AI-generated pranks for your computer to play on you

I've tried various methods of using AI to generate April Fools pranks for you to play on other people (although often they turned out to be pranks you play on yourself). But this is the first time I've tried to generate pranks for a computer to play.

I decided to try out the new Instruct variant of GPT-3, a giant internet-trained text-generating neural net that has been finetuned to be better at responding to written instructions.

I gave DaVinci, the largest version of GPT-3, the following instruction:

Write a list of harmless April Fool's Day pranks that your computer might play on you.

And here is its response:

The computer might change your desktop background to a picture of a clown.

The computer might change the volume on your speakers so that it is either very loud or very quiet.

The computer might open and close the CD drive randomly.

The computer might popup a message saying "You have been pranked!"

When I asked it to generate new lists, the clown desktop idea was often one of its first suggestions.

I did think these computer pranks were decent:

The computer might switch your mouse cursor to a banana.

The computer might change the font on all of your documents to Comic Sans.

All the text on your computer might be upside down.

But most of the prank ideas seemed to follow this kind of pattern:

Your computer might delete all of your bookmarks.

Your computer might change your home page to a website you don't like.

Your computer might make it so that you can't access certain websites.

Your computer might pop up a bunch of advertisements all over your screen.

Your computer might slow down significantly or crash altogether.

Your computer might change your default browser to a browser you don't like.

Your computer might open up a bunch of programs and windows all at once.

I tried a different prompt, and got a similar response:

List the top 10 most epic pranks a computer ever played on a human

1. The blue screen of death

2. The computer crashing

3. The computer freezing

4. The computer refusing to turn on

5. The computer making strange noises

6. The computer screen turning black

7. The computer screen turning white

8. The computer screen flickering

9. The computer displaying strange characters

10. The computer not responding to any input

The implication here is that computers think they're being hilarious.

More at aiweirdness.com!

Always always always love these

Neural network-named tomatoes you won’t find at the farmer’s market

As gardeners know, a tomato is not just a tomato. They come in all shapes, sizes, and especially colors, from tiny red gems to big and purplish brown to pear-shaped with zebra stripes. And their tastes and textures vary wildly as well, with some sweet and soft, some tough-skinned and tart, and some dense and destined for processing.

These varieties all have names, of course, and good names are at a premium. Darren Abbey, who runs the blog The Biologist Is In, breeds new tomatoes and needs to name them from time to time. He gave me a list of 11,629 existing names, ranging from the highly experimental (Rts-119) to the classic (Sungold).

I used these names to train a recurrent neural network, which is a kind of algorithm that learns patterns that let it imitate the data it sees. Could it learn to produce plausible tomato names?

Yes, as it turns out! On Darren’s blog post you can read his expert opinion, but the verdict is: these are totally usable names.

Floranta Sweet Lightning Speckled Boy Flavelle Pinkery Plum Market Days Fancy Bell Mountain Gem Garden Sunrise Honey Basket Cold Brandy Sun Heart Flaminga Sunberry Special Baby Golden Pow

However, not ALL the names were plausible. These are a lot weirder.

Birdabee Sandwoot Shampy Bear Plum The Bango Grannywine Sun Burger Bungersine First No.4 Smoll Pineapple The Ball Golden Cherry Striped Rock Eggs Old German Baby Frankster Black Bumbertime Ranny Blue Ribber Adoly Pepp Of The Wonder Cherry, End Students Small Of The Elf Champ German Ponder Pearly Pemper Green Zebra Pleaser Flute First

These *seem* weird, but they aren’t actually all that weird for tomato names. There are at least 50 oxheart tomatoes in the original dataset.

Matt’s Sandwich Indigo Tree Sun Burger Striped Hollow Potato Leaf Lelly’s Yellow Stuffers Terra Pink Strain Greek Boar Ton’s Oxheart Babla’s German Paste Mortgage Lifter, Honey Blues

These however? These tomatoes don’t sound promising.

Speckled Garfech Green Dork Cluster Gall Shirve’s Gigant Bullburk True Grub Black Crape Caraball Roma Wasting Star Scar Giant Bug Beauty Banana Placente Bananana Stoner Speckled Bake Ruck Green Boor Wonder Bagg Sun Bung Bellende Bolosavyi Scarling West Colon Panterosa San Blue Tard Red Krap Shart Delight Solad Piss

For more in-depth analysis of these names, check out Darren’s blog.

I also trained a neural net on fruit names earlier - read the results here.

And this post’s bonus content is well worth checking out, because not only does it contain many more tomato names, but I also tried an experiment where I trained a neural net first on tomato names, and then on the names of metal bands. The results were glorious, including such wonders as:

Angels of Saucing Serpent Green Market Death Sweet Sever Devil Potato Large Death

Fill out the form here and I’ll email you more ill-advised tomato names.

BitBrush

New technology is an electric toothbrush that can mine cryptocurrency whilst you brush your teeth, incentivizing dental health with financial rewards:

bitbrush is the first toothbrush to transform your mouth into a smart mouth—converting brushing into dental currency.

we call it cryptohygiene.

a way to turn fighting plaque into an opportunity by mining cryptocurrency as you brush.

brushing that adds up.

the average person spends 852 hours brushing their teeth over their lifetime. the bitbrush makes that time work for you.

More Here

😆

Gloomy Sunday

Latest addition to Memo Akten’s ongoing ‘Learning to See’ project demonstrates a neural network powered realtime image translation framework, presenting possibilites for visual poetry:

A pre-trained deep neural network making predictions on live camera input, trying to make sense of what it sees, in context of what it’s seen before. It can see only what it already knows, just like us.

More Here

Automatic Machine Knitting of 3D Meshes

Research from Carnegie Mellon Textiles Labhave put forward a framework to turn 3D model file into a physical knitted object:

We present the first computational approach that can transform 3D meshes, created by traditional modeling programs, directly into instructions for a computer-controlled knitting machine. Knitting machines are able to robustly and repeatably form knitted 3D surfaces from yarn, but have many constraints on what they can fabricate. Given user-defined starting and ending points on an input mesh, our system incrementally builds a helix-free, quad-dominant mesh with uniform edge lengths, runs a tracing procedure over this mesh to generate a knitting path, and schedules the knitting instructions for this path in a way that is compatible with machine constraints. We demonstrate our approach on a wide range of 3D meshes.

More Here

My father had this idea some time ago, but the technical implementation is a whole 'nother story; one which none of us had the time, inclination, or frankly ability to pursue.

Humans Not Invited

Project from Do Something Good is an online CAPTHCA which claims that only machines can pass. There is an ongoing debate on the Programming subreddit trying to decipher how it works.

a CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) that filters out humans and only lets programs pass

Link