Encrypted Life: A Day in the Life of a16z Crypto Partner

Original Title: “Info Diet: a16z Crypto's Elena Burger”

Author: Elena Burger, a16z

Compiled by: Kxp, BlockBeats

Info Diet invites different builders to document the articles they read and browse over the course of two days from their own perspective. Today, we are fortunate to have Elena Burger, a trading partner at a16z Crypto, who currently resides in New York and focuses on the fields of NFT, Web3 media, and infrastructure.

Sunday, July 10

10 AM: For me, this is an unusually lazy day. Normally, I wake up around 7 or 8 AM, have a cup of coffee, go for a run in Central Park, and then check Twitter and my emails. But today, I happen to be visiting a family member's beach house with friends, so I’m not as enthusiastic as usual this morning.

I went to the kitchen for a sip of coffee and casually flipped through an issue of The New Yorker that was on the table. This is also something I don’t usually do: I might have read a lot of print media before, but now I prefer digital versions. As people flooded into the breakfast room, I was looking at a profile of gospel/blues singer Mavis Staples.

12 PM: On the Jitney bus back to Manhattan, my friend shoved one of his AirPods into my ear and asked, “Have you seen Channel 5?” I said I hadn’t.

Channel 5 is a street interview show by YouTuber Andrew Callaghan, who always chooses controversial groups (pickup artist camps, anti-vaccine rallies, etc.) and films their daily activities. We watched an episode from the rap festival he covered (I suggest you turn the volume down a bit before clicking the link), which featured several segments of a 15-year-old prodigy rapping. I really enjoyed this episode; it’s somewhat similar to shows like “Rescue Me” or “How To with John Wilson.” Although I don’t watch YouTube much, just like with the profile on Mavis Staples, if it’s right in front of me, I’ll still watch it with great interest.

Then I put on my AirPods and listened to Lex Fridman’s interview with Grimes from a few weeks ago. After listening, I found myself really falling in love with her. I didn’t know she had such deep knowledge of Crypto, and she has a very tech-utopian view of Web3. She mentioned using DAOs to support all mothers and open-source research, and talked about how Web3 or AI could replace many intermediaries. To me, it’s fascinating that artists can explore more ways to use Crypto compared to many people in finance or traditional tech. After listening to her interview, I’m now also curious about Mavis Staples’ views on DAOs.

Additionally, in the podcast, Lex mentioned that he lives next to a well-known anarchist. This made me wonder: if you are an anarchist and you do own private property, what anarchist principles are you actually upholding? Anyone who wants to discuss this topic with me is welcome to reach out.

3:30 PM: After I got home, I opened the SITALWeek newsletter on my laptop, which I read every Sunday. This newsletter is compiled by Brad Slingerlend, who focuses on topics in traditional finance and tech: AI, semiconductors, the streaming economy, inflation, and oil and natural energy. As someone who spends most of my time thinking about Crypto, this is a great “knowledge-gaining” time for me.

This issue has an article about the impact of robotic surgery on the skill levels of medical students, another discussing China’s dominance in the lithium-ion battery market, and some links to articles about the urgent need for streaming services to tighten their budgets.

In my view, even though this isn’t a “Crypto” newsletter, I always find ways that Web3 can learn from some of the articles within it (especially in the streaming/online media space). In terms of user preferences and stickiness, it seems we’ve reached the limits of what large streaming companies can model effectively and accurately, while newer business models in Crypto can allow more artists to enjoy a larger share of the value they create, while also allowing communities to share intellectual property outcomes.

4:30 PM: I went out for a walk and called my brother. He’s an electrical engineer, and for the past few months, I’ve been trying various ways to get him into the Crypto space. Recently, I might have gone a bit overboard because the first thing he said when he answered the phone was, “Elena, I’m not coming to the Stanford Blockchain Conference.”

We started talking about recent developments in AI and image generation, and we also discussed DALL-E, which I find very interesting lately. My brother sent me an article from Astral Codex Ten titled “A Technical Guide to Designing Stained Glass Windows with Robots,” where the author describes the entire process of guiding DALL-E to gradually present the images he envisions.

When I use DALL-E, I sometimes give specific prompts, and other times I intentionally provide it with open-ended, conceptually vague information. For example, a few weeks ago, I asked it to generate a scene of “Edison selling pirated DVDs on the street in the 90s.” The first two images it generated looked like the 90s (the colors were saturated, and Edison was standing next to a pickup truck), but the other four images looked like the 1890s—black and white with a grainy texture, and the DVDs were oddly proportioned and oval-shaped, like something a visual gramophone would produce. This is a creative leap that’s hard for people to achieve, and DALL-E can realize this possibility, intentionally or unintentionally, which really surprises me.

