Consensus 2024 Recap: Trends, Highlights, and The Future of the Crypto Industry
Exciting projects, the zeitgeist, and where I think things are heading at the intersection of web3 and AI.
Hey friends! I got back this weekend from Consensus 2024, hosted this year in Austin, and wanted to share my insights and thoughts. This piece will be heavy on photos and nuggets of knowledge, and won’t be as much of a deep dive into any project in particular.
With that said, let’s get started!
The Appetizer: GenAI Summit in San Francisco
Before we get into the meat and potatoes of this piece, I want to briefly mention the GenAI Summit, hosted this year in San Francisco. I had the opportunity to go to both of these conferences back to back last week, and it was a great way to get a feel for the general zeitgeist of both spaces and how they’re overlapping and converging into each other.
GenAI Summit was hosted in the beautiful Palace of Fine Arts. The building has an interesting history, so I do recommend reading up on the history of it, if that’s your thing.
I can only speak to the first day of the conference, as I only attended on Wednesday. What I was looking for at the conference, and where I was focused, is where AI and web3 are intersecting. That is, how are projects leveraging the decentralization and benefits of web3 within their AI solutions. Think AI customer service for web3, AI agents on-chain, and making web3 either easier to navigate or automating aspects of it.
Some of the biggest names you might expect from the AI space had booths here, and many projects that had just started in the last 12-18 months that are heavily utilizing APIs from major providers. No surprise there.
The venue’s design and open layout made listening to speakers quite difficult, and getting passes on the first day was utter chaos. I have hopes that next year’s event will be organized better. There’s clearly a massive amount of interest for large AI events like these, and tickets were oversold. Too many people, not enough room.
Gripes aside, here are my major takeaways as they relate to where AI and web3 are heading:
It is clear that the industry is looking for ways to democratize access to GPU clusters (like H100s), and demand is far exceeding supply. Decentralizing training of AI models is an unsolved problem that could help the industry expand faster and offer incentives to those who offer up spare compute. That’s where web3 comes in.
AI was already an institution and an industry. Now, more clearly than ever, it is becoming something that is rivaling web3 for both hype and attendance for conferences. AI is cool and exciting.
Every significant advance in AI will kill off a lot of the startups that are trying to make it in this space. Better multimodal models that are usable and accessible on mobile and traditional computers could easily wipe out many of the services being offered. OpenAI’s next model will kill off countless small businesses, and they have access to the compute and raw intelligence and talent to do this.
Democratizing access goes beyond just getting access to cloud compute to train. We want a multipolar future, one where multiple companies, organizations, DAOs, and individuals own the future, not just Microsoft and OpenAI. Rivaling their progress amounts to having access to compute, so we either need to get a bunch more GPUs into the hands of all these small businesses or figure out how to create solid methods of distributed access to compute globally. This is where decentralized compute marketplaces powered by open networks like Akash can come in.
Every year, open source local models trail behind closed source models usability and efficiency. They still trail, but we have open source models you can run locally that rival GPT 3.5 (and some might argue 4.0) at this point. This is slowly becoming the case with multimodal models as well.
We haven’t expelled or exhausted all of the text on the internet for training, but good data is extremely important for efficient and usable models.
We can only get models so smart with text as we approach AGI. The next leaps are going to happen with models that have access to sensory data and incorporate them into their digital brains. This is akin to expecting a child that cannot see to have the same depth of intelligence about the world as one that can see, touch, taste, and feel. Intelligence as we think about it as humans is inherently multimodal.
The Turing Test has already been passed. Many of these models already surpass human intelligence in plenty of domains. As we get increasingly uncomfortable, the sliding scale keeps on sliding.
That’s enough for GenAI. Let’s move onto Consensus.
Main Course: Consensus 2024
Consensus this year took place in blisteringly hot Austin, TX, from May 29th to the 31st. It wouldn’t be web3 without countless side events, and especially this year, an appearance from Robert F. Kennedy Jr. I attended the main two days of Consensus, Thursday and Friday. By Friday mid-day, the Austin Convention Center had cleared out. One thing is for sure: we’re not in a bull market. Bitcoin might be high, but this convention wasn’t as jam packed as it was last year. There’s still excitement and institutional interest (see Google’s presence in the image above), but it definitely felt like things were just puttering along.
