Cultural and regulatory conversation worldwide today show that there is a lot of fear around AI. AI is going to take over the world, it’s coming for everyone’s job, it’s going to destroy Hollywood: choose your own dystopian adventure. I don’t want any of those things to happen, either, but I’m an AI optimist at heart. There are a lot of things that humans are pretty bad at, and we should want AI to help us with those things, and work together with us to fix them, especially when it comes to solving complex problems.
It will soon become incredibly difficult to know for sure what is real and what is fake.
We’re already seeing more maliciously generated content online, especially around high-stakes events like elections or periods of social unrest. While political misinformation is by no means a new problem, it can be scaled up dramatically with artificial intelligence (AI) tools — as well as personalized to an extent not possible before now.
Imagine not just fake videos of real politicians speaking, but fake videos of entirely fake politicians spewing fake quotes targeted at a particular audience’s beliefs or even at one single person’s views. This is already possible with powerful LLMs like ChatGPT and Bard, and it could spin out of control sooner than we think.
Large language models (LLMs) can gain outsized influence quickly because they are language-based, as is most of our society — from media and communication, to politics, to law — making it harder to detect which information is real versus generated.
In the very near future, one of the biggest negative impacts of powerful AI models is this inevitable explosion of generated content and fake news –– and the best way to manage this is with blockchains.
By introducing blockchain-based “sign-and-trace” systems for content authentication to important media outlets, we can reliably source and validate real news stories when they are published, leaving a legitimate trail of sources that can be followed online.
It’s not all bad
To be clear, I am an AI optimist at my core.
Before founding NEAR Protocol, I spent a decade as an AI researcher primarily at Google. I’m also a co-author of the landmark AI paper, “Attention Is All You Need,” which introduced transformers, the model architecture powering the LLMs in the news today. I am well aware of both the amazing potential and the risks that today’s LLMs present to society.
I also strongly believe that AI and Web3 can complement each other, and that blockchains and Web3 communities are a good match for addressing the current system challenges that LLMs help exploit. The Web3 solutions that I’m proposing will be more effective than old-school Big Tech regulation, which is neither practical nor possible at this point in time.
We can introduce provenance tracking around a piece of content — an article, a quote, or even an image or video — when it is created as well as each time users interact with it: hence, sign-and-trace.
Content authentication systems are well-suited to run on blockchains for a few simple reasons: They can be linked to a persistent digital identity, there is an on-chain immutable record of every action or transaction that can’t be changed later, and they can be connected to a governance structure and set of rules that must be followed to interact with the system.
Read more from our opinion section: TV is dying, but decentralization is the cure
This type of system — using cryptographic signatures to verify content — will make it easier to trace where information comes from, as well as who is using that information and how they are using it. Sign-and-trace also makes it possible to clearly see what information online has been mislabeled or misattributed.
Whenever anyone references content within this system, it is recorded and audited in a tamper-proof way: This creates an immutable record of every interaction that is visible to everyone. These interactions can be linked to digital identities that can build reputation over time, so people know which sources are trustworthy and verified.
Because these records would be open source and transparent, the burden of auditing and monitoring content could be spread across different communities with a greater diversity of opinions. Putting this monitoring and authentication work on an open market — which blockchains are perhaps better at than anything else — would supercharge our collective ability to tackle this important, but likely overwhelming, mission.
Combating misinformation and fake content is a clear example of the kind of problem Web3 is best at solving: natively open and global, transparent yet verifiable, and proactively managed by a community without single-party bias.
Blockchain-based content management and sign-and-trace systems for media are necessary: We need to get a handle on what is real online and keep it that way, because soon, it will be too late.
Get the day’s top crypto news and insights delivered to your email every evening. Subscribe to Blockworks’ free newsletter now.
Want alpha sent directly to your inbox? Get degen trade ideas, governance updates, token performance, can’t-miss tweets and more from Blockworks Research’s Daily Debrief.