SWNFT
NFT Education

Generative Art vs AI Art on the Blockchain

Generative art follows rules an artist wrote. AI art learns from data. See how each works on the blockchain, and how Art Blocks mints work.

James Killick7 min read

TL;DR

  • Generative art is made by rules the artist writes in code. AI art is made by a trained model that learned patterns from lots of images.
  • Art Blocks is a platform where the artist's code sits on the blockchain. It runs when you mint. Each mint creates one output.
  • Code-made art can be rebuilt exactly from its code. Model-made art usually can't be rebuilt from the token alone.
  • Many projects mix both. Read what a project says about how each piece is made.
  • NFTs are high risk. Most lose value, and this post is about how art is made, not what to buy.

If you're curious about NFTs and AI, you've seen "generative art" and "AI art" used like they mean the same thing. They don't. Generative art follows rules an artist wrote in code. AI art comes from a model that learned patterns from many images. Both can end up as NFTs, but they work very differently.

Quick risk note first. Most NFTs lose value, and scams are common. This is education, not a buying guide.

What is generative art?

Generative art is art where the artist hands some of the decisions to a system. Sotheby's describes it as work made with algorithmic code or mathematical formulas. The artist gives some or all decision-making to an autonomous system. It says the genre began in earnest in the 1950s (Sotheby's, June 2023).

So it is much older than modern AI image tools. The artist writes the rules: shapes, colour palettes, how things move. Add a random starting number and the same rules give a different picture each time. For the basics, see generative NFT art explained.

What is AI art?

AI art comes from a machine learning model. The model is trained on a large set of images, then makes new ones. The artist does not write the shape rules. The model works them out from the data.

An early famous example is a portrait made by Paris collective Obvious. An algorithm called a GAN made it. The GAN was trained on 15,000 portraits painted between the 14th and 20th centuries. Christie's sold it in October 2018, as Time reported at the time.

Today's AI NFT work often uses text-to-image tools. The artist types a prompt, then picks from the results. What are AI NFTs covers that side.

What is the difference between generative art and AI art?

Here are the practical differences:

| | Generative (code-made) | AI (model-made) | |---|---|---| | Who sets the look | The artist, through written rules | A trained model, shaped by its data | | Same input, same output? | Yes, the same seed gives the same piece | Not always. It depends on the model and tool | | What you can inspect | The script | Usually only the finished image | | Typical home on-chain | Script stored on the chain | Image file stored elsewhere, token on the chain | | Big question | Is the code good? | What was the model trained on? |

The line between them blurs. Sotheby's treats AI as one tool inside the wider history of generative art. It isn't the definition of it. Many artists use code and a model in the same piece. So ask a simple question: what part of the work is a rule, and what part is a learned pattern?

What is Art Blocks?

Art Blocks is a platform for generative art on Ethereum. Its docs say it launched in November 2020. Its first project was Chromie Squiggle by Snowfro. Artists write scripts, and the docs say these are stored on the Ethereum blockchain.

Sotheby's calls it a vital trading platform and notes that Dmitri Cherniak's Ringers were released there. The best-known project is Fidenza by Tyler Hobbs. It's a series of 999 works built on a flow field algorithm. It sold out in minutes (Sotheby's again).

Hobbs describes the base of Fidenza in his own write-up, published June 2021. It's a flow field algorithm. Think of it as invisible currents that lines follow across the canvas.

How does an Art Blocks mint work?

The Art Blocks docs explain it in two steps. When a collector mints, a unique hash is generated on-chain and fed into the artist's script. The same hash always makes the same output. A hash is a long string of letters and numbers that acts like a seed.

That has a big effect. Hobbs wrote in August 2021 that the output goes straight to the collector. The artist has no chance to curate or cherry-pick. Nobody knows in advance what a mint will make. He calls this "long-form generative art". He says artists must make sure the worst output is still good, because bad ones are as visible as good ones (The Rise of Long-Form Generative Art).

Compare that with a weekly AI pick like Botto. There, a community chooses one winner from many. Same blockchain, very different way of choosing.

Where does the art actually live?

This matters for how long a piece lasts. Sotheby's notes Autoglyphs by Larva Labs sit in "the very limited space available on the blockchain". Most NFTs link to outside files instead. That's the strongest form of on-chain art.

Model-made art works differently. A model runs on someone's computer. The finished image is then minted. The token proves who owns it, and the file usually sits on outside storage like IPFS. I'd say that's a fair general pattern, but check each project. For how ownership is recorded, see how NFT ownership works on-chain.

How can you tell which one you're looking at?

Use this short checklist before you read the hype:

  1. Find the project's own description of how each piece is made.
  2. Look for a script or a "how it works" page. Generative projects usually name the algorithm.
  3. Look for model names or tool names. AI projects usually say which ones.
  4. Ask whether the artist curated the outputs or the code did.
  5. Check whether the project says anything about what the AI was trained on.

If a project won't answer these, treat that as a warning sign. How to read NFT rarity is a good next skill. Rarity works differently in code-made sets.

Frequently Asked Questions

Is Art Blocks AI art?

No. Art Blocks hosts generative art made by artists' scripts. The script runs when you mint. Some artists may use AI in their own work. The platform itself is about code-made art.

Is Fidenza generative art or AI art?

Generative art. Hobbs says its main basis is a flow field algorithm, a rule system he wrote. It is code-made.

Can generative art be AI art too?

Yes, they can overlap. Sotheby's treats AI as one tool within generative art's longer history. Always check how a specific piece was made.

Who owns the copyright in either kind?

Owning the NFT does not automatically mean owning the copyright. That depends on the licence and the law. See what rights you own with an NFT.

This is education, not financial advice. NFTs are high risk, and most lose value.

Next step: pick one Art Blocks project and one AI NFT project. Use the checklist above to write down how each piece is made. Start with what AI crypto means if you want the wider picture.

About the Author

James Killick

AI Product, Web3 & Sales Automation Specialist

James Killick runs this site →

Digital Product Development

Devwiz is a Sydney-based digital product development agency specialising in launching and scaling SaaS products, web apps and mobile apps. James is a shareholder and advisor, shaping product strategy and PLG approach.

Visit Devwiz

// Related

Keep Reading