ETHDenver Recap: Emerging Trends in web3 and AI
Where we're at, where we're heading, and the return of Kevin.
Hey there! I got back from Denver a few days ago, and it was quite the trip. I was in town from Tuesday evening until Sunday morning, and really jammed a lot into my schedule in a short period of time. I have a lot of thoughts surrounding various projects and evident trends at the conference.
This article will first offer a brief recap of the conference on the whole, and then dig into some of the emerging and popular trends happening right now in the space. For those interested, most of the talks I attended are available on YouTube. Some parallel conferences, like Proof of Data Summit, do not appear to have the talks online. But, feel free to reach out to me for audio recordings or transcriptions, I’d love to share them.
So, let’s get into it. If you’re just interested in emerging tech trends, skip to the second half of this article.
But First: Denver and Conference Vibes
I arrived in Denver late on Tuesday. I’ve been in Orlando for the last few months, so the cold took some getting used to. I stayed downtown, with a short commute to the conference daily. I rented a small garage apartment and had no issues getting around the city.
During lunch on the first day in the city, I walked over to Colorado’s state capital within Civic Center Park. Despite having visited Denver many times, I had never actually made it over to this part of the city until this trip. The Colorado State Capitol was constructed in the 1890s with Colorado white granite, and is intentionally reminiscent of the US Capitol. The rest of Civic Center Park is worth walking around, especially including the Denver City Council building (and the statue of Broncho Buster, capturing the legendary soul of the old west). Checking out part of the city was a nice contrast from the densely packed conference and non-stop socializing that would embody the rest of the week.
I’m not going to pretend to completely understand the logic behind the layout of the conference. Let’s just say I got lost multiple times and I felt like the stage names could have contrasted between the buildings. What is a Sporkwhale? What is a Bufficorn? The merch at ETHDenver is popular, and completely sold out by the end of the conference. And no, I didn’t see Vitalik. I think I saw the back of Charles Hoskinson’s head, though. That counts for something.
The conference, as many already know, is free to attend. It’s an interesting amalgamation of things happening on the internet, with a common theme: people love Ethereum and the community that builds it and on top of it. With that said, there was still a Solana booth and Solana side events, and plenty of other notable projects that are only tangentially related to Ethereum projects.
After what feels like a lifetime, cryptocurrency-centric web3 gaming projects are finally starting to look like games that could be fun or could be worth spending some time with. With Epic’s store supporting games that have crypto mechanics, I think it is entirely possible that the impending bull market really brings attention to many projects that have been building during the bear. If digging into more web3 gaming projects interests you, and you’d like a separate write-up focused just on that and the GameFi space, leave a comment and I’ll write something up.
Unlike Art Basel back in December, ETHDenver placed a much lighter emphasis on NFTs and on-chain art trading. This is probably a good thing! As time goes on, we’ll see more conferences oriented around strictly digital art, and part of that space will include NFTs.
The conference wasn’t exclusively cryptocurrency focused, as I’ll expound upon in more detail later in this article. Besides AI, the presence of RFK Jr. at the conference was notable. Pro-crypto and running as an independent, the fireside chat at the end of the conference on Saturday was likely the most attended talk by a longshot. Heckling ensued, of course, but Robert certainly had a command and knowledge of the space in a capacity not seen by any other candidates running for POTUS in 2024 (Nikki dropped out today, so now we’re back to 2020 all over again). I’m not saying I’m a supporter, but it was worth sitting in on regardless.
And, of course, no conference could be complete without Kevin, from the ‘Worst NFT project ever’, with art so bad you have to love it. Kevin became a meme in 2022 with Pixelmon, and Coinbase brought Kevin back as an NFT on Base. Base (Coinbase’s low-cost ETH L2) had a notable presence at the conference, but that shouldn’t come as a surprise.
