Tech Brew Ride Home - Fri. 03/10 – Silicon Valley Bank(Run)

Episode Date: March 10, 2023

There’s really only one story right now. The troubles at Silicon Valley Bank and how it has literally everyone in the startup ecosystem worried right now. It’s a classic bank run, people. Plus, Meta is developing a Twitter clone. And, of course, the weekend longreads suggestions. Links: This just broke, FYI: Silicon Valley Bank is shut down by regulators, FDIC to protect insured deposits (CNBC) Meta is building a decentralized, text-based social network (Platformer) Weekend Longreads Suggestions: In AI, is bigger always better? (Nature) Can Xerox reinvent itself for another 100 years? (The Verge) Why Are So Many Guys Obsessed With Master and Commander? (GQ) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the tech meme right home for Friday, March 10th, 2023. I'm Brian McCullough today. There's really only one story right now in tech, the troubles at Silicon Valley Bank, and how it is literally everybody in the startup ecosystem worried right now. It's a classic bank run, people. Plus, meta is developing a Twitter clone and, of course, the weekend long-read suggestions. Here's what you miss today in the world of tech. For months, crypto has been doing bank runs like they've been going out of style. And I guess Silicon just got a little jealous. Silicon Valley Bank is a bank that a lot of startups do business with because a lot of VCs have done business with them for decades. And when your VC says, hey,

open your bank account here, you tend to do it, the better for your VC to, you know, wire you your money. In theory, this has always been the way of things in Silicon Valley because traditional banks traditionally shy away from doing business with risky startups. But Silicon Valley Bank is literally the regional bank of Silicon Valley. So, you know, they've done the banking that the major industry in their region has required. More on that in a second. So, funny thing, I had been hearing whispers about Silicon Valley Bank being maybe not so sound for a couple months now, but it was nothing solid, just a few sort of not prominent accounts wondering about how they were doing business. I know it's easy to say that in retrospect, but it is true. Maybe I should get more

in the habit of talking about rumors. But again, I'm not a journalist. I'm just an analyst. really. I don't break news, but apparently I wasn't alone in hearing these rumors. Apparently, VC firm Green Oaks, among others, has been warning their founders of potential problems at SVB since November. Apparently about a dozen of their portfolio companies withdrew around a billion dollars in recent months. And then earlier this week, Silicon Valley Bank announced it was basically dumping a huge portfolio of its assets all at once, which is the sort of thing a bank does when it's trying to raise cash to cover deposits. It also launched a $2.25 billion share sale after losing around $1.8 billion on the $21 billion or so of U.S. Treasuries and mortgage-backed securities it

held. Again, not stuff that a healthy bank does. So, I don't know, maybe Silicon Valley Bank got wrong-footed by all the Fed rate hikes, I guess. Anyway, the stock plunged and at the same time a bunch of VCs yesterday started sending emails to their portfolio companies, urging them to take their money out of Silicon Valley Bank Tutte Sweet. I know this happened because I've seen the emails. I've been sent the emails, though I've not actually sent out any of those emails myself. It would be ironic. If VCs bankrupt a bunch of startups because they rushed to try to save a bunch of their own startups, but I guess that's the definition of a bank run, right? You try to get your money out first and everybody has the same idea all at the same time. So Silicon Valley has its bank run now,

just like crypto has had over and over again these past few months. Though crypto has, it should be noted, been crashing, I guess in sympathy, so there's that. And there are folks in the Valley being like, let's calm down. If we all just relax and keep our money where it is, then a bank run technically can't happen. It's basically that scene from the movie, It's a Wonderful Life with a tech version of George Bailey basically saying, I don't have your money right now, Sally. It's in Tom's SaaS startup, and it's part of Davindra's Series A round. Be thankful I didn't do a Jimmy Stewart impression for you. Anyway, as I write these words, Dan Premack is saying that Silicon Valley Bank's stock has been

