- Meta’s ulterior motive with generative AI
- Apple’s mixed reality headset is right around the corner
- The Court will come to order… in a metaverse
Dear Reader,
The last 72 hours have been an absolute whirlwind.
I’ve never seen the early stage tech community in complete panic mode like the last three days. I was in contact with numerous founders over the weekend in an effort to sift through all of the information and help with damage control.
The issue at hand was the sudden demise of Silicon Valley Bank, known as SVB, the 16th largest bank in the U.S. It is now known as the second largest bank collapse in history.
And it was woven into the fabric of Silicon Valley. More than 2,500 venture capital firms banked there, even more early stage private companies, not to mention an even larger number of individuals from the industry.
Because so many individuals and organizations were tied to SVB, its collapse is catastrophic. SVB literally experienced a bank run. There were lines of people waiting to get their money out of the bank.
While the $250,000 FDIC insurance provided some protection, 93% of all SVB deposits were above the $250K limit. That meant that anything beyond that could vanish without some kind of emergency support.
Other investment banks were scrambling over the weekend to assess whether or not a buyout of SVB was interesting and at what price. Many were hoping that a deal would happen and ensure that all depositors’ funds were kept whole.
But that didn’t happen and the lender of last resort, the U.S. government, had to step in to stop an all-out collapse of the U.S. banking sector. After all, if it can happen at the 16th largest bank, it can happen elsewhere.
Specifically, the U.S. Treasury, the Federal Reserve, and the Federal Deposit Insurance Corp. (FDIC) stepped up to guarantee that all depositors at SVB will be made whole. It also outlined a Bank Term Funding Program providing $25 billion worth of one year loans to banks where needed.
What the heck went wrong?
Did the execs at SVB misappropriate customer funds and buy yachts or invest those funds in highly risky assets? Is this another FTX?
No, not at all.
In fact, SVB, and many other banks invested the vast majority of their assets into U.S. Treasuries, about as close to a risk-free asset class as anyone can get. This is what makes SVB’s collapse that much more incredible.
After all, if the assets are in rock solid U.S. Treasuries that are guaranteed to pay a yield and return all of the capital at maturity, it’s hard to imagine how anything can go wrong.
But it did. And the problem was caused by duration risk.
Like most banks, SVB maintained a certain level of desired liquidity in short duration, low yielding available-for-sale (AFS) securities. And like most banks, a large portion of assets with higher yields are designated as held-to-maturity securities. They function exactly as their name, intended to be held until maturity when the par value of the Treasury bonds is returned to the bank.
In the case of SVB, the large majority of its Treasury bond portfolio was considered held-to-maturity. Held-to-maturity securities tend not to be marked to market, in other words, even if the value of the bonds decline, it doesn’t matter because the bank intends to hold until maturity. Overtime, it is guaranteed not to lose money regardless of any volatility in the bond’s price.
But it does matter in banking.
As a public company, everyone could see the size of SVB’s long duration securities. And when interest rates climb, the value of a Treasury bond drops accordingly. By September last year, SVB’s marked to market losses on its long duration bonds was $15.9 billion compared to its tangible common equity of $11.8 billion.
If those long duration assets had been marked to market, the bank would have been considered insolvent.
The SVB execs intended to raise additional capital to provide short term liquidity, but something happened that made it difficult to do so. Deposits fell to $173 billion at the end of December, down from $198 billion in the first quarter, and by the end of February deposits had dropped to $165 billion.
And that’s the pickle that SVB found itself in.
If it sold its long duration securities it would have to do so at a massive loss, which would exacerbate the liquidity crisis. And it was out of short duration assets to sell. In the absence of a liquidity injection, SVB was done.
A lot of fingers are being pointed right now at SVB and its executives. Yes, it’s possible that they could have raised additional capital much earlier. And yes, it’s possible that they could have sold off some of their long duration Treasuries at losses earlier and shifted to shorter duration. It’s not clear at all if that could have been done without the same result; but they were options. But there is a much larger culprit at play here.
The Federal Reserve broke SVB.
That’s right. The Fed hiked rates 450 basis points in less than 12 months at record speed. It did that after printing more than $5 trillion in stimulus as the U.S. government continues to run multi-trillion-dollar deficits.
What did it think was going to happen?
I’ve been writing since last summer that the Fed and the U.S. government has been aggressively hiking rates with the intention of breaking something. Clearly, it’s already happening. The mortgage market, and thus the housing market, is collapsing, the auto loan industry is falling over, and now banks are collapsing, all caused by the current monetary policy of the Fed.
SVB is the largest and most dramatic, but First Republic Bank was on the verge of collapse this weekend before the Fed and JP Morgan stepped in with some additional capital to provide liquidity. And Signature Bank in New York collapsed over the weekend as well with the government also guaranteeing to make depositors whole.
We’re being told that taxpayers won’t have to pay a dime for this mess; but that’s simply not true. The Federal Reserve (the government) will have to print more money to pay for the losses that it caused through irresponsible fiscal, monetary, and economic policy.
And that means devaluation of the U.S. dollar.
Meta’s false attempt at magnanimity…
We had a look last year at how Meta (formerly Facebook) has been working on its own large language models. Their purpose is to train generative artificial intelligence (AI) and apply that technology across its social media empire.
Well, Meta just did something unusual. The company just released one of its models, including the entire source code, to select institutions for free.
This stands in sharp contrast to what Google, Microsoft, and OpenAI have done.
