Category Archives: AI

Crypto and AI

Posted on by
Reply

There have been dozens of “crypto and AI” posts and essays published over the past year.

At first (and second) glance, most of the conversations seem to be a bit like “Chainwashing” from a few years ago: shoehorning one popular trend into another.

Are there any legs to the overall theme?  After all, we have had “AI” in TradFi for years; is this variation any better than continued human involvement?

Based on my own experiences and interactions with experts, I think there could be at least three genuine areas of overlap:

  1. Search and model training (e.g., Venice)
  2. Automated on-chain agent interactions (e.g., trading, market making bots)
  3. Trusted execution environments (e.g., TEEs used by Oracles and Validators such as Switchboard)

We could argue over the utility of each category, whether or not a blockchain is needed or helpful at all in the lifecycle of the activity.

One counter-argument is that in each of these areas: traditional, centralized infrastructure is more efficient (quicker) at achieving the end-goal for the users.  But I think – like past cases involving any blockchain – the merits and demerits of the infrastructure usage is specific to the circumstances.  That is to say, claiming a priori that X is better than Y because it is faster or cheaper lacks nuance.  There are tradeoffs with centrally owned and operated infrastructure that may push certain uses to decentralized infra.

CoinGecko recently had a poll surrounding the (2) topic, asking participants: which crypto AI agent use case(s) are you most excited about?

  • KOL/Influencer
  • Executing buy/sell orders optimally
  • Market intelligence & recommendations
  • Trading automation
  • Investing automation
  • DeFi automation
  • Onchain data analysis
  • Auditing smart contracts
  • Detecting potential hacks or scams
  • DAO governance assistance
  • Chat
  • Gaming/Metaverse NPCs

We don’t have the space to dive into each of these, but I think one category that merits inclusion in the future combines investing and trading automation with market intelligence: prediction and forecast modeling (e.g., Ocean Protocol).

With that said, I do think there is a bit of unwarranted hype around a number of “trading agents” efforts which seem to rely on getting unsophisticated traders to dogpile into zero sum outcomes.  For more on that topic, I think readers will be interested in a newly published report from Delphi Research: The Battle of the AI Agent Frameworks.

How fast will these niche areas grow over the next year?  Will there be meaningful ‘normie’ adoption or will it continue to be dominate by a handful of participants sybilling?

Automated prediction markets

Posted on by
Reply

There are a couple of different ways to categorize prediction-markets as they exist today.  One way to slice them is based on their infrastructure:

Centralized:

  • One of the longest lived AI-driven ‘prediction’ markets today is Numerai.  It’s contests have been a gravity well for data scientists for nearly a decade.
  • Another arena for professional and amateur modelers alike is Kaggle, from Google.
  • Metaculus caters to a similar intellectual community: quants that are attempting to model the future.
  • Kalshi and PredicIt (and a slew of imitators) which provide capped bets and wagers for a variety of events (the imitators often focus on sports betting).

Decentralized

  • Polymarket, dYdX, and Drift currently have the largest marketshare of ‘prediction’ markets in the blockchain universe.

Infrastructure differences aside, what commonality do all of these projects and organizations all share?  Contract issuance and trading are silo’ed.  Even blockchain-related platforms mimic the existing TradFi futures exchange model (e.g., can only trade WTI on Nymex).  One other difference is that a prediction market (as they are often advertised today) is based around a specific, discrete event (such as an election or a football game), whereas a forecast contract or forecast market is based on a reoccurring event / time series (such as CO2 emissions, unemployment rate, inflation, etc.).

What seems to be missing are tools for forecast modelers to be able to take an arbitrary time series and construct a futures product that continually rolls over (e.g., a perpetual).  This can be done in centralized or decentralized manner but to actually be decentralized, the clearing of the contract arguably should also be separated from the trading venue (crypto assets ‘took off’ because they could be traded independent of venue, yet all crypto perpetuals are venue dependent).  Will 2025 be the year in which such a decoupling finally takes place?  Has anyone created the GPT for these types of contracts and markets?