halted in trading. I'm assuming that that means that this might work out, that there's a sale of assets, or maybe the bank itself, or a bailout or something being arranged. like I'm not sure that the U.S. would stand by and let Contagion hit the most vibrant sector of its economy. Maybe a forced sale to J.P. Morgan or something like that, but who knows? Maybe the worst happens, and this wipes out a bunch of startups. I'm not being glib about it. It's more gallows humor, my mood right now. Which brings me back to Silicon Valley Bank being a regional bank. Let me tell you a story. In the early 2000s, I was living in Ann Arbor, Michigan, and I was founding my second startup. I got a small business loan to do so because this was right in the teeth of the bubble bursting and there wasn't a lot of

VC cash floating around and I got the loan from Bank of Ann Arbor. In retrospect, it's insane that they gave me the loan because again, this was right after the dot-com bubble, but I did have my first business, a profitable business in the background backing everything up. So it all worked out. I got the loan. I did the startup. In the end, I paid off the loan. Everyone was happy. But it was insane dealing with Bank of Ann Arbor. You couldn't do simple things like log into your account and just move money around. The amount of times I had to pick up a phone or even go in person to the bank to conduct business was crazy. At the time, I thought, hey, you know, they're a small regional bank. You can't expect them to do things a bigger bank can do. Fast forward to the last decade of me doing business regularly with, Silicon Valley Bank, and I always weirdly had the sense when I was doing so that I was still actually doing business with Bank of Ann Arbor. Have any of you out there done business with Silicon Valley Bank? Have you tried to, I don't know, log on to their website? I'll stop right there.

The amount of times over the years I've been unable to just log into my account and had to be like, gee, I guess I'll try later. Or have you ever tried to do something simple like send a wire? The Silicon Valley Bank website, again, the website for the bank of Silicon Valley for the tech industry, has always existed in some sort of liminal space where the great advances of Web 2.0 technology, simple things like, I don't know, Ajax and CSS simply apparently hadn't been discovered there yet. They recently, as recently as this very month, revamped their website, and they had an action list for me of like 18 things I had to do just to enable the handover of my own banking services to their new banking services. They literally had like two hours

worth of work for me because of a change they were forcing on me. It was great. Shoot, long-time listeners of this show will remember when Silicon Valley Bank had ads on this show. And yes, advertising on podcasts is still a very personal, very hands-on, very 20th century sort of style business, but I never had a harder time just getting an ad campaign running on this show in the five years of doing it than I did with Silicon Valley Bank. It literally took three months. Every time I'd be on the phone for an hour getting sign off from one group, I'd have to jump on another set of calls with another group, then another, then another, then another. All along, I've just assumed that Silicon Valley Bank was basically what Bank of Ann Arbor was, a small regional bank,

and I shouldn't expect to be able to do modern banking things with them. Heck, I've always assumed that the reason why places like Brex and Mercury and all those neobanks have sprung up in recent years was because they saw what I saw, which was just you could do banking for startups in a modern way with like, you know, modern UI and such. All you had to do was do the things Silicon Valley Bank seemed incapable of doing. So what I'm suggesting here is this. I guess I was right. Silicon Valley Bank is just a regional bank. Maybe a little too tangled up in the industry of its region. Maybe it wasn't a very good idea for every startup in the world to do business

with the same bank. Maybe a tad too parochial for an industry based in the future. Just asking questions here. Or maybe I'm being too harsh. You know how Warren Buffett famously said that the airline industry has never made money in its entire existence because every couple of decades, their whole business blows up and they have to be taken over and bailed out and reorganized. Well, let me tell you about the banking industry going back to, I don't know, the Hundred Years War, bankrupting the Renaissance banks of Italy because the Kings of England defaulted on their loans. Congratulations, Silicon Valley. I hope not, but it looks like you now have a bank run to call your own. As I type these words, news is coming across the transom that Silicon Valley Bank has hired advisors to try to do a fire

