Original title: Foresight Ventures: AI x Crypto Report
Original author: Foresight Ventures
Compiled by: Alvis, Mars Finance
Key Insights
Foresight Ventures believes that AI x Crypto may be the most exciting industry in the next 50 years. Crypto enhances AI by solving centralized control, data privacy and security issues, making AI more trustworthy and efficient. AI can give crypto intelligent functions, enable smart contracts to make AI-driven decisions, enable blockchain to perceive the physical world, and improve the user experience of blockchain applications.
The intersection of AI and cryptocurrency can be divided into three layers: infrastructure, models, and applications. Innovations at the infrastructure layer, such as decentralized computing power and data storage, will reduce costs while improving the efficiency and security of AI applications, bypassing traditional cloud providers. The model layer will enable peer-to-peer model inference networks, promote collaboration and innovation in the field of AI, reduce costs, and allow small businesses to participate in the development of AI. In addition, AI-driven smart contracts will enhance blockchain applications such as biometric authentication, fraud detection, and AI trading robots. Finally, the application layer will integrate the infrastructure layer and model layer into various consumer products.
The infrastructure and model layers of AI x Crypto are already relatively mature, but application development is still in its infancy. There are significant opportunities to expand the application layer to leverage the developed infrastructure and models.
Al x Crypto brings many new research areas with great research potential, such as ZKML, FHE-ML, distributed ML training, network quantization, decentralized databases, fully homomorphic encryption, Al hardware, and FHE hardware.
Preface
“Blockchain revolutionizes production, while AI changes production processes. AI x cryptocurrency is likely to be the most exciting industry in the next 50 years.”
Forest (Co-founder of Foresight Ventures)
AI x Crypto Introduction and Overview
AI x Crypto projects have shown strong asset returns over the past two years. The intersection of AI and cryptocurrency has significant value. Cryptocurrency allows us to trust AI, and AI makes blockchain smarter.
The AI x Crypto system can be divided into three layers: infrastructure layer, model layer, and application layer.
The infrastructure layer provides the computing and storage capabilities required to efficiently execute models and applications. Many projects such as io.net, Akash, APUS, PingPong, etc. belong to the infrastructure layer computing power category. Projects such as 0G, Glacier, and SpaceAndTime provide decentralized data storage for AI.
The model layer involves the algorithms and models used in AI systems. Model networks build peer-to-peer reasoning networks or create new base models through community contributions. Projects such as Bittensor, PIN AI, Cerbo AI, and Sentient belong to this category. On-chain AI is a subset of model networks, except that its reasoning results can be used in smart contracts, allowing them to use AI for decision making. Projects such as Ora, TheoriqAI, Nesa, and Modulus are advancing on-chain AI. They may use infrastructure layer compute and storage providers to power these models.
AI applications like MyShell, Story Protocol, and Sleepless AI package the lower levels into a cohesive consumer product.
AI x Encryption Infrastructure Layer
1. Decentralized computing power projects: collect GPUs and CPUs from underutilized sources such as independent data centers, crypto miners, and hardware networks such as Filecoin and Render. They provide decentralized and cost-effective computing power, bypassing cloud oligopolies such as AWS. Decentralized computing power addresses the growth in GPU demand driven by AI and provides competitive pricing advantages. For example, an A100 on AWS costs about $4.10 per hour, while io.net charges $0.76 per hour. Although decentralized computing generally provides lower quality than large data centers, this situation is starting to change.
2. Decentralized database/file system/DA layer: Provide data storage for decentralized AI, enabling people to own, control, and utilize AI assets in a decentralized manner. Decentralized storage systems provide cost-effective storage solutions and promote collaborative AI development environments. The key design challenges of decentralized storage systems include economic models, storage verification algorithms, degree of decentralization, and improving retrieval performance. These four factors are interrelated and require trade-offs.
AI x Encryption Model Layer
1. Model Networks: Build peer-to-peer model reasoning networks to incentivize the development of machine intelligence. Model Networks promote openness and collaboration in AI. They improve the quality of AI product outputs by labeling AI models and agents. They also help reduce AI costs and encourage the development of small businesses.
