The Svalbard archipelago is a two-and-a-half-hour flight from the northern coast of Norway, about half-way to the North Pole. Roughly 3,000 people live in the region, which is also home to arctic foxes, polar bears, dolphins, seals and a unique, unusually small subspecies of reindeer.
Svalbard also has permafrost – ground that remains below freezing. And that’s only one reason why it’s an excellent place to store things underground. “It’s a point that’s so far north, that if you want to watch the northern lights, you have to look towards the south,” says Ben Kovalis, cofounder of art.ai, which is engaged in a mix of artificial intelligence and digital art.
Kovalis and his partner, Eyal Fisher, returned recently from a visit to Svalbard. At first glance, such a silent and frozen place near the end of the earth couldn’t be further away from the frenetic internet world in which the pair work – a sector in which the talk is about blockchain, Web 3.0 and NFTs. But Fisher and Kovalis were there for work.
That’s because Svalbard is home to an institution called the Artic World Archive, whose declared purpose, as stated on the website of the company that runs it, is to be the most secure archive for the world’s digital memory and to ensure that those files of art, culture, history and literature are well preserved for the coming generations.
Among the items digitally stored in the archive are Dante’s “Divine Comedy,” books by Olga Tokarczuk, a Polish laureate of the Nobel Prize for literature and the painting “The Scream,” by Norwegian artist Edvard Munch. And as of mid-March, due to the importance that the archive places on blockchain technology, it has begun storing select collections at no charge of NFTs – non-fungible tokens stored on blockchain – at no charge. That’s Kovalis’ and Fisher’s connection to the place.
Something totally different
Kovalis and Fisher are only 30 but they’ve known each other for 26 years. They were childhood friends who grew up together in the Tel Aviv suburb of Hod Hasharon. As adults, Kovalis became involved in sales and marketing in the high-tech sector while Fisher earned a PhD in statistical genetics and began working in the field of statistics and artificial intelligence.
- The NFT marketplace born in an ultra-Orthodox school in Israel
- Caring for your CryptoKitties has never been easier, thanks to these Israelis
- New president Herzog presented with digital copy of father's signed oath
“We wanted to start a company together, and we waited for the right idea,” Kovalis explained. The idea surfaced in 2018. A few months earlier, the portrait “Edmond de Belamy,” which was produced by a French art collective, became the first such art to be produced by artificial intelligence that was offered for sale at a public auction. It was sold in October of that year at Christie’s in New York for $432,500.
“It wasn’t a particularly attractive portrait,” Kovalis acknowledges. “It was very pixelated and not high quality. We’ve always been attracted to art and Eyal, who specialized in visual image processing, thought we could develop an algorithm ourselves and make better works with it through artificial intelligence. The idea of creating a startup dealing with cool and fun things was attractive to us. The minute Eyal mentioned the idea, from my standpoint, it was ‘Let’s go.’”
They linked up with a third partner, Guy Haimovich, another childhood friend, who founded the company with them. (Haimovich left the company late last year). At its initial stage, in 2019, Art AI, opened an online store on Shopify and sold paintings generated by artificial intelligence for about $100 a piece. “Every painting was unique and was just sold once. We would put new paintings online every day and we sold to several thousand customers, who would receive the painting as a canvas print,” Kovalis explained.
In the meantime, the field of NFTS was coming into being. Succinctly put, NFT can be described as ownership of a digital asset. It’s not digital currency, like bitcoin for example, which can be exchanged for other bitcoin without it having any particular significance. NFT, on the other hand, is a unique asset, such as an image, sound clip, video clip or a character in a virtual game. As such it is non-fungible, and each NFT - or non-fundbile token - is a unique digital file that unlike other tokens (like digital coins) cannot be broken down into smaller units, thus allowing the “original” file to be bought and sold in very much the same way a painting can be sold to art collectors.
Last year the NFT sector exploded: Initially few people even knew what it was, but by the end of the year, people had spent nearly $41 billion on NFTs.