sale to find someone to buy them or bail them out. Yes, tech industry, you too, can have a Lehman moment. I hope not, but it sure looks that way right now. There have also been rumors for months that Meta was going to take a shot at doing its own Twitter clone. Well, Casey Newton is reporting that Meta is indeed exploring a standalone decentralized social network for sharing text updates. The app, codenamed P92, will reportedly let users log in with their Instagram accounts, quoting Platformer. Details about the project are scant. The product is still in its earliest stages, sources said, and there is no time frame for it being released. But legal and regulatory teams have already started to investigate potential privacy concerns around

the app so they can be addressed before launch, we're told. Adam Maseri, who runs Instagram, is taking the lead on the project. Sources said, the most remarkable aspect of the project is that meta plans for the network to be decentralized. While the company would not elaborate beyond its statement in a decentralized network, individual users are typically able to set up their own independent servers and set server-specific rules for how content is moderated. Building a decentralized network could also give META the opportunity for its new app to interoperate

with other social products, a previously unheard-of gesture from a company known for building some of the most lucrative walled gardens in the industry's history. Building a decentralized social network could let meta experiment with an app that pushes back on standard criticisms of Facebook and Instagram. Individual servers would let different groups set their own community standards, though likely with a floor of rules set by meta in a fashion similar to how Reddit's individual communities work. And making its network interoperable with others could help meta as it faces ongoing scrutiny over whether it has maintained its market power through anti-competitive acquisitions.

Still, no one has built a profitable global-scale decentralized network till now. The Fedaverse had around 2.6 million users last month a rounding error for a company like meta, which has more than 2 billion daily users across its family of apps, end quote. Abbreviated weekend long read suggestions for you today because today was so newsy. But first up, from nature, a pretty important question right now. As generative AI models grow larger and more powerful, is bigger always better? Like, is this how this is going to go? Just bigger models, trillions of data points, or is there another way to go?

Could you do lean AI? quote, large language models such as Chach EPT and Minerva are giant networks of computing units, also called artificial neurons, arranged in layers. An LLM's size is measured in how many parameters it has, the adjustable values that describe the strength of the connections between neurons. Training such a network involves asking it to predict masked portions of a known sentence and tweaking these parameters so that the algorithm does a little better next time. Do this repeatedly over billions of human written sentences, and the neural network learns internal representations that model how humans write language. The more robustly AI,

the better it generalizes to unseen data. Bubeck and his colleagues have shown mathematically that increasing the number of parameters in a model increases robustness and hence ability to generalize. The law proves that scaling up is necessary for generalization, but not that it's sufficient, says Bubeck. Nonetheless, it is being used to justify the move towards bigger models, he says, which I think is a reasonable thing, end quote. Even when LLLL get the answers right, however, there is no understanding involved, says Charlotte. When you try to probe it a little bit, it becomes immediately obvious that it's all empty. ChatGPT has no model of what it is talking about, he says. You're watching a puppet show and believing the puppets are alive,

end quote. So the article goes on to note the logical endpoint of all this is just larger and larger models consuming ever greater compute resources and also crucially power. One of the arguments for building a Dyson sphere in science fiction is that if your AI gets powerful enough, you're going to need to harness all of the energy of your star or even your galaxy just to power it. That's one of the proposed solutions to the Fermi paradox, by the way. Anywho, say hello to new theories of smaller, more efficient AI. Not only would that cut off the power problem, but it might also be the innovation necessary to make artificial general intelligence possible. Quote, for comparison, our own brains are much more complicated and larger than any LLM with 86 billion neurons and some 100 trillion synaptic

connections. And yet the human brain consumes somewhere between 20 and 50 watts of power, says Friedman Zenkie at the Friedrich Meischer Institute for Biomedical Research in Basel, Switzerland. Meanwhile, researchers are experimenting with different ways to make existing LLMs more energy efficient and smarter. In December 2021, DeepMind reported a system called Retro, which combines an LLM with an external database. The LLM uses relevant text retrieved from this database during inference to help it make predictions. DeepMines researchers showed that a 7.5 billion parameter LLM coupled with a database of 2 trillion tokens outperforms LLMs with 25 times more parameters. The researchers wrote that this was a, quote, more efficient approach than raw parameters scaling as we seek