In Google’s case, it hasn’t even made its AI software available… much less the source code. And in the case of Microsoft and OpenAI – they have allowed users to access their generative AI… but they have kept the source code to themselves.
Companies and institutions are only able to connect with their generative AI using application programming interfaces (APIs). That means that they can use it for a fee, but they can’t see how it works and they can’t modify the software.
So Meta is going above and beyond here. It’s allowing select institutions to use its generative AI. And those same institutions can pour through the underlying code as well. That means they are free to make edits and additions to create a modified version if they so choose.
But there’s a nuance here…
Meta isn’t releasing its latest and greatest language model. Instead, the company is making a smaller model available. It’s actually a smart move.
To be fair, the older language model is no slack. Meta claims it can outperform OpenAI’s GPT-3.
And get this – Meta’s model can run on a single NVIDIA graphics processing unit (GPU). It doesn’t require massive computing power. That means that for probably less than $10,000 of hardware, any research organization or institution can work with this technology.
Of course, this begs the obvious question: Why would Meta do this?
The answer is simple. Goodwill.
Meta is making its technology available to government departments and academic research labs by request. In doing so, Meta is building goodwill with the government, academia, and other influential organizations.
That could come in handy down the road… especially if Meta comes under additional regulatory scrutiny for its business practices again. And it could open doors down the road if government agencies want to integrate generative AI into their systems.
We should keep an eye on this. Meta never does anything purely out of the goodness of their heart.
This is not a magnanimous organization, as we’ve learned its motives are pure profit and ultimately “mind control” of its users. There’s another play here. By giving the software away for free, it can supplant the use of its heavily biased algorithms in various institutions and achieve even wider adoption.
The latest leak around Apple’s augmented reality headset…
An interesting leak just came out of Asia. It’s regarding Apple’s upcoming augmented reality (AR)/virtual reality (VR) headset.
As a reminder, Apple is referring to this technology as “extended reality” (XR). And it’s first high-end XR headset is rumored to be called Reality One.
And it turns out a company called Luxshare Precision Industry is already working to manufacture the Reality One. It’s in production right now.
I’ve been predicting that we’d see Apple announce its new XR headset in the May/June timeframe … and that prediction is looking better than ever. If the device is already in production, it’s only a matter of time before Apple spills all the beans.
I doubt anyone will recognize the name Luxshare Precision Industry. But this company acquired Pegatron a little while back. And Pegatron is the company that has historically manufactured a variety of Apple’s products like the AirPods.
What’s more, Pegatron was working on Microsoft’s AR headset, the HoloLens. Had Microsoft not scrapped the HoloLens, Pegatron would have been its manufacturer.
So Luxshare, with the addition of Pegatron, is the logical choice to produce Apple’s Reality One product.
The initial product launch will reveal a high-end device. It will retail for about $3,000. From there, we can expect Apple to release less expensive, mass-market versions of the XR headset into 2024/2025. That’s when augmented reality will go mainstream.
And there’s an important reason why Apple is taking the plunge into augmented reality…
The XR device will provide a far more immersive experience than our smartphones. We’ll be able to do everything on the XR headset that we can do with our smartphones today.
That’s why I’m convinced that augmented reality will be the next mass-market consumer frenzy. And one day it very well may replace smartphones entirely.
And for Apple—a company that still derives roughly half of its revenue from iPhone sales—that will be a problem if they don’t adapt.
That’s why Apple is going full steam on augmented reality. Just as it did in the smartphone industry with the iPhone and its iOS operating system, Apple wants to set the gold standard for AR/VR/MR technology with its Reality One hardware/software.
The world’s first court hearing in the metaverse…
We’ll wrap up today with an interesting development. The world’s first court hearing in the metaverse just took place a few weeks ago.
It happened in Columbia. The judge used Meta’s Horizon Workrooms metaverse to host a two-hour session.
And they went through the same processes as they would if the case were held in a court room. The only difference is that participants used digital avatars to represent themselves.
Given that this is a historical first, I took some time to think about it more deeply. And I believe it has some interesting and lasting implications.
We may ask – why do this online? And why use digital avatars for such a serious situation?
To answer this we have to think back to what’s happened over the last several decades…
Back before the internet, whenever we had a problem – business or personal – we usually resolved it by having an in-person meeting. Those discussions could be uncomfortable at times. But back then, that’s how issues were handled.
Over the years, in-person meetings gradually gave way to phone calls. Then over the last ten years or so, phone calls gave way to email… then to online chat services. And more recently, people often take to social media to air out issues in front of anyone who would pay attention.
This is a cultural evolution.
Our society has gradually been migrating towards methods of communication that require less and less personal presence. It’s a kind of disintermediation, intentionally creating barriers between two parties. And we’ve de-humanized these interactions over the years.
That’s why I think we’ll see more and more activities take place in a metaverse. It’s just the next step in this cultural evolution. While the court hearing itself may be unavoidable in certain circumstances, this approach using avatars in a metaverse will likely be preferred by many who are uncomfortable in what could be a stressful situation.
Plus, if we think about a court hearing, people can have all their documents laid out in front of them for reference, and nobody will know they are using their notes. That could be a benefit.
And while I don’t think this is a healthy dynamic, and it will likely lead to even more societal decay in resolving issues peacefully and professionally. It seems unavoidable to me given that the technology is now widely available.
And it’s worth mentioning that hosting events and proceedings in a metaverse is far less expensive than doing them in-person. A savings in time, money, and a more comfortable environment is going to be a strong argument for adoption.
Regards,
Jeff Brown
Editor, The Bleeding Edge