• Google isn’t out of the generative AI race just yet

  • A CRISPR pioneer launches a new company

  • Police sketch artists’ days are numbered…

“Only buy something that you’d be perfectly happy to hold if the market shut down for 10 years.”

“Our favorite holding period is forever.”

These two well-known quotes from Warren Buffett stand in stark contrast to his recent moves at Berkshire Hathaway.

In November last year, Berkshire Hathaway disclosed that it had bought a $4.1 billion stake in the worlds largest semiconductor manufacturer, Taiwan Semiconductor Manufacturing (TSMC).

It was an unusual investment for Buffett, who has largely eschewed technology companies to his own peril. But I don’t disagree with him.

TSMC is a fantastic company to own with a time horizon of ten years. It’s one of the most well managed businesses in history and arguably the most important semiconductor manufacturing company on the planet…a reality that Buffett likely realized once he understood how critical TSMC was to one of his favorite positions, Apple.

Which is why it came as such a surprise to find out that Buffett turned around and sold 86% of his position in the fourth quarter of last year. And I suspect that by now, he has already sold off the remainder of his shares. We won’t know for sure until Berkshire Hathaway files its 13F with the SEC for the first quarter of this year, which will happen in a couple months or so.

Clearly, something is afoot.

Berkshire Hathaway doesn’t tend to trade in and out of individual positions from one quarter to the next. That’s not Buffett’s investing style. He buys and holds for long term capital gains. But something clearly changed to cause such an abrupt about face.

And that something is almost certainly China.

Buffett has ties to the most powerful politicians and businesspeople in the world. He has an inside track on what’s going on behind the scenes that informed his investment strategy. And this latest move is telling us something.

He closed out his position reportedly for a very small profit. But that’s not what the move was about. He wanted out. Quickly.

A quick reversal like this by an investor like Warren Buffett tells us that the risk of a move on Taiwan by China is very high. It’s the simplest explanation for the sale. And considering that I’ve predicted that China will make its move on Taiwan within the next twelve months, I completely agree.

In fact, before Buffett’s sale of TSMC became known, we closed out a profitable position in TSMC in The Near Future Report model portfolio.

A move by China on Taiwan will cause TSMC’s share price to collapse. Gaining control over TSMC could threaten not only the entire semiconductor industry, but every company that manufactures products that use semiconductors that are manufactured by TSMC…which is to say most companies.

It’s hard to imagine a more critical throat to squeeze that would have such an immediate and tangible choke hold on the world’s economy. And that’s why we closed our position, and that’s why Buffett did as well. Tensions have increased significantly since the third quarter of last year.

But these geopolitical risks are not the end of TSMC. With crisis comes opportunity. Presuming the takeover happens, we could have a once-in-a-generation opportunity to buy into TSMC. I believe that the collapse in share price would be short lived, and shares would trade briefly at a depressed valuation never seen before.

The reality is that China needs Taiwan’s economic engine to succeed. And while it wants access to TSMC’s manufacturing technology and knowhow, it also needs TSMC to continue to succeed, which can only happen at scale.

TSMC’s scale is what allows it to invest $36 billion this year in capital expenditures. That’s the magnitude of investment required to stay on the bleeding edge of semiconductor manufacturing. And that kind of investment can’t happen in the absence of massive scale and a whole lot of free cash flow.

While I wish for peace and an absence of geopolitical conflict, I find myself oddly excited about the investment prospects of a fire sale on TSMC. We don’t know exactly when it will happen, but we’ll be ready to jump in when it does.

A hundred-billion-dollar mistake?

A few weeks ago Google revealed its own generative artificial intelligence (AI). It’s called Bard.

As a reminder, generative AI refers to an AI that can interact intelligently with humans, answer questions, and complete certain tasks upon command.

As we discussed, Google’s hand was forced. It had to rush the release of Bard because it was caught flat-footed. The incredible success of OpenAI’s ChatGPT took the tech giant by surprise.

And the fact that Microsoft now controls about 75% of OpenAI and has incorporated ChatGPT into its own search engine, Bing, made this an even bigger problem for Google. So much so that the company issued a “code red”.

But Google’s Bard got off to a bumpy start when the company demonstrated it for the first time a couple weeks ago.

Bard was asked about the new discoveries from the James Webb Space Telescope. And Bard responded that it had taken the first pictures of an exoplanet – a planet outside of our solar system.

Unfortunately for Google, that answer was wrong. The first picture of an exoplanet happened back in 2004. Nearly twenty years ago. A fact that was quickly pointed out by many.

Bard’s error led investors to dump shares of Google’s parent company Alphabet. Get this – Alphabet lost over $160 billion in enterprise value in the three days following Bard’s gaffe. We don’t see that happen very often.

As regular readers know, I’m not a fan of Google’s business practices. But remaining objective, this was a massive overreaction.

We know Google had to rush the release of Bard. We also know that this is brand new technology. It’s only been around for about two and a half months. It is far from perfect.

So, the fact that Bard made a mistake is not as big a deal as investors made it out to be.

The fact is, these generative AI improve every single day with use. Two months from now, Bard will be far more robust than it is today.

And we shouldn’t fall into the trap of thinking that Google is out of the generative AI race because of one mistake. Quite the opposite.

