The Most Interesting Companies At The Intersection Of Crypto And AI

With its technological capabilities accelerating everyday, AI feels capable of overcoming almost any obstacle. But those at the bleeding-edge of AI research will tell you that the technology has inherent blindspots. Some innovative companies have been searching for solutions in an unlikely arena: blockchain, the decentralized and permissionless verification technology which makes cryptocurrencies and NFTs possible by facilitating automatic, trustworthy, peer-to-peer asset stewardship and transaction.

This week we saw the announcement of the Superintelligent Alliance token merge: Fetch.ai FET , Ocean and SingularityNET AGIX proposed to merge their tokens into a new one, $ASI. This merger, if approved by the community vote, would mean that the alliance would have a "fully diluted market capitalization of $7.6 billion across 2.631 billion tokens", according to Fetch.ai’s blogpost.

The interest of venture capitalists in the intersection of AI and blockchain remained strong in 2022-23, despite a broader market slowdown affecting global VC funding. Coinfund, a crypto-native VC fund invested in Gensyn, decentralized compute for Machine Learning, and Giza, decentralized Machine Learning inference computation. Paradigm, another prominent crypto firm has invested in Numer.ai, an AI powered and crowdsourced hedge fund.

Here are four companies in four different sectors which creatively marry AI and Blockchain technologies:

Generated Images And Artworks: Kojii.ai

With an unsustainably high amount of artworks made everyday using AI engines (MidJourney, Civitai, etc.), AI artwork is developing a public stigma similar to that of NFTs after the hyper-speculation of 2021. NFTs, however, are nothing more than media files (jpg, png, mp4, etc.) that can be transferred and stored verifiably using Blockchain technology.

Kojii.ai seeks to marry customizable Generative AI technologies and NFTs to elevate their collective public perception. The platform enlists artists to train (or “data-set,” used as a verb) unique AI models tailored to a specific aesthetic or thematic vision.

Artists design bespoke UI/UX interfaces where users can manipulate AI outputs with weighted sliders, toggles, and drop-down menus. Users generate images from the model with the click of a button (think OpenAI’s art-generation program, Dall-E, with overarching conceptual frameworks and far more customization). While generating images is free, by purchasing credits that can be used to tokenize select images, users “immortalize any generated image(s) as non-fungible tokens (NFTs) on the Avalanche AVAX blockchain,” which can then be traded, sold, and stored.

Artists are paid via this subscription-style credit system, and kojii.ai users can co-monetize the imagery they create. This project establishes a new paradigm for AI artwork, one which rewards artist and user creativity symbiotically.

Compute Marketplace For AI: PrimeIntellect

Training and deploying AI models requires massive amounts of computing power. This computing power (known as “compute”) comes exclusively from powerful chips called Graphics Processing Units, or GPUs. Owing to both singular technical excellence and manufacturing capacity, chip-maker Nvidia has a near stranglehold on the GPU market, and their products are priced accordingly. Many smaller AI research firms and open-source AI developers decry this near monopoly.

Blockchain technology allows peer-to-peer transactions between individuals irregardless of geographic locales, language barriers, Fiat currency denominations, and other external factors. AI research firm PrimeIntellect takes advantage of this capacity by allowing individuals and groups to sell access to their compute in a decentralized, permissionless marketplace. PrimeIntellect describes their business model as enabling “the training of large AI models using distributed resources, reducing costs and democratizing AI development.”

To access compute from anywhere in the world, on-demand, in a low-communication and trustworthy environment is to challenge the traditional computing infrastructure itself.

Open-Source AI Research: Nous Research

While multinational corporations like OpenAI, Microsoft MSFT , and Google GOOG all have massive in-house teams responsible for improving and innovating their proprietary AIs, it is open-source actors who are oftentimes responsible for the greatest advancements in any new tech. Open-source developers have traditionally worked for free, though their products can regularly disrupt monopolistic industries.

Nous Research, holds daily competitions for open-source developers to create AI models capable of “beating” OpenAI’s GPT-4, currently the gold-standard for public-facing AI models. This is judged via a complex, underlying system of automatic guidelines. Unlike platforms such as HuggingFace, which simply collect participants on a real-time leaderboard, Nousresearch rewards participants using Bittensor’s proprietary cryptocurrency, $TAO, based on their standings each day.

Using cryptocurrency as an automated reward both economizes and incentivizes the open-source developer community. As a handful of tech giants further ingrain their products into the AI market, the empowerment of open-source developers will prove vital for keeping costs down, maintaining innovation, and democratizing concerns over privacy, accessibility, etc. Nous Research’s crypto-and-competition-based business model could theoretically be applied to tech verticals the industry over, providing permissionless and powerful encouragement to developers with less resources.

Privacy And Security For AI-Based Oracles: ORA

As programmer and proclaimed “cypherpunk” Eric Hughes wrote in his landmark 1993 text, “A Cypherpunk’s Manifesto”: “Privacy is not secrecy… Privacy is the power to selectively reveal oneself to the world.” Much of the blockchain world operates off this underpinning. Blockchain transactions are deified because of their completely private nature. Not even identity needs to be disclosed. Certain necessities common in high-level blockchain transactions, however, potentially undermine this privacy.

Blockchains themselves operate via decentralized codes that run automatically, triggered by certain events or programmed to execute at certain times. These are known as “Smart Contracts,” and Smart Contracts govern every single blockchain transaction. However, in certain sectors (like Decentralized Finance, or “DeFi”), Smart Contracts are often prompted to interact with outside programs, receiving information from them (like real-time market and pricing updates), which is then used to execute transactions with accurate mathematical data. This outside program is called an “Oracle ORCL .”

Unfortunately, the calculations that an Oracle provides, and which are then used by the Smart Contract, must be physically confirmed by either external programs or outside developers. While this is usually harmless, the presence of these external programs allows external parties unpermitted access to one’s data, transactions, responses, and assets.

While the terminology is rather complicated, privacy-focused AI company Ora seems to have solved this issue. They’ve pioneered “zkOracle,” which is “a new type of decentralized oracle network that utilizes zero-knowledge proofs” to confirm calculations within Smart Contract transactions.

In other words, Ora uses secure, AI-based Oracles to interact with Smart Contracts, removing the need for outside actors and keeping data definitively safe within each transaction.

While the market sector of AI + crypto is at its infancy, the first solid trends are here: art, open-source AI, and oracles. These streams all require the power of blockchain for payments and immutable and trustless proof. When welded together, the two technologies are helping the nascent AI industry find and ways to remain creative, economically competitive, growth-focused, and reliably private.