Bridging the Gap: Fairness, Attribution, and AI in the Age of Web3

Share this Article
Fairness, Attribution, and AI in the Age of Web3

In the throes of the AI revolution, we find ourselves facing dilemmas that are as much ethical as they are technological. Among the most poignant of these concerns is the unease of creative professionals — authors, musicians, actors — who fear their livelihoods are being threatened by AI’s rising proficiency in content creation. Their worry is valid as can be seen in this article “‘Bargaining for our very existence’: why the battle over AI is being fought in Hollywood.

From AI that can write compelling prose to those that can generate lifelike deep fake videos, we are living in an age where the traditional boundaries of creativity are being blurred. But could these same technologies that incite fear also hold the key to an equitable future for creatives?

In this article, we look at ways in which blockchain and AI technologies can help.

Fairness, Attribution, and AI in the Age of Web3
Image Courtesy of The Guardian article above | Actors and writers march the picket line outside Netflix on 20 July. Photograph: Clutch Pockets Wambli/Shutterstock

Where do DAOs play a role?

The recent surge in interest around decentralised technologies, in particular, DAOs (Decentralised Autonomous Organisations) and Web3 infrastructure, provides a fresh perspective on this issue. A DAO, governed by smart contracts on a blockchain, could serve as an impartial, transparent system for attribution and reward distribution among contributors. Think of it like a futuristic version of a royalty collection society, but one that is not centralised and is transparent in its operations.

Imagine a futuristic version of a royalty collection society, but one that is not centralised and opaque in its operations. Instead, it’s a transparent, auditable, and democratic system. Every transaction and decision would be recorded on the blockchain, visible to all, and could only be modified if the collective agrees.

In this DAO, the contributors — authors, musicians, artists, and even AI programmers — would have their works protected and their contributions acknowledged. The AI’s outputs would also be monitored, and their inspirations traced back as accurately as possible. If a creator’s work is identified in an AI’s output, the smart contract could automatically trigger a royalty payment, ensuring swift and fair compensation.

Web3 infrastructure further enhances the possibilities of this DAO approach by offering a decentralised and interoperable network where these interactions and transactions can happen seamlessly. It brings the promise of a fair, open internet where creators are directly rewarded for their work, and consumers can access and interact with content in a more direct, peer-to-peer manner.

Nevertheless, the application of DAOs to tackle this complex issue is in its infancy. There are numerous legal, technical, and ethical hurdles to overcome. But the promise is there — an impartial, decentralised solution to one of the most pressing problems of the AI age.

The role of Zero-Knowledge Proofs

In the context of large language models (LLMs), however, things get complicated. Blockchain technology, specifically Zero-Knowledge proofs (ZKPs), offers a promising method for auditing changes in data without revealing the underlying data itself. This could prove invaluable for verifying the authenticity of inputs to AI models and maintaining the integrity of the process. Yet, LLMs do not simply copy their inputs; they learn and extract patterns, transforming this knowledge to generate outputs. This means it’s challenging, if not impossible, to directly map a piece of input data to a specific output. Nevertheless, the potential of Zero-Knowledge proofs (ZKPs) in this context should not be disregarded.

Despite the inherent difficulties, ZKPs can serve as a powerful tool in tracking the utilisation of data in the training of large language models (LLMs). This cryptographic principle allows for the confirmation of a claim about specific information, such as data input into an LLM, without revealing the data itself. This could provide a form of tagging system, albeit one that operates differently from what we might intuitively imagine.

Through ZKPs, changes — or the lack thereof — in the input data can be audited and proven without directly exposing the data. This level of security and privacy is paramount in scenarios dealing with sensitive or proprietary information. Suppose a piece of music, a novel, or a confidential document is part of the input data for the LLM. In that case, ZKPs can provide a robust means of ensuring that the usage of such data can be tracked and proven, while maintaining the necessary confidentiality.

It’s important to note that while the ZKP can verify whether a certain set of data was included in the model’s training set, it doesn’t reveal how that specific data influenced the model’s output. Therefore, while ZKPs add an important layer of security and accountability in the data utilisation process, the complex nature of how LLMs learn and generate outputs remains a challenge to be addressed in the attribution discussion.”

What role does AI itself play in a better solution?

Enter AI once again, this time as a potential solution to its own quandary. By analysing the stylistic and thematic elements of an AI’s output, we may be able to infer the source of its ‘inspiration.’ Although such a method would likely have its limitations — high false positives/negatives, the complexity of style analysis, and potential privacy concerns — it’s a direction worth exploring.

Picture an AI system, akin to the deep learning models currently used for authorship attribution or music genre classification. By training on a wide variety of styles and themes, this AI could then analyse the outputs of an LLM and deduce patterns, rhythms, and stylistic quirks reminiscent of certain inputs. In essence, it would be identifying the ‘fingerprints’ of various sources, even if they were abstracted through the learning process of the LLM.

However, this method is not without its limitations. For one, the risk of false positives and negatives could be high. Given the large volume and diversity of data that LLMs are trained on, the AI could attribute outputs to incorrect sources, or miss valid connections entirely.

