The AI Invasion of Knitting and Crochet

Disclosure: I am not a knitter/crocheter. Though many of my loved ones are, and I turned to them for help writing this story, I obviously don’t know as much as those who participate in it. As such, mistakes may be made, and corrections will be posted as needed.

In August 2019, Researchers at MIT’s Computer Science & Artificial Intelligence Laboratory (CSAIL) debuted a new computer-aided knitting system. The idea was straightforward: you could take a photograph of a garment or even your hand, and the system would create a pattern that could then be sent to a knitting machine.

The process has been likened to 3D printing, and since then, several other systems have been developed to do something similar.

But, despite the promise these systems have shown, a lot has changed in the public availability of generative AI systems over the past six years. Since the public launch of ChatGPT in November 2022, AI systems have been able to generate almost anything. This includes text, images, videos and even knitting and crochet patterns.

However, this has created a significant problem for human knitters and crocheters. Many of their favorite sites for patterns are slowly being filled with AI-generated works. While some of these patterns may be acceptable, many either don’t match the images or are physically impossible.

So what can fiber artists do? Some are starting to fight back and call out these patterns as they see them. However, the simplest solution may be simple awareness and more people being more conscious about where they purchase patterns from.

Why AI Fails at Knitting and Crochet

On the surface, knitting and crochet patterns seem like the perfect task for AI to tackle. 

First, such patterns are already similar to a programming language in format and structure. They are comprised of a limited number of stitches that can be reordered to make just about anything. Much like a musician doesn’t come up with new notes, instead, they rearrange and reorder those notes to come up with new and novel sounds. Fiber artists find new ways to recombine the stitches they have to create new works.

Second, copyright protection in this space is limited. Though most knitting patterns contain some protectable elements, the instructions themselves are considered “procedures for doing, making or building things” which aren’t protected by copyright. Much like how cooking recipes don’t enjoy copyright protection, neither do fiber patterns.

Finally, most knitting and crochet patterns involve a lot of repetition and require a fair amount of math. These are things that, historically, computers have been great at. However, these are also areas where AI is struggling.

That’s because AI relies on probabilistic prediction rather than spatial logic. It is attempting to “guess” the next word or phrase rather than fully understanding what it is saying or doing.

This might be passable for responses to queries or even generating images, but it creates serious problems when creating a pattern. Creating a pattern requires considering the entire work; each step has to fit with and work with all the others. Blindly selecting the next step without that consideration will, more often than not, fail. This is especially true since AI can’t “test” the pattern after writing it, which is a big part of what humans do.

As such, AI is prone to hallucinating patterns that don’t work, are physically impossible, or don’t produce what is advertised. However, this hasn’t stopped shady sellers from turning to various sites to sell AI-generated patterns. 

This has become a growing problem for crafters, even on major sites.

The Fight Against AI

Two of the major sites for knitting and crochet patterns are Etsy and Ravelry. Etsy expressly permits AI-generated works as long as they are disclosed, and Ravelry doesn’t appear to have a set policy, but has been removing obviously AI-generated patterns. 

However, in both cases, there is the very real problem of patterns being published without disclosure. Unfortunately, many won’t realize that the pattern is AI-generated or impossible until after they purchase it. Given that most patterns cost well less than $20, few will bother to request a refund.

The most vulnerable are new and inexperienced fiber artists. They may miss the signs that a pattern is problematic and assume that it didn’t work because of their skill level. 

That said, there are a few things that buyers can do to reduce the chance of being duped by an AI pattern.

1. Look for Photos of People: Since most AI-generated patterns also use AI-generated sample images, look for pictures of the finished product on people, ideally, multiple people. AI still struggles in this space, and a pattern featuring humans in the photo is much less likely to be AI.
2. Read Reviews: Reviews will often warn that a pattern is AI-generated. However, people are frequently out of the habit of checking reviews for smaller purchases, so these can fly under the radar.
3. Look at Account Ages: If an account is older than 2022 and has posted patterns before then, they are unlikely to be using AI. Most AI accounts are newer and started out using AI, not accounts that started as human and switched.

However, the best and simplest advice is to buy from patternmakers that you trust. If you know someone who is a human making high-quality patterns, turn to them first. Rewarding known human creators rather than chasing the cheapest pattern is the best way to avoid buying AI slop.

Bottom Line

To be clear, much of this advice would be the same for cooks looking for recipes, builders looking for plans and so forth. Though the fiber arts community has been hit particularly hard, it’s primarily due to the rapid growth in the space and the relatively low entry point for new crafters.

Simply put, AI spammers see opportunities in this space and are exploiting it. Even if the major sites want to stop this, they may be powerless to do so.

However, the story is very representative of the general frustration with AI. As MIT research showed, there are real possibilities for AI to actually help human crafters. Instead, most find themselves dodging questionable AI-generated patterns on online storefronts.

While I feel that AI is dubious in and of itself, the way that it’s being used only amplifies the mistrust of the technology that so many feel.

It’s going to be a big ask to get crafters to support AI, especially when their entire experience has been scammy patterns produced for the sole purpose of turning a quick dollar. 

But this failing is as much a human one as a machine one. It belongs as much to the scammers who produce the patterns as to the AI systems that struggle to understand knitting or crochet. 

So, while it’s easy for me to blame AI for this problem, it’s important to remember that the humans behind the bots deserve at least an equal share of the blame.