2. On-chain AI Model Network: Enable smart contracts to use AI for decision making, giving blockchain the ability to perceive the physical world. The AI Model Network uses ZKML (zero-knowledge machine learning), OPML (optimistic machine learning), AI oracles, and AIVM (artificial intelligence virtual machine) to ensure that AI reasoning is both trustworthy and accessible in blockchain smart contracts. Users can use AI to make decisions in smart contracts. For example, they can participate in DeFi projects and earn income based on AI predictions, or use AI for facial recognition to enable smart contracts to recognize individuals.
AI x Encryption Application Layer
1. Data Collection Dapps/Apps: Use labeling to earn money and data/bandwidth contribution to earn money mechanisms to collect data efficiently and economically while protecting the rights of contributors. The data collection and labeling market is huge. Data collection applications use crypto incentives to create a trustless global labor market, thereby reducing costs and enabling global participation. They also use gamification to make labeling tasks easier and more fun.
2. AI Dapps: Package the infrastructure and model layers into a cohesive consumer product. Crypto AI platforms use asset tokenization to help AI creators address funding challenges and increase profitability. They establish a sharing economy for AI, create opportunities for new business models, and provide users with a safer, more privacy-focused experience. There are many types of AI dApps and applications, including AI creator platforms, AI-enhanced social, gaming, and entertainment applications, AI applications for prediction, trading, AMMs, and DeFi, and a variety of AI tools.
Conclusion, future trends and social impact
In summary, driven by a massive influx of capital and rapid technological advancement, generative AI will drive significant business transformation over the next 50 years. Cryptography enhances AI by addressing issues of centralized control, data privacy, and security, making AI more trustworthy and efficient. The intersection of AI x cryptography is still in its early stages and holds significant value. The infrastructure and model layers of AI x cryptography are relatively mature, but application development is still in its infancy. Significant opportunities exist to extend the application layer to leverage the infrastructure and models that have been developed.
Future trends in the AI x Crypto space include several key developments:
Decentralized infrastructure: Innovations in decentralized computing power and data storage will reduce costs and increase the efficiency and security of AI applications, bypassing traditional cloud providers.
Model Network: A peer-to-peer model inference network will foster collaboration and innovation in the AI field, reduce costs and enable small businesses to participate in AI advances.
On-chain AI: AI-driven smart contracts will enhance blockchain applications including biometric authentication, fraud detection, and AI trading bots.
Data Collection Dapps: The integration of AI and encryption will make data collection more efficient and cost-effective, with projects focusing on gamification of data labeling and enhancing user rights protection.
AI Dapps/Apps: New AI-driven decentralized applications will emerge in various fields, providing enhanced security, privacy, and user experience through blockchain integration.
Regulatory and ethical considerations: In the evolving AI x Crypto space, clear regulations and best practices are critical to foster innovation while protecting user rights and data privacy.
The social impact of AI x Crypto in the next 50 years could be huge. Blockchain reshapes the way production is done, while AI changes the production process. We eagerly anticipate the unprecedented explosive growth that will occur when these two disruptive innovations collide. We envision a future where AI will have enormous power that is not suitable for centralized control. This urgently requires the establishment of decentralized networks to manage larger AI systems, starting with computing power and data, and gradually expanding to algorithms and applications.
Blockchain reshapes production methods: Blockchain technology improves the transparency and efficiency of asset management and transfer through distributed ledgers and smart contracts. For example, blockchain can be applied to supply chain management to ensure that every stage from production to sales is traceable.
Artificial intelligence reshapes production processes: The application of artificial intelligence in manufacturing, service and other fields has improved production efficiency and accuracy. AI-driven automation systems can run around the clock, reduce human errors and optimize resource allocation.
Decentralized governance of AI: Decentralized networks use distributed consensus mechanisms and transparent governance models, which have the potential to mitigate the risks of AI centralization. Similar to blockchain, decentralized AI networks can establish transparent and fair governance structures through smart contracts to prevent any single entity from exerting excessive control over AI systems.