I’ll be linking all the notable AI x web3 talks once we get to that section, but one of many talks I think would be worth watching would be with Chris Dixon, promoting his new book. A lot of the material will be a refresher or not useful for the average web3 gigabrain or degen, but this could be a great book to buy your parents if they’d be willing to read it. So much of the future of this space requires it become less insular and more mainstream. Of course, part of what makes web3 appealing is the exclusivity factor (at least when it comes to the social aspects!).
The multifaceted nature of the web3 landscape in 2024 is fascinating. On one hand, we have projects actively trying to make crypto accessible and understandable to the average person (someone who isn’t into technology and simply wants to transact as a means to an end). On the other hand, we have insular and bizarre projects that have so many layers to them that attempting to explain it to someone outside of the space would make you sound insane. What exactly is a HARRYPOTTEROBAMASONIC10INU ETH Denver Party? What does that mean? Why is this memecoin relevant in the space, and who is ultimately behind promoting and driving seemingly homegrown projects like these? How many licks does it take to get to the Inu center of the Cryptoverse? The world may never know.
Make no mistake, this conference is still primarily aimed and geared towards people building in the space. However, I do think things are getting more accessible, generally speaking. For someone new coming into the space, they might ask ‘why do we need all of these projects?’ or ‘what is the difference between everything happening here?’ These are questions that are difficult even for people within the space to articulate, but I think that, relatively speaking, crypto still feels nascent. It still has a sense of magic to it, and the competitive nature of the space is driving people to create truly innovative solutions to/within decentralized finance.
Denver was a blast. I met up with a good amount of friends, colleagues, and met a couple new people, too! If we spoke or hung out during the conference, know that I loved every second of it.
Having a beginner’s mindset and forcing yourself to talk to people you don’t know helps tremendously for events like these. Many of the people here will share similar interests to you, given that they’re showing up as well as investing a non-trivial amount of time and effort into being here.
With all of this said, let’s dig into the nitty gritty. The reason why I came to the conference in the first place, AI and web3. What are the themes we’re seeing here in 2024? What movements or ideas are capturing the hearts and minds of the people involved in this space? Are web3 and AI as complementary as we think they are upon first glance?
Key Takeaways: Emerging Trends in web3 and AI
This year, web3 x AI has become the bread and butter of the conference. Taking a look at the conference talk list, a substantial portion of the talks oriented around AI.
To keep it simple, the themes I saw present at ETHDenver from a web3 x AI perspective can be encapsulated into three common themes:
AI integration: Decentralizing AI training and model deployment, fraud detection, decentralized storage, and utilizing AI to optimize decentralized systems.
Decentralized identification. Enhancing security and trust by providing verifiable credentials that are cryptographically secure. Reputation systems that are underpinned by DIDs. Using DIDs and VCs to support decentralized financial systems and e-commerce in general.
Agents/Agentics: Automated asset management, better and more personalized user experiences, better security, agents and their place within DAOs, and AI agents interacting with smart contracts.
Let’s go down the list and start first with AI integration more generally. There’s a lot to cover there, but I promise it is interesting.
AI Integration (Decentralizing AI, Storage, and Compute)
Doug Petkanics gave this DePIN talk twice last week, once at Proof of Data Summit, which I attended on Friday separately, and again on Saturday at ETHDenver. DePIN stands for Decentralized physical infrastructure network. As described by CoinGecko, DePIN “is a term that connotes projects developing blockchain-based management systems for real-world facilities like marketplaces for computing power and real-world data.” Doug’s talk, since he’s from Livepeer, emphasized the importance of decentralizing infrastructure to avoid regulatory capture and ensure broader access to AI technology.
As I’m sure you all are aware, there are serious issues right now when it comes to getting access to high powered or desirable GPUs for training and computing resources for AI. Startups have a difficult time affording to buy the GPUs, or even to rent them on a limited basis to train models. This is a major problem! Many startups have already failed because they lacked the resources necessary (or the connections) to get the right amount of compute at the right time. Blockchain will be used in the future to democratize AI access, reduce costs, and lead to innovation by leveraging underutilized computing resources globally.