So, what changed from last year?
Bye bye NFT projects, hello GameFi. There’s a larger presence of more substantial, GameFi-oriented projects, or projects made in Unreal engine with a heavy focus on in-game economics, Solana has become the obvious network of choice here. I don’t think I saw a single NFT project booth this year. However, there is still some presence of the notion of the metaverse, with projects like Somnia having a booth.
Brands in web3 have notably vanished. Unlike last year, where there were a lot more talks about how brands could incorporate web3 tokens and NFTs into their marketing, this is virtually absent now. Even projects that had a strong presence in the space have deleted their Discord servers, like Budweiser. I suspect this will come back again once there’s another bull.
Back to fundamentals. Less hype, more serious talk. This is more or less obvious considering the state of the market, but the only projects with a presence here are the ones that can afford to be. Solid mixture of legacy projects (essentially anything that has survived for the last few years) and new projects that have a heavy emphasis on identity and authentication, and acting as DePIN for GPU access.
Less bullish on Ethereum. There’s more of a push for new L1s that can scale massively to accommodate the kinds of transaction volume we might expect for DeFi to take over in the place of traditional finance. A couple of talks had speakers mention ‘is Ethereum going to die?’ This gets into the notion of legacy blockchains, not dead blockchains – chains that will exist in the future purely for nostalgic purposes and have little to no active use or utility. This isn’t the case right now, but it came up more than once at the conference.
Liquid re-staking. Eigenlayer conversations took up a good chunk of the conference. Look into it if you’re unfamiliar.
AI x web3. This is probably the biggest notable change from last year to this year. More and more projects are talking about things in terms of democratizing access to GPUs, on-chain agents, and finding synergies between the current deficits and issues with access for AI training and how web3 can help solve those problems.
Overall, less people. I found myself with others at the conference saying ‘wow, this is dead’. Many talks I thought would be filled to the brim with people listening were poorly attended (for example the talk Neal Stephenson gave on Friday afternoon). Even talks with famous individuals in the space like Anthony Pompliano were easy to sit in on.
But hey, more room to wiggle around, right? In stark contrast to GenAI, which was over-attended and poorly conducted, Consensus was lacking in attendance but impeccably run. For those who didn’t attend that want to get a better feel for what was talked about, check out the schedule here.
This is my general take, now let’s dig into three particular intersections and topics of interest that really caught my attention: DePIN, AI agents, and identity. All of these intersecting spaces tie into each other.
DePIN
Decentralized Physical Infrastructure Networks (DePIN) are emerging as a critical component of the web3 ecosystem, with far-reaching implications for the future of AI and decentralized applications. By enabling the decentralization of storage and compute resources, DePIN projects are laying the foundation for a more accessible, efficient, and resilient infrastructure layer that can support the growth and evolution of AI and web3 technologies.
DePIN had a massive presence this year at Consensus. But why? Why does this matter? We’re seeing a clear intersection between needs in AI and where web3 can step in.
Key benefits of DePIN:
Democratizes access to hardware resources needed for AI training and inference
Provides a cost-effective and efficient alternative to traditional cloud computing providers
Helps address key challenges facing the AI industry, such as data privacy, security, and bias
Enables the decentralized storage and processing of data for web3 applications
Perhaps democratized access to GPUs utilizing web3 tech will be what we need for massive AI adoption. Many projects seem to think so.
A list of DePIN projects to check out:
Fluence Network - “Cloudless computing platform for the future of the internet”
Spheron Network - “On-demand DePIN for GPU compute”
Akash Network - “The World’s premium decentralized compute marketplace”
Hivello - “DePIN manager”
Sia - “The future of cloud storage”
Koii - “A practical and complete dePIN network”
NeuroMesh - “World’s largest AI model, trained by everyone, for everyone”
This is just scratching the surface, and as we move into talking about AI agents there’s a great degree of overlap.The above resources are a great place to get started.
