The Phenomenal Rise of ChatGPT and Its Implications
In the realm of technological innovation, the rapid ascent of ChatGPT, an artificial intelligence (AI) chat application, has been nothing short of remarkable. Within a mere two months of its launch, it amassed a user base of 100 million by January 2023, marking it as the fastest-growing consumer application in history. This meteoric rise underscores the profound and transformative impact AI has already imprinted on our society. However, as we stand on the precipice of this new era, it is crucial to acknowledge that the evolution of AI technology must be paralleled by societal adaptations to balance its immense potential with the inherent risks it presents.
In this context, blockchain technology, with its inherent transparency, open-source ethos, and global permissionless development, emerges as a potential counterbalance to the societal risks posed by AI and related technologies. Although we are in the nascent stages of this technological convergence, a multitude of projects are already harnessing the power of AI and blockchain to revolutionize fields as diverse as machine learning and the arts.
A GLIMPSE AT WORLDCOIN
Worldcoin, a verification and identity-based protocol that raised $115 million in a Series C funding round in May 2023, is one such initiative. Co-founded by OpenAI CEO Sam Altman, Worldcoin's "proof of humanity" system uses retinal scans to authenticate human users, combating the increasing presence of bots online. The company is developing World ID, a digital identification system generated using a person's iris, with a focus on preserving privacy. World ID aims to verify personhood and the authenticity of online content or interactions through the blockchain, offering potential solutions to the proliferation of bots and deepfakes. Worldcoin and its token officially launched this month to considerable controversy mostly in regard to their VC-heavy fundraising strategy, (arguably) predatory tokenomics model, ethics surrounding their marketing, ability to protect users’ privacy, and more
Blockchain: A Potential Mitigator of AI Risks
the rapid adoption of generative AI models has also led to an increase in the production of falsified content. This is a growing concern that needs to be addressed to ensure the responsible and ethical use of AI technologies. But so why a blockchain to try and solve this problem?
The crux of the matter lies in the effectiveness, confirmability, and predictability of these cryptographic signatures and their proof of authenticity. At present, the detection of deep fakes largely hinges on machine learning algorithms, such as Meta's "Deepfake Detection Challenge," or Google's "Asymmetric Numeral Systems" (ANS). These algorithms strive to identify patterns and irregularities in visual content, but their accuracy is not foolproof and they are grappling to keep up with the escalating sophistication of deep fakes. Often, the task of assessing authenticity falls on human reviewers, a process that is both time-consuming and expensive.
Envision a world where every piece of content is accompanied by its own cryptographic signature, empowering everyone to conclusively prove the origin of creation and highlight any manipulation or falsification. This is the potential of blockchain technology—a bold new world where authenticity is not a matter of speculation but a verifiable truth.
Addressing the Challenge of Deepfakes with Blockchain Technology
In the United States, the prevalence of deepfakes has seen a staggering thirteen-fold increase, from 0.2% in 2022 to 2.6% in the first half of 2023. The Department of Homeland Security has warned that AI-generated deepfakes can incite violence, sabotage corporations, and manipulate elections, among other serious consequences. As the technology behind deep fakes becomes increasingly sophisticated, distinguishing between real and manipulated content has become a pressing concern.
In response to this threat, builders are investigating ways in which leveraging blockchain technology can establish a consensus on the authenticity of information, a feature intrinsic to blockchain technology.
Digital Signatures: Cryptography’s Solution to Deepfake Dilemma
One potential solution to this problem lies in the realm of cryptography, specifically, cryptographic digital signatures. By using a private key to sign content, the identity of the content creator can be verified. This signature, created using a private key known only to the creator, can be verified using a public key accessible to everyone. By attaching this signature to the content, it becomes possible to prove that the content was indeed created by the original creator, be it a human or an AI. This process ensures that the content has not been tampered with and that it originates from a verified source.
To fully grasp the potential of blockchain in thwarting deepfakes, it's crucial to understand the fundamentals of hashing algorithms, cryptographic signatures (the process of transforming data from a readable format, or plaintext, to a secure format, or ciphertext, and back to the readable format), and blockchain timestamping (a secure method of tracking the creation and modification time of a document).
While cryptographic algorithms and hashing are not exclusive to blockchain, these specific features of the technology are amplified when combined with other blockchain characteristics – such as its immutable nature.
Consider a cryptographic hash as a unique identifier for a specific content or file. For instance, the image below demonstrates how data passing through a cryptographic hashing algorithm takes a text input of any size and generates an output of a fixed length. This output is known as a hash. Anyone else with similar data can use that algorithm to produce the same hash, and comparing hashes can confirm that you share the same information. This can be instrumental in proving that a piece of text, file, or content has remained unaltered over time.
Establishing a Verifiable Provenance Record
Another approach to managing fake content involves establishing a verifiable provenance record. This record serves to differentiate between original and manipulated data. A joint initiative between Bundlr and Arweave is currently underway to develop a standard known as the Digital Content Provenance Record. This standard mandates that digital content and data assets adhering to its specifications include an immutable cryptographic signature provided by the content creator and a cryptographic timestamp recorded on-chain.
Arweave's permanent storage capabilities ensure the long-term persistence of the provenance record. At the same time, Bundlr's precise and millisecond-level time-stamping functionality establishes it as a trusted provenance layer for the content.
AI and the Issue of Centralized Power
GPU Shortage and Blockchain's Democratizing Role in AI Development
AI's capital-intensive nature poses another risk: the centralization of power among a few key players, which could limit alternative options for consumers. Network effects and economies of scale from owning massive data facilities could position a few winners in the AI market to gatekeep users and developers by charging high fees or creating user lock-in effects. These dynamics are already evident as increased demand for training machine learning models and computation requirements has led to a shortage of GPUs.
