In August 2022, a Japanese developer named Nao_u released a simple scrolling shooter, in which all the art was made by the Midjouney neural network. He had to work with the settings, and achieved impressive results. With the development of neural networks people started talking about how this technology might replace artists in the future. In this article, we will figure out whether this is actually the case, find out how neural networks work, and discuss if it can be integrated in crypto.
How neural networks work?
For now neural networks are very far from the human brain and cannot make truly creative tasks. However, they already have some benefits. Effectiveness of such services directly depends on computing power, so the development of neurons has become possible relatively recently. As the name of the technology implies, such networks imitate the work of neurons and synaptic connections in the human brain. The more connections in a neuron, the more efficiently it works. And this requires a lot of power, so the use of neurons is an expensive thing.
Today most of the art neural networks use a public domain license, so their pictures can be freely used in business. You can either generate art for your T-shirt or use it in your game – no one will say a word. Basically, neural network services are stepping into the field of Pinterest and other aggregators, which have always been the best sources of references.
This, for example, is what the neural.love service is aiming for – to become a photo stock, where everyone can find anything they want. Neural network services publish the work of their "robots" mainly in the public domain, and so users can browse through the archives and take what they need.
There is a popular myth that neural networks collect their work from the art available on the Web, but this is not true. They draw it themselves, turning the words into a picture based on what they have learned. A human artist knows what a forest looks like. He has been there, seen it in photographs and other art. The neural network is exactly the same. It saw a forest in the works of other artists, so it can draw it. In order for the neural network to understand what to create, the user must enter keywords – prompt.
Can neural networks replace artists?
Neural networks are like a new tool for the artist. A way to quickly get a set of necessary images in order to work with them. Now there’s no need to study image aggregators for inspiration, people might just use any of the neural network services – it’s faster and requires less effort.
Also neural networks help visualize ideas very fast. This will be useful not only for artists, but also for game designers. With the help of the generated image, it is easier to show the style and the atmosphere that you want for your project.
Neural networks and crypto
Many have tried to use neural networks to forecast the Bitcoin price. For example, in 2017 Ognjen Gatalo developed a Twitter bot. This service should post Bitcoin predictions every two hours using historical data. The technology was based on blockchain.info API (today blockchain.com) – there the bot took price data in order to predict the future. However, this bot wasn’t working correctly, it was impossible to make a forecast with the only historical data only.
There were plenty of other attempts to come up with a solution. In 2020 Federico Riveroll developed a neural network that could predict Bitcoin price based on the historical data (Bitcoin\USD) and news titles from Fox Business News. The algorithm was looking for connections between keywords in news titles and Bitcoin price. That still wasn’t enough to make accurate forecasts. But the algorithm has shown that it’s possible to predict the price if it uses more data.
Final thoughts
As we can see, neural networks made a huge step forward in the art sphere. Sadly, they’re not so advanced in the Crypto World. Yet. Developers are working on services that can help us predict crypto prices. They have already made relatively accurate algorithms, but they are still unsuitable for long-term forecasting. High volatility, unpredictability and lack of data complicate developers' work.
However, everything could change very soon. As modern art neural networks show us – everything is possible. It just needs time to grow into something fascinating. The more effort researchers put into collecting data and better understanding the market, the sooner tools for price prediction will appear.
Would you like to learn more about crypto? Then you should check out our blog! You might like our articles “Cryptocurrency And ArtLogan Paul’s NFT game, was it all a puff?