AI Agents
AI agents, agentics, agents on-chain, etc are becoming increasingly important in the web3 ecosystem, as they enable the creation of intelligent, autonomous entities that can interact with decentralized applications and services. These agents can take many forms, from simple chatbots and virtual assistants to more complex, self-learning systems that can adapt and evolve over time.
Key benefits of AI agents in web3:
Enable the creation of more engaging, interactive, and personalized user experiences
Can automate tasks and processes, reducing the need for human intervention
Have the potential to make web3 applications more accessible and user-friendly
Can help to drive the adoption and usage of decentralized services
Agents can operate in decentralized, trustless environments, unlocking new possibilities for how we can use web3. Yes, there are still major concerns and hurdles to overcome for security, governance of agents, and interoperability (and incentive mechanisms and general user education). Despite all of this, plenty of projects are seeing the potential here, for using AI-powered agents to smartly unlock the often opaque world of web3.
Here’s a list of AI agent projects that are oriented around web3 worth looking into:
Morpheus - I wrote about this a few weeks ago!
Valory - Autonomous agent economy platform
Gaianet - “Decentralized and incentivized GenAI agent network”
Theoriq - “Modular and composable AI agent base layer”
Soon, agents will be everywhere. We’re already seeing them slowly seep into e-commerce, but agents will in time be thought of as assistants, often ones that follow you wherever you go. Siri will eventually become this. But your super Siri is going to need a name soon.
Identity
Lastly, there are concerns about identity. We can’t have all of these nameless and faceless entities running around the internet without some accountability. As more and more agents get spun up, we need a way to know who is controlling them. This ties into the idea of content provenance – as the proliferation of AI continues, how can we trust anything that is on the internet? Digital watermarking becomes even more important! Identity will come to be an increasingly important component of the internet and how we engage with content. In the context of web3, identity is often associated with the concept of self-sovereign identity (SSI), which gives users control over their personal data and enables them to selectively disclose information as needed.
How does identity specifically tie into web3, though?
Key benefits of identity in web3 (and just generally on the internet):
Enables users to securely and transparently prove their identity
Gives users control over their personal data and privacy
Facilitates secure and efficient interactions between users and decentralized applications
Helps to prevent fraud and identity theft in the decentralized web
Lots of projects are trying to figure out how to tackle the authenticity problem on the internet. Offering biometrics is a solution, and utilizing biometric passports is another. Zero knowledge can come in handy here, as we often want to preserve privacy while conclusively saying you are who you say you are. In any event, there are a lot of projects in the space solely focused on this, and identity will eventually encompass both people and the agents they govern (at least until AGI breaks free).
Here are some identity projects worth exploring:
Verus - ‘A new paradigm in digital identity and ownership’
Web3Auth ‘go beyond web2.5 for users to offer a fully self-custodial login experience with WaaS (wallet as a service)’
Worldcoin - ‘Give me your eyeballs!’
Prove - ‘The modern way of proving identity’
Concordium ‘Creating a future based on trust, privacy, and safety with blockchain tech’
Privacy preserving identity solutions are top of mind in the space right now. It isn’t sexy, but it will be necessary for the future. The projects above all have their pros and cons, but they’re onto something, and aspects of these will be incorporated into whatever replaces the modern form of SSO logins.
Wrapping Up
I’ve provided a few projects that piqued my interest, but this barely scratches the surface of all of the things that were talked about or showcased at Consensus this year. I have high hopes for the industry, and I think it is trending in the right direction. With AI getting further enmeshed, things are getting really interesting.
It is clear with the GenAI summit that AI interest and collective attention is the highest it has ever been, while the web3 space continues to putter along. Both will need each other to succeed in the future, as their offerings complement each other. There aren’t perfect solutions for anything yet, but I think we can all agree that we want a future that is privacy-preserving, egalitarian, and one where we collectively own the future.
That’s all for this week. If you enjoyed this piece, let me know. If you want to talk about more projects, AI, web3, or spatial tech, reach out!
Have a great weekend.
Chris