“We had heard about the potential of the field, and we knew that it was worth our getting into it, but we also knew that if we went in, it needed to be something entirely new,” Kovalis said. “For anyone on the outside, this entire market might sound like one big mess, but one of the cool things about it is that people in it really love innovation and anyone who produces innovation is compensated accordingly.”
Instead of creating paintings and then selling them, at Art AI, they built an interface that enabled people to produce art themselves through artificial intelligence and then turn it into an NFT. “We built a machine that’s capable of creating an unlimited number of works and to make use of various algorithms to create different kinds of art,” he said.
The company has so far raised $1.8 million from angel investors, Kovalis said, including the developer and investor Rafi Gidron, who is the chairman of the company, and Liron Rose. The NFT field captured their interest.
“This world is developing at an unusually fast pace. Being on top of things requires having a finger on the pulse all the time, reading many hours a day and listening to a lot of Spaces on Twitter,” Kovalis added, referring to live conversations on a specific topic on Twitter.
Kovalis and Fisher were not the first to spot that NFTs would be the next big thing. Venture capital funds have been investing in them and celebrities have been showing off NFTs, increasingly a status symbol, or are even creating their own NFT collections. Meanwhile, thousands of digital creators have tried to get some foothold in the field and to make a name themselves - though few have managed to gain as much prominence as Kovalis and Fisher have done with their project, which is called Eponym. (Following its success, the pair also created the Epolabs brand for their work in the NFT space).
Linking art and artificial intelligence has also engendered reservations from art lovers who question what value of artificial intelligence can really have in the art world and whether works of art generated by AI can even be considered art. From the standpoint of the NFT world, the debate looks different, Kovalis explains.
“In October, we came out with our NFT interface. The aim was to enable people to create an NFT for themselves using artificial intelligence and to permit them to get a taste of how it feels to use artificial intelligence to create art. The project was enthusiastically received and prompted people in the community to debate the philosophical aspects of the use of artificial intelligence to promote art, and what we as private individuals can do to advance the new internet and the metaverse. People fell in love with this right away,” he said.
That affection comes at a price. According to Kovalis, people are prepared to pay a lot to be part of what’s considered pioneering technology in the NFT world. Eponym was actually an NFT collection created overnight by 10,000 people - each of whom created a work using artificial intelligence and paid around $300 for the service.
How was participation in the project carried out?
“Every participant went into our interface and was asked to write text: a word, a sentence, a quote from a movie or anything else that inspired them – up to 66 words. They could also write instructions about what to show in the painting or ask for a specific style. And then with the click of a button, within half a minute a picture is created based on the text entered in the interface.
“Within a few hours, we sold all 10,000 in the collection for a sum of nearly $3 million. The revenue from the project didn’t end there, since the works that are created through the project continued to be traded after that – so far through secondary transactions worth $20 million, and a certain portion of this revenue has also gone to the company. Some aspects of the NFT market are like the art market but on steroids - particularly when it comes to trading. When people see something that they believe will increase in value, instead of holding onto it for 10 years like in the regular art market, they may hold onto it for just days or even hours until its value increases.”
A single corrupted bit
The success of the Eponym project is what led Kovalis and Fisher after several months to visit Spitsbergen Island, the largest island in the Svalbard archipelago. This group of islands became part of Norway as a result of the Svalbard Treaty of 1920, which provided that despite Norwegian sovereignty, the area would be permanently demilitarized and the signatory countries to the treaty would be able to conduct economic activity there if they wished. What was anticipated at the time was mainly the mining of coal, which is considered of particularly high quality in the region.
The fact that the place is always frozen, as well as being remote and demilitarized, makes it safer than most other places around the world to store items that people wish to preserve. In addition, the archive is in the depths of a coal mine – a dark, dry and low-oxygen location, which makes it ideal for storage. The archive is ultimately a business, run by the Norwegian tech firm Piql and Norway’s government coal-mining company.