Furthermore, the complexity of style analysis should not be underestimated. Styles are often subtle, subjective, and multifaceted — qualities that can be challenging for AI to accurately grasp. Can an AI truly understand the nuance of a poignant metaphor, the subtle rhythm of a literary cadence, or the emotional resonance of a particular chord progression? Our current technology still has limitations in this respect.

Finally, privacy concerns cannot be overlooked. If an AI is analysing output to infer sources, safeguards must be in place to prevent potential misuse of this capability, especially when dealing with sensitive or confidential inputs. We need to ensure the balance between the pursuit of attribution accuracy and respect for privacy.

Despite these challenges, this approach represents a fascinating convergence of technology and creativity, offering a potential path to reconcile the tension between AI-driven content generation and the rightful attribution to human creators. It might not provide a perfect solution, but it opens a new field of exploration in our quest for fair and accountable AI systems.”

But what about the grey area?

At the same time, we must also consider the nature of creativity itself. Artists, whether consciously or not, draw from a broad spectrum of influences to create their work. Musicians learn and borrow from one another, producing remixes and mashups. Is AI-generated content not just another form of this tradition? Rejecting AI because it accelerates the creative process might be akin to dismissing electronic music production tools for their mechanical origins.

Yet, the analogy isn’t perfect. While musicians engage with their influences actively and knowingly, AI models absorb data passively, lacking the contextual understanding and intention that human artists possess. Therein lies a grey area and a compelling argument for a careful, nuanced approach to AI in creative fields.

Some tough questions ahead

Moving forward, we must grapple with some difficult questions. How can we ensure fair compensation for creators whose work contributes to training AI? How much ‘influence’ should AI be allowed to exert on its output, and how should this be regulated? And perhaps most importantly, how can we maintain a human-centric perspective on creativity while leveraging the power of AI?

As we tread into this uncharted territory, the guiding principle should be one of fairness, giving due credit to those whose work paves the way for new technology, while also acknowledging the incredible potential of AI. In the end, it’s not about pitting humans against machines, but rather, finding ways in which the two can coexist and, better yet, mutually benefit.

Therein lies the promise of the Web3 wave — not just as a tool for creating decentralised applications, but as a framework for a future where innovation and tradition coalesce, fostering a space that is as respectful of its roots as it is eager for what’s next.

Email Us

we@notcentralised.com

Visit Us

Lvl 14, 5 Martin Place, Sydney NSW 2000

Get in Touch

Fill out the form to find your solution.

Get in Touch

Fill out the form to find your solution.

Stay in the Loop

Subscribe to receive updates.

Stay in the Loop

Subscribe to receive updates.

Mark
Monfort

Mark drives innovation with his deep understanding of AI, blockchain, and data technologies. His experience spans over 15 years of contributions to finance, technology, and operational strategy across Australia, Europe, and North America.

In 2021, he transitioned from Head of Data and Technology at a leading Australian accounting firm to startups. Prior to this, he worked in equity and macroeconomic research in the capital markets space.

Mark brings a passion for data and insights to NotCentralised. His understanding of AI and blockchain technology is central to the development of workplace productivity and financial system modernisation products, including SIKE and Layer-C. Mark’s dynamic and solutions-focused methods enable the navigation of complex technological landscapes and new market potentials.

Mark holds an Executive Master’s and a Bachelor of Commerce. He led the creation of the Australian DeFi Association and serves on the advisory board for the Data Science and AI Association of Australia. His commitment to such communities demonstrates his enthusiasm for emerging technologies and vision of positive change through technology adoption.

Nick
Bishop

Nick spearheads product strategy and institutional business development, leveraging a rich background spanning 23 years in capital markets and financial services across the UK, the US, and APAC.

In 2020, Nick transitioned into startups, bringing extensive experience in asset management and corporate advisory from roles including Director, Head of Australian Fixed Income at Abrdn and Managing Director, Head of Corporate Credit at Gresham Partners. His expertise extends to client management across the government and private sectors.

With a First Class degree in Law and Criminology and Chartered Financial Analyst experience since 2002, Nick is known for his energetic and creative approach, quickly appraising business models and identifying market opportunities.

Beyond his role at NotCentralised, Nick actively contributes to multiple startups and SMEs, holding various Board and advisory positions and applying his institutional expertise to early-stage ventures. Nick is fascinated by emerging technologies with significant societal impact and loves to immerse himself in nature.

Arturo
Rodriguez

Arturo leads product development and software engineering, applying over two decades of experience in technology, capital markets, and data science. With his years of programming expertise, Arturo smoothly transitioned into blockchain, AI, and machine learning.

Arturo has built and sold technology startups across Europe, following quant derivatives roles in global investment banks. His prior experience includes data projects for the NHS in the UK, Strategic Technology Advisor at Land Insight, and Senior Advisor to OpenInsight, where he built predictive models for vessel usage in commodity markets.

A mathematics and statistics graduate from Stockholm University, Arturo’s early grounding in logic problems and data manipulation techniques is evident in his practical applications. His work building equity derivative pricing models for Merrill Lynch and Royal Bank of Scotland showcased Arturo’s highly specialised skillset.

Arturo relocated from London to Australia in 2020. Beyond NotCentralised, his passion for technology and industry involvement extends to the Australian DeFi Association, which he co-founded, and regular contributions to the Data Science and AI Association.