Web3, as a self-certifying, permissionlessly composable web, is key to achieving seamless interoperability required for planetary intelligence. Planetary intelligence can be described as the acquisition and application of collective knowledge on a global scale. Think of the coordination of a bee colony, and one can imagine how the inevitable future of the web necessarily entails further and deeper connectivity. We already have a hivemind, but it is fractured compared to where things are heading. This feeds into the idea of agent (software representing entities) frameworks, in a future where the internet will be populated by both humans and AI.
The Proof of Data Summit was all about this, and all of the talks oriented around decentralized computing and their relation to decentralized identification. This is quite the rabbit hole to fall into. While we could speak more broadly, and say ‘decentralizing everything is where things are heading’, I think the space is more focused right now on decentralizing GPU access and compute.
Greg Osuri at Akash Network gave a great, succinct presentation on Saturday, advocating for blockchain-based marketplaces that provide access to GPUs and compute in general. Much like how cars spend most of their time sitting doing nothing, the same is often true for computers globally. With big tech companies competing to enhance their GPUs, cloud-based solutions for training models is one of the better solutions. The current wait time for top-tier NVIDIA GPUs is over 2 years! Decentralizing compute is one aspect of AI, but another is storage.
Clara Tsao at Filecoin gave a great presentation at the end of the conference entitled: Building the Next Generation of Data Storage. Filecoin claims that decentralized storage will provide a more equitable and reliable solution for ML datasets. Think more decentralized, open, and transparent. We want this, and such models or solutions will be more resilient to monopolistic data ownership. If we can have proof of storage and reliable data integrity on-chain, that would be a good thing! We want to know that the data we’re utilizing hasn’t been tampered with, and is what it claims to be. Again, it seems obvious, but most data online is gathered, stored, and provided by a handful of companies (and you know what companies those are).
While decentralizing compute remains a top priority, decentralizing data storage should be 2nd on the list. Data is the fuel of AI, and blockchains are perfect for creating a safe and transparent environment for the future of AI usage globally. With the global AI market set to grow substantially in the next decade, we need somewhere to park the data. Do we want all of that living on AWS or Google’s servers? I don’t think so.
Decentralized Identification (DIDs, VCs, and DIDComm)
Golda Velez from Ceramic gave a great talk, entitled ‘Decentralized Trust and AI’
Decentralized identification and the concepts that encompass it are less well known, even amongst people within the web3 space. Having a decent intro to these ideas seems appropriate given most people’s lack of familiarity.
But first, why are DIDs important, what do they enable? Why should I care? Why does it matter?
Enhancing Security and Trust: DIDs enable a more secure and trustable environment by providing verifiable credentials that are cryptographically secure. This helps in reducing fraud and enhancing the reliability of transactions and interactions within the network.
Empowering User Privacy: By allowing individuals to control their identity data, DIDs shift the power from centralized entities to users, promoting privacy and data sovereignty.
Facilitating Reputation Systems: In the context of decentralized systems, DIDs can underpin reputation mechanisms, where users' interactions and behaviors can be securely recorded and assessed, fostering trust and accountability.
Enabling Interoperability: DIDs contribute to creating a more interconnected and interoperable web, where identities can be consistently verified across various platforms and services, removing silos and enhancing user experience.
Supporting Decentralized Finance (DeFi) and E-Commerce: DIDs are instrumental in DeFi and decentralized e-commerce platforms, where they can provide secure identity verification, reduce the risk of fraud, and enhance trust in transactions.
DIDs could be part of a larger future, where society is transformed through decentralized incentives. Programming society incentives on-chain could lead to dramatic shifts in how economies and societies function, addressing large-scale challenges and optimizing for human values. Capital could be routed more effectively to catalyze impactful projects and initiatives. We’ve seen proto-examples of this with DAOs to great success already, and many of these ideas are similar to some ideas posed and suggested by Balaji. The usage of DIDs will ultimately enhance security and trust in decentralized networks, leading to more reliable verification of entities.
Why is verification of entities important?