The big takeaway here is that these tech giants are moving to advance generative AI at an accelerated pace. And the stakes are enormous.

To me, it’s clear that generative AI represents the next generation of search technology. It’s highly likely that this tech will ultimately replace search engines entirely. No more scrolling through several pages of links looking for an answer. Instead, the AI will simply deliver the answer to us with the context that we need.

Companies like Google and Microsoft know this. That’s why they’re racing forward with generative AI right now.

And “search” won’t just be limited to search engines. Corporations and governments will use generative AI to study their respective organizations body of knowledge. From there, search can be used in a contextually relevant way for the functioning of any given organization.

This is an incredibly powerful concept and one that will radically improve operational efficiencies. And it’s just getting started.

Fourth generation CRISPR technology out of stealth…

Feng Zhang is at it again.

Long-time readers will remember Zhang as the founder of CRISPR Cas9 genetic editing technology. He also co-founded first-generation CRISPR company Editas Medicine (EDIT) and second-generation CRISPR company Beam Therapeutics (BEAM).

And in the last few days we’re learning more about Zhang’s latest CRISPR company, Aera Therapeutics.

Aera has been in stealth mode since its founding in 2021. There’s been very little information about the company. But that’s starting to change…

In the last few days Aera has hired two high-profile executives to help lead the company. One was the head of oncology at Alnylam Pharmaceuticals. The other was Alnylam’s founding CEO.

This is telling. For two reasons…

First, the fact that Aera plucked Alnylam’s head of oncology suggests that the company plans to focus on therapies that could cure forms of cancer. That’s obvious.

Second, Alnylam specialized in a technology called RNA interference (RNAi). RNAi is designed to deliver medicines directly to cells to stop them from producing certain proteins that cause disease.

And RNAi tech has proven to be a successful approach. In fact, Alnylam has grown into a $26 billion biotech company on the back of this tech.

So it stands to reason that Aera Therapeutics wants to incorporate RNAi technology into its own approach.

In addition, we know that Zhang has developed a new approach to genetic editing called SEND. That stands for “Selective Endogenous eNcapsidation for cellular Delivery”.

SEND focuses on delivering medicines directly to cells. And it has the potential to expand the universe of diseases that can be treated by genetic editing technology.

Put it all together and we may be looking at what is effectively a fourth generation CRISPR company here.

If Aera Therapeutics successfully combines CRISPR, RNAi, and SEND technology, it’s safe to say the company is skipping over the third generation entirely and pioneering an approach that’s never been tried before.

It goes without saying that this is absolutely a company to watch going forward.

And Aera is launching with an impressive $193 million venture capital raise backed by ARCH Venture Partners, Google Venture, and Lux Capital. ARCH is one of the best early stage biotech venture firms, and Lux is known for being very early in investing in breakthrough technologies.

And I’m excited about what Zhang and Aera are going to accomplish with CRISPR and its novel delivery system.

Generative AI will quickly replace the sketch artist…

We’ll wrap up today with another interesting application of generative AI. This one courtesy of OpenAI’s DALL-E 2.

If we remember, DALL-E 2 is a text to image AI. Users simply describe what they want the AI to create using text prompts, and the AI creates the image from scratch.

We’ve looked at some creative images that came from DALL-E 2 in previous issues. And as it turns out, DALL-E 2 can serve practical purposes as well.

A few weeks ago, two developers entered into a “hack-a-thon” centered around AI. They ended up using DALL-E 2 to create an AI-based forensic sketch artist. And it’s absolutely remarkable.

Here’s a look at the AI-powered sketch artist at work:

Generative AI Applied to Police Sketch Art

Source: LabLab.AI

Here we can see someone inputting information regarding a suspect’s physical appearance. Then, in a matter of seconds, the AI generates an incredibly life-like image.

And if the image doesn’t match the user’s memory, all they have to do is make some tweaks to their description. They can iterate the sketch on the fly.

Today, law enforcement often uses forensic sketches like this to try to identify the perpetrator of a given crime.

Historically, this has required forensic sketch artists to interview witnesses who try to recount the criminal’s physical appearance. Then the artist renders a sketch based on the information provided.

The challenge is that this is a very time-consuming process.

For starters, forensic sketch artists aren’t always immediately available to interview witnesses. Yet, every hour that passes tends to reduce the clarity of human memory. Thus, the longer it takes to do the sketch, the less accurate it is likely to be.

A tool like this solves that problem. Witnesses can use it immediately after a crime to create the most accurate sketch possible. And where a human artist would need a few hours to produce a sketch, the AI can pull one together in seconds.

It’s far more efficient and can produce photorealistic images that can not only be useful in identifying a suspect, but it could also be used to compare to a database of known perpetrators.

And here’s the thing – this tool was developed over the course of a day or two at the hack-a-thon and yet it is already capable of incredible quality and functionality. With a solid development team and a couple months of work, this kind of AI can be dramatically improved and ready for production.

The bottom line is that generative AI is going to quickly become ubiquitous in our society. What makes it so different from traditional software development is the speed at which it can now be made functional.

Not employing this incredible technology will quickly become a competitive disadvantage.


Jeff Brown
Editor, The Bleeding Edge