Grayscale’s Decentralized AI Fund Targets Blockchain Solutions for Deepfake Challenges

Key Takeaways

• Grayscale has launched a new Decentralized AI Fund.
• The fund holds Filecoin (FIL), Livepeer (LPT), Near (NEAR) and Render (RNDR).
• The asset manager said it is investing in “solutions to centralized AI-related problems, including authenticity checks against bots, deep fakes, and misinformation.”

Responding to the rise of AI threats like deepfakes and social media bots, technologists have proposed various solutions to the AI authentication problem, i.e. how to distinguish the human from artificially-generated. 

One approach has been to embrace blockchains and other decentralized networks to power verification systems that don’t have the innate vulnerabilities of centralized databases.

To help fuel this revolution, on Wednesday, July 17, the digital asset investment firm Grayscale launched a new Decentralized AI Fund that will invest in a basket of prominent AI tokens.

How Blockchains Can Solve AI Challenges

In a press release  announcing the new fund, Grayscale distinguished different categories of decentralized AI assets. Perhaps the most intriguing of these it defined as: “Protocols building solutions to centralized AI-related problems, including authenticity checks against bots, deep fakes, and misinformation.”

The asset manager’s emphasis on “AI-related problems,” hints at the capacity for technologies like Bittensor to address the societal challenges posed by modern AI.

Among the most ambitious AI blockchain platforms, Bittensor recasts various forms of compute, data and storage, as well as finished AI models and applications as digital commodities. Through interconnected “subnets,” Bittensor envisages a decentralized AI economy that spans every stage of development and distribution.

For example, the BitMind Subnet  (currently testnet 168) is designed as a decentralized deepfake detector, allowing users to verify the authenticity of images without having to rely on centralized technology providers.

Grayscale Backs Decentralized AI

Decentralized AI is an umbrella term that covers a variety of distributed solutions for AI training and deployment.

The two tokens with the greatest weighting in Grayscale’s new fund are Filecoin and NEAR. At inception, Filecoin received 30.59% of the fund’s allocation while NEAR received 32.99%.

Launched in 2014, Filecoin is a blockchain-based data hosting solution that lets users rent disk space from a decentralized network of over 3,000 storage providers in exchange for FIL, the platform’s native currency. While it can be used to store any kind of digital data, Filecoin is increasingly used by AI developers who don’t want to rely on Big Tech cloud providers.

Meanwhile, NEAR is a smart contract platform that has emerged as a popular AI development stack thanks to its cross-chain interoperability and early emphasis  on machine learning requirements.

Alongside Filecoin and NEAR, Grayscale has invested in the decentralized GPU network Render, allocating 24.86% of the fund’s capital to RNDR. The fund also holds smaller amounts of Bittensor (TAO) and Livepeer (LPT). 

CCN reached out to Grayscale for

Trust and Content Provenance

Ultimately, blockchain’s biggest contribution to the AI market is trust. Or rather, trustlessness. 

From Bitcoin onwards, blockchains have enabled trustless interactions between decentralized networks of peers by removing the need for centralized intermediaries who might be able to manipulate the flow of information. 

An example can be observed in content provenance, which has become increasingly important as AI-generated media has evolved to the point where it is often indistinguishable from content created by humans. 

While the likes of OpenAI, Meta and Google are all developing their own digital watermarking solutions to identify AI-generated content, without a record of changes made to digital files, there is no way of knowing metadata, or even hidden watermarks haven’t been tampered with. 

To address this challenge, Livepeer has joined  the Coalition for Content Provenance and Authenticity (CCPA), integrating its open provenance standards into the Livepeer AI Subnet.

As a blockchain-based infrastructure layer for video streaming, Livepeer is ideally placed to flag media manipulation and increase transparency by recording the lifecycle of videos uploaded to the platform on-chain. By decentralizing the task of verifying content provenance, “history can’t be edited or modified at the whim of a single person who controls a database,” explained CEO Doug Petkanics in a recent podcast .

For Livepeer and the other protocols backed by Grayscale, trustless verifiability and censorship resistance are the biggest draws of blockchain-based decentralization, which could help empower AI builders and users and the technology evolves.