Well, because soon enough, a majority of the entities operating on the internet aren’t going to be human. We need a way to identify these entities, which we call agents.
To create trusted spaces, we need two things:
Long-term identifiers for entities
Ability to share data about their behavior
Blockchains are uniquely qualified to assist in both of these needs. Without these, we cannot have trust in a fashion that would be beneficial or useful for those of us who want to keep using the internet once AGI occurs. If every post or data event is signed with a DID, you can permissionlessly make attestations about that DID and models can roll up the claims to make predictions.
Are you seeing how all of these ideas tie into one another? We want infrastructure and standards that can scale and support the next inevitable era of the internet as we know it. This transitions right into our next topic, agents.
Agents/Agentics
AI Agents with Crypto Wallets: That’s a Good Thing…Right?
When we think about the ‘killer app’ for web3, it hasn’t happened yet. And it hasn’t happened yet because we would all know, right? We can think of Facebook and traditional social media as being the ‘killer’ app for web 2.0, with a focus on user-generated content and usability for end-users, especially when compared with web 1.0. Web3 entails a more semantic web, with increased interconnectivity and the usage of digital identities with blockchains making some of the decentralized magic actually happen.
As I’ve written about before, more and more experts in the field of AI agree that we are approaching a point where artificial general intelligence is not only likely, but inevitable. Once AGI-level intelligences are running around on the internet, we will hope and pray that society has landed on standards that can be used to verifiably identify, audit, and manage them online. Security for such entities online that exceed human intelligence is an open and unanswered question, but it is a problem that is increasingly shifting from science fiction to reality.
Agents, or whatever we’ll eventually call them, will result in an evolution in how we interact and transact online, particularly within decentralized environments. AI agents will be used to autonomously manage digital assets, trade, and make investment decisions. We can imagine AI agents interacting with smart contracts and automating processes that were formally manual. This will raise the bar for how things are done online, and eventually it will seem odd to do it manually, much like how driving a car in a world of autonomous cars is envisioned (eventually, c’mon Elon!).
As we’ve seen with products like Rabbit R1, similar models either on your phone or computer will hyper-personalize everything. This may even result in the complete abstraction of how you engage with programs and interfaces altogether, with the agent model tuning the interface based on your knowledge, level of understanding, or simply tuning it based on what it thinks you’re doing (think token prediction but with a UI). The future is looking more and more like a natural language interface.
Because the internet will be increasingly agentic, we will soon find ourselves in an arms race with AIs. You’ll want the best possible agent that can monitor and respond to security threats in real time, and the bad guys will want a model ever so slightly smarter to hack you.
Most of the participants of DAOs in the future will be AGI agents, as virtually all of the interaction done through or with most DAOs is online, requiring little more than an internet connection and a chromium browser.
Takeaways
The crypto space is heading towards a more integrated, decentralized, and innovative space, where AI, decentralized identification, and AI agents will play a significant role. Platforms are incorporating AI to decentralize video streaming, and using blockchain to distribute AI tasks across the network. AI is increasingly being used within the web3 space to detect anomalous behavior and ensure a safer and more secure space online.
As AI gobbles up more and more GPU and compute resources, it also gobbles up more storage space, and decentralizing storage is crucial for the data-intensive nature of AI. Decentralized identification is viewed as a necessary primitive to AI agents on-chain, empowering user privacy and enhancing security and trust across the board. Agents can be viewed as a natural stepping stone for functionality within the web3 space, eventually offering automated asset management and more personalized and human-centric user experiences.
As we move towards the future, the line will continue to be blurred within the web3 and AI spaces. They are converging and will continue to converge as we approach the singularity. These spaces are complementary, so learning about one will help you better understand the other.
These notes constitute just a handful of the talks present at ETHDenver this year. I encourage anyone reading this to really dig in. Watch the talks and learn more about the networks and projects mentioned. Decentralized AI, DIDs, and agents aren’t going away anytime soon.
Thanks for reading, and peace be with you.
-Chris