After reading this article, I realized that if I could leverage GPT-3, I could answer my previous question (What would Mavis Staples think about DAOs?), and I could access GPT-3 through OpenAI’s playground page. So I wrote the prompt: Write a paragraph in the style of The New Yorker about Mavis Staples discussing Crypto and decentralized autonomous organizations.

The result was as follows:

Mavis Staples has been singing gospel and soul music since 1950, respected as a contemporary legend and beloved as a performer. Last fall, she released her latest album, “If I Were Black,” a collaboration with producer and musician Jeff Tweedy. Staples is also a venture capitalist, and in early December, she attended a web summit in Lisbon focused on Crypto and decentralized autonomous organizations. At the conference, she stated, “While I’m not a technical expert, I’m working hard to learn.”

I must say, this result is quite impressive. The only inaccuracy I noticed (aside from the fictional element of Mavis Staples suddenly appearing as a VC at the Lisbon Crypto conference) is that her most recent album is “We Get By,” released in 2019. This is likely because the training data used by GPT-3 hasn’t been fully updated yet, but I can’t be entirely sure about that.

7:30 PM: I don’t remember when I logged out of that website, but I’m now reading an article about Minecraft servers. This article is from Rock Paper Shotgun, published when Hypixel first launched in 2015, at which point Hypixel was the largest independent server in terms of concurrent users globally. It peaked at 216,000 players in 2021, with an average of 45,000 to 70,000 players daily. I’m curious how Minetrack quantifies CCU in Hypixel games, as the server guides players into small “parties” to play (so you won’t be surrounded by thousands of players). But I might be mistaken; after all, scaling games to tens of thousands or even millions of CCU is still a technical challenge—Water & Music recently explored this topic in detail in their article “9 Design Principles for the Music Metaverse.”

Monday, July 11

7 AM: I ran a few laps in Central Park after getting up. I only listen to music while running (I tried podcasts for a while, but my mind wanders when I jog), and now my playlist features a lot of 2000s artists like Meek Mill, SOPHIE, and Ryan Leslie.

10 AM: I’m back to reading my “non-crypto” articles, but actually, a lot of news in various fields—chip shortages, inflation, entertainment events, etc.—are related to Crypto. I also enjoy checking Liberty's Highlights, which covers news from various fields: semiconductors, chips, traditional media, etc. Today’s issue actually cited yesterday’s SITALWeek article on surgical automation, along with some articles from The Wall Street Journal and Compound Capital Advisors CEO Charlie Bilello, covering energy efficiency in electric vehicle batteries and inflation in the Eurozone.

8 PM: I spent most of the day on phone calls or replying to emails/Slack messages, leaving little time to browse the news (aside from tweets). I do check a lot of funding pitches and emails from founders, but I only read those in-depth articles at night (or on weekends).

Tonight, I watched the Ethereum Foundation research team’s Reddit AMA from a week ago because I wanted to stay updated on the Ethereum Foundation’s latest opinions, especially as we approach the merge. Then I found that the core developers’ opinions are often not necessarily unified, and there can be disagreements during the research and implementation processes.

The types of applications people are looking forward to in the future, at least according to Vitalik, Justin Drake, and Danny Ryan, should be innovative applications centered around identity, reputation, and privacy.

(I also frequently browse the website “What’s New in Eth2,” compiled by Ben Edgington from ConsenSys. Additionally, when we talk about this topic, it’s necessary to thank Tim Beiko for his very detailed notes during the Ethereum All Core Devs calls.)

1 AM: After dinner, I continued to check information and browse tweets on Twitter. Then, I opened a deep article about Minecraft. Before going to bed, I read an OpenAI research paper about training neural networks to find and create items in games. Impressively, in the video used to train the neural network, only a small portion was labeled with corresponding mouse and keyboard actions, while the rest was just a massive amount of videos uploaded by players online. In other words, the neural network can learn to infer from this small portion of labeled data and apply it to a larger quantity of unlabeled data.

The more I pay attention to what’s happening in the field of AI, the more I want to explore how to combine AI with Crypto. I wonder if there will be more tools and protocols in the future that support on-chain AI, available for anyone to inspect and upgrade, and owned by the community? A few months ago, 0xPARC did a proof of concept demonstrating an on-chain digital image classifier within zkSNARK. Similarly, a few years ago, our team member Ali posted on Twitter about how to launch a self-sustaining neural network on-chain. I hope we continue to see more applications of this type in the future.