GPUs are integral to the operation of machine learning models, facilitating both their training and inferencing processes. These tasks are known for their high computational demands, making GPUs a crucial resource in the AI landscape. As we progress into Q3 2023, the demand for computational resources, driven by the rise of AI and other emerging technologies, shows no signs of slowing.
This heightened demand for GPUs has led to a significant supply shortage. Major cloud service providers, including Amazon Web Services (AWS), Microsoft, Google, and Oracle, have been compelled to impose restrictions on GPU availability for their customers, even affecting some in the crypto industry. In some cases, customers have reported wait times extending into several months.
Gensyn: Pioneering Decentralized AI Development
Blockchain technology can counter these forces and democratize AI development by offering decentralized data and computation marketplaces. Gensyn, a blockchain protocol, is taking a crack at this issue. Designed to "connect and verify off-chain deep learning work," Gensyn tailors its approach to machine learning. The decentralized nature of Gensyn's model eliminates intermediary service providers like AWS, connecting developers looking to train their AI models directly with available compute from a variety of sources, including Central Processing Units (CPUs), GPUs, and personal gaming computers.
The Promise of Decentralized Computing Networks: Solving GPU Shortage
The advent of decentralized computing networks has opened up a new avenue for addressing the global GPU shortage. These networks connect entities in need of compute resources with systems that have idle CPU and GPU computing power. By leveraging the vast supply of idle GPUs worldwide, these networks present a viable solution to the current supply-demand imbalance. Moreover, they offer competitive pricing compared to centralized providers, as there are minimal additional costs for individuals or organizations to supply their idle resources.
Decentralized compute networks can be broadly categorized into two types: general purpose (GP) and specific purpose (SP). GP compute networks function akin to a decentralized cloud, offering computing resources for a myriad of applications. They operate on a model similar to Airbnb, where clients can rent server space from providers who have the autonomy to set their own prices.
Akash Network: A Glimpse into GP Compute Networks
Akash Network is a prime example of a GP compute network, offering services ranging from deploying machine learning models for inference to rendering workloads.
Conversely, SP compute networks are designed for specific use cases. They typically adopt an architecture that amalgamates compute resources into a unified pool, akin to a supercomputer. The cost in SP compute networks is usually determined through a gas mechanism or a parameter controlled by the community.
Ensuring Accuracy in Compute Networks: A Review of Verification Methods
Ensuring the correct execution of compute jobs is a critical aspect of compute networks. Various mechanisms, including reputation models, Stake-for-Access models, and Trusted Execution Environments (TEEs), are employed for this purpose. However, these approaches have their limitations and potential vulnerabilities.
The Advent of zkML: Leveraging Zero-Knowledge Proofs in AI
A new wave of compute projects is now leveraging zero-knowledge proofs (ZKPs) to verify computational integrity for machine learning, giving rise to a sector known as zkML. ZkML not only enables AI on-chain but also offers broader applications such as privacy-preserving machine learning and verifiable machine learning.
Exploring the Render Network: Decentralized GPU Rendering
The Render Network is a decentralized GPU rendering network that is built on the Ethereum, Polygon, and, soon, Solana blockchain. The network allows users to share their unused GPU power with others, and in return, they earn rewards.
The Render Network is designed to solve a number of problems for digital creators. First, it can save them time and money. By sharing their unused GPU power, they can avoid having to purchase their own high-end GPUs, which can be expensive. Second, the Render Network can help to reduce the environmental impact of rendering. By decentralizing the rendering process, the Render Network can help to reduce the amount of energy that is used for rendering.
Large TAM: Render’s applications extend beyond the confines of the crypto market, catering to an array of industries such as gaming, metaverse development, augmented and virtual reality, architecture, animation, interior design, product design, and artificial intelligence (AI) tooling.
Stable Diffusion support: The integration potential with AI-centric tools is considerable. As seen in the recent surge in demand for Large Language Model (LLM) tools like ChatGPT, with reported daily active users exceeding 13 million, the application of LLMs to Render Network is a plausible future consideration.
Q1 2023 marked notable advancements for the Render Network, with the incorporation of Stable Diffusion services into the creator portal and substantial enhancements to Cinema 4D ORBX export speeds. These improvements reflect Render's commitment to both user experience and its strategic expansion into the AI space. Stable Diffusion services now form an integral part of the Render Network's portal. The integration of Stable Diffusion's intuitive prompt creation suite is the first step in a series of future initiatives poised to delve deeper into the AI domain.
Our Take
At Event Horizon Capital (EHC), we believe select cryptoassets will outperform all other asset classes over the next five, ten, and possibly even twenty years due to their superior qualities as new money/assets for the internet age. Because of this, we seek the best risk-adjusted exposure to protocols that personify the blockchain benefits outlined above. With crypto markets being one of the world’s most dynamic markets, our agile and active management provides the flexibility required for swift, decisive action while also never compromising on security.
EHC’s multi-strategy approach is built upon:
Qualitative fundamental research,
Quantitative tools and valuation metrics
Narrative and sentiment-driven market swings
This newsletter from Event Horizon Capital is intended for informational and illustrative purposes only and has been prepared to provide insights on the market. It should not be construed as an offer, solicitation, or recommendation to buy or sell any security or financial instrument, nor participate in any investment strategy. The opinions and information expressed in this newsletter are as of the date it was written and are subject to change without notice due to various factors, including changing market conditions and regulations. This newsletter is not intended as investment advice and should not be considered as such. Third-party data presented in this newsletter is sourced and deemed reliable, but no guarantee is made as to its accuracy or completeness. All investments carry risk, and there is no assurance that any specific investment, strategy, or product referenced directly or indirectly in this publication will be profitable or suitable for your portfolio. |
Comments