Kovalis says that after the Eponym project was launched, he was contacted by a successful NFT firm, Top Dog Studios. “They decided to focus on projects that advance the NFT field and help people understand what’s involved. They talked about their connection with the Arctic archive and said that they wanted to join forces with several innovative projects in the field to store [their output] in the archive. I was immediately taken with the idea.”
The reason for the archives is the fundamental problem involved with the digital age – the difficulty in storing digital files securely over time. As Maranke Wieringa, a researcher from Utrecht University in the Netherlands, described it in a 2017 article in the journal Junctions, the preservation of digital information is a problem whose scale and severity has grown with the rise of digital media. Although it may seem that anything can be backed up in the cloud or on a memory stick, the file itself can actually disintegrate, Wieringa claimed. “Whereas non-digital media such as paper degrade slowly, a single corrupted bit can irrevocably harm a digital file,” she wrote.
Another problem is the rapid development of software, hardware and file formats that make them obsolete and complicate their future use.
Kovalis recounted that after several months of discussions between the archive and Top Dog Studios, the moment that they were waiting for arrived – the trip to Svalbard, where digital copies of the NFT collections that were selected would be buried deep underground. “The archive is about 300 meters [1,000 feet] deep in the coal mine, which is under a snow-covered mountain,” he remarked. “We put on hats with flashlights on them and walked a length of track until we got to the archive.”
At the cave-like archive, they held a short ceremony and placed well-wrapped photographic film into a safe there. The works stored in the archive are on special film made of particularly resilient material that is supposed to be able to last for at least 500 years – and in the conditions in the mine for up to 2,000 years. The film includes photos of the works or documents themselves or a QR code where the information is stored.
It’s at extremely high resolution and contains massive amounts of data. The film includes all the necessary information to decipher the encoded material so that people in the future will be able to access it.
Kovalis recounts that at the entrance to the archive are the flags of the dozens of countries from which material is stored. “We are the first Israelis to put something in the safe, and now they will be adding the Israeli flag, and for us, that’s a great honor. From the standpoint of our artistic activity, the Arctic archive is preserving a historic night in which 10,000 people created works of art at the same time. It’s exciting to think that that moment will be preserved for thousands of years.”
So how does it work? Kovalis explains what happens when two sets of artificial intelligence compete with one another: Art.ai’s technology includes the development of two sets of artificial intelligence that function as two rival entities (a model called GAN for short). Kovalis suggests one can be thought of as the artist and the other as the art critic.
Their technology, he explains, works as such: The Critic algorithm is exposed to millions of works of art so it’s able to identify a work of art and in a sense judge its quality. The Artist algorithm is not exposed to the art, but it learns - or teaches itself - how to create art on its own. Initially the Artist’s work will get a low grade from the critic since these are mere abstract scribbles that are in no way similar to the works of art that the Critic has been exposed to and is trained to identify and rate.
But over time, through an iterative process between the two, the Artist’s paintings will improve, and it will be harder for the Critic to decipher them from those done by real flesh-and-blood artists. In fact, the two sets of artificial intelligence will continue to improve each other since the Critic will continue to improve to discover the paintings done by the Artist and the Artist will continue to try to deceive the Critic so that the Critic will think that its work was done by humans.
Additional artificial intelligence being developed at the company is capable of comparing the Artist’s picture with millions of pre-labeled pictures on the internet and to determine whether their content is in keeping with other online pictures – meaning whether they comply with the demands that the user has inputted in the interface. If not, it can issue instructions that will help the Artist algorithm create a picture that is more suitable.
Kovalis’ and Fisher’s startup is not the only one to generate such works using artificial intelligence through text, but Kovalis claims that Art AI has an advantage over the other companies – and that’s the hundreds of thousands internet users who have visited the company’s online gallery over the past two years.
“People have wandered through the gallery, ‘liked’ specific works, lingered at others or have bought them. We’ve taken all these interactions and taught the algorithm about what people like to look at and what they like to buy so that it’s connected to human sentiments.”