to build more powerful language models, end quote. In the same month, scientists at Google Research reported another way to increase energy efficiency at scale. Their generalist language model, or glam, has 1.2 trillion parameters, but these parameters don't represent one giant neural network. Internally, they are distributed between 64 smaller neural networks alongside other layers. The LLM is trained such that during inference, it uses only two of its networks to complete a task. Overall, the LLM uses only about 8% of its trillion plus total parameters for inference per token. According to Google, Glam uses the same amount of computing resources as were needed to train GPT3, but consumed only about one-third of the power because of the improvements in training software and hardware.

During inference, Glam used half the computing resources that GPT3 needed, and it outperformed GPT3 when trained on the same amount of data. To improve further, however, even these more energy-efficient LLMs seem destined to become bigger, using up more data and compute. Researchers will be watching to see what new behaviors emerge, with scale, quote, whether it will fully unlock reasoning, I'm not sure, says Bubeck, nobody knows, end quote. Next, Nilai Patel has a fun conversation with the CEO of Xerox on the occasion of its 100th birthday, talking about printing, IT services, automation, the Palo Alto Research Center, Care AR, Mergers and Acquisitions,

and more. Quote, I started with a couple of things. First, we are Xerox. There is a pride in the heritage of who we are and where we came from. It's not just technology, it's not just printers, and it's not just our products. It's our employee resource group. It's the Palo Alto Research Center that has created trillions of dollars of value around the world. Not necessarily just for Xerox, but for other companies as well. They're very famous in Silicon Valley for helping companies with products and services.

You go to Silicon Valley today. It's still one of the most proud locations that people visit. It's our proud heritage of innovation, end quote. And finally today, on the 20th anniversary of the movie Master and Commander, GQ takes a look at why so many people like me, dads of a certain age, are so obsessed with that movie. Quote, any nostalgia stirred up by Master and Commander is also nostalgia for a different era of Hollywood. This sort of richly detailed big-budget historical epic rarely gets a chance in today's movie landscape. And even if the action isn't the point, the battles absolutely kick ass, using practical effects that would probably be weightless CGI these days. They bought a ship in Rhode Island and sailed it through the Panama Canal and a hurricane to a six-acre filming tank in Mexico.

Nando Villa, the head of Studio at Exile Content Studio, told me, quote, I think why a lot of guys are liking it now is because Aubrey is so charming and swashbuckling and swaggery. You believe that all these sailors are into Lucky Jack and they'll follow him to the far side of the world. You don't see that kind of brawny. We're just going to fucking go to the far side of the world. Who's with me? That's not a movie that gets made anymore, end quote. Has anybody else noticed that when Matt Levine announces he's taking a day off from his newsletter, shit immediately goes down?

Like, the most recent edition of Money Stuff hit my inbox at like 223 yesterday, with a note at the top about how Matt was going to plan to take today off, and, you know, wave's hand, look at what happened. If you liked my Silicon Valley Bank segment today, maybe tweet this episode out. Maybe I can slip into Matt Levine's shoes just this once if he really does take today off. We've got one bonus episode for you this weekend. It's a portfolio profile episode. What if I told you we've invested in a company alongside Andresen Horowitz, Founders Fund, Thrive Capital, Coastal Ventures, General Catalyst, just a murderer's row.

Find out exactly what sort of company could attract all those names and even the Ride Home Fund. By the way, if you want to invest in companies like this, remember, you can join the rolling fund at any time at right home fund.com. Also, this founder maybe got the most personal about how his passion for a particular vision drove him to start this company. It gets sort of deep, actually, so enjoy that. Talk to you on Monday, assuming the tech industry still exists by then.