At the moment, I think one of the biggest unknowns for a trader entering into forecast futures is the question of how they would hedge their exposures.  E.g., a market maker for a futures market today can always construct a hedge using the spot market… but what if the spot does not exist?

Non-technical Prompt Engineering and GPU count

Posted on by
Reply

Creating (and refining) prompts used by LLMs has been a side-hobby for the past year.  Below are three non-technical prompts that illustrate using an LLM to answer casual questions.

For instance, my young daughter currently plays hockey. Last March one of her coaches mentioned that there were outsized scholarship opportunities from NCAA Division III colleges for women’s hockey. To start fact-checking that statement, I began with a prompt:

There are 67 teams in NCAA Division III women’s hockey. Please list them in order based on the school rankings from U.S. News & World Report.

At the time, ChatGPT 4 said it did not have access to UNSWR or Forbes and couldn’t complete the ranking.  

Fast forward to today, while ChatGPT 4o still does not have direct access to these two ranking indices, it did pivot by asking if it could provide a different sorting (although not by academic ranking).

During the summer, when Ask Venice launched, I tried asking what the etymology of a “no-coiner” was.  While it didn’t provide the history of the phrase, it did provide some footnotes which at the time few LLMs did.  Fast forward to today, the same question still does not provide a satisfactory answer (e.g., it lacks history: it doesn’t say who the first people were to use the term).

This past fall I queried:

What was the participation in the PSAT / SAT over the past 10 years. And, assuming birth rates / immigration trends continue, what the participation rate would be the next 15.  

Following those I asked what the previous participation in the AMC 8, AMC 10, and AMC 12 were the past 10 years. And, likewise, what the trend lines look for the next 15 years.

ChatGPT 4 drew some decent line charts to visualize those trends.  The motivation for this were various parents discussing the rat race of cram schools in the Boston area during a time in which some matriculation numbers for the city are declining.

A number of other casual searches involved finding local activities for my daughter and even writing or reading contests.  The answers typically are often better than what Google provides.

The refinement process for prompts is probably something that will not go away even with more accurate reasoning models (such as DeepSeek R1) are deployed.  Does that mean people being paid $200k a year for their help chatting with bots in 2023 translates into a long-term moat for creative prompt writing?  No, but there does seem to be an art for conjuring and revising them at this time.  

Counting GPUs

We have discussed the flexibilities of GPUs over the years… I even wrote a “How-to” guide for mining Dogecoin over a decade ago.

I saw a germane tweet this past week that piqued my interest in part because it was a little inaccurate:

Source: David Holz

Someone is (a little) wrong on the internet!

My quibble, and one that was echoed by a couple of others in that thread, is that there are a lot more than 5 million GPUs manufactured each year.  For instance, in Q1 2024 around 70 million discrete and integrated GPUs were shipped by Nvidia, Intel, and AMD combined.  These were for both desktops and laptops.  In Q2 2024, approximately 9.5 million discrete GPUs for desktop PCs (not laptops) were shipped by the same three manufacturers.

And one other GPU segment that David appears to have missed in his calculation are video game consoles.  For example, more than 60 million Playstation 5’s have been sold since 2020.  Likewise around 31 million Xbox Series X and Series S have been sold in that same time frame.  While neither of these are particularly powerful relative to a new GPU from Nvidia today, when they were first released they were all considered high performance.

As LLMs become more accurate, with better reasoning abilities, a scenario that David and others rightly ask is: wouldn’t the demand for GPUs outstrip the supply?  That is to say, if an AGI (however defined) is capable of running on a discrete GPU, how many people with the means, will purchase and install one at home versus renting them off a cloud provider?  Maybe the TAM for on-premise AGI would be as large as on-premise e-mail users?

Let’s check in again on this in a couple of years to see what level of artificial intelligence can run off the forthcoming 5090 and its successor.  Hopefully they’ll be more accurate, because during the drafting of this post, ChatGPT 4o used outdated video game console numbers (without footnotes!) that I eventually tracked down with… Google.

And if you have access to the newly released DeepResearch, feel free to leave a comment about what it assesses are the merits and demerits of Jeffrey Emanuel’s essay on Nvidia as well as the energy and resource usage estimates from Rohit Krishnan.