Vitalik Buterin Titok AI supports onchain image storage

Vitalik Buterin Titok AI supports onchain image storage


Ethereum founder Vitalik Buterin has endorsed the new Token for Image Tokenizer (TiTok) compression method for the blockchain implementation.

Not to be confused with social media platform TikTok, the new TikTok compression method significantly reduces image size, making it more practical to store on the blockchain.

“320 bits is essentially a hash,” Buterin said on decentralized social media platform Farcaster, highlighting Tito's blockchain capabilities. There is a bit that goes on the chain for each user.

The development could have significant implications for digital image storage profile pictures (PFPs) and non-perishable tokens (NFTs).

okex
Source: Thomas

Related: Traders: Ethereum Is ‘Most Bullish Altcoin' As ETH Returns $3.5K

Titok image compressor

Developed by researchers at ByteDance and the Technical University of Munich, Titok allows compression in 32 small pieces of data (bits) without reducing image quality.

According to Titok's research paper, advanced artificial intelligence (AI) image compression allows Titok to compress a 256×256 pixel image into “32 discrete tokens”.

Titock is a 1-dimensional (1D) image simulation framework that “breaks the grid limitations in 2D simulation methods” leading to dynamic and compact images.

“As a result, it brings a high speed in the sampling process (for example, 410 × faster than DT-XL) and obtaining a competitive quality of generations.”

689a0e6f 5019 45fa 9441 d52e02893ad9
Titok's research paper showing comparisons of image compression rates. Source: Titok

Related: Spot Ethereum ETFs May Begin Trading on July 2 – Bloomberg Analyst

Machine learning images

TiTalk uses machine learning and advanced AI using transformer-based models to transform images into tokenized representations.

The method uses regional redundancy, which means it uses redundant information in different regions of the image and reduces the overall data size of the final product.

“Recent advances in generative models have highlighted the critical role of image simulation in the efficient synthesis of high-resolution images.”

According to the research paper, Titok's “compact latent representation” can produce “more efficient and effective representations than conventional techniques.”

0e92708f 0ab1 4733 8ede 54d0fefe3ed3
An example of image reconstruction (a) and generation (b) with Titok framework (c). Source: Titok

Related: Vitalik Buterin-backed Nocturne protocol shuts down operations overnight

It's Tik Tok, not Tik Tok.

Despite the similar name, Tik Tok, a social media platform, did not receive support from Buterin.

Ethereum's founder gives credit to the new AI-based image compression method by highlighting Tito's blockchain capabilities.

“Unlike existing 2D VQ models that consider the hidden space of an image as a 2D grid, we propose a more compact formula for mapping an image to a 1D hidden sequence.”

The proposed new method can “represent an image 8 to 64 times” with smaller tokens than “2D tokenizers”, and the team hopes the research will shed light on “more efficient image representation”.

Magazine: $1M Bet ChatGPT Won't Lead to AGI, Apple's Use of Intelligent AI, AI Millionaires on the Rise: AI Eye

Leave a Reply

Pin It on Pinterest