- The “magic” of The Perceptron
- The risk of holding through an acquisition
- Readers weigh in on “Enceladian sushi”
Welcome to our weekly mailbag edition of The Bleeding Edge. All week, you submitted your questions about the biggest trends in technology. Today, I’ll do my best to answer them.
If we have a question we’d like answered in a future edition, we can submit it right here.
Now, let’s turn to the mailbag.
The “magic” of The Perceptron…
Your [recent alert for Neural Net Profits] is quite interesting.
But in the next to last paragraph, you state that you have “no idea” what the perceptron is “thinking.” How can you not know what it is thinking?!! Didn’t you help design it?
– Graham L.
Hi, Graham. This is a fun question because it speaks to the complexity and mystery of using artificial intelligence to perform some kind of function. The question also highlights a key tradeoff that we have to make when using artificial intelligence, that of performance versus explainability.
For the benefit of readers, the Perceptron is the name we’ve given to a proprietary neural network that I’ve built—alongside a specialist in the field—to do one thing: Find digital assets on the verge of shooting higher.
Yes, my team and I architected and programmed the Perceptron. We also gave it some guardrails due to the volatile market that we’re in right now.
More specifically, we designed the Perceptron to perform better in the kind of tough markets that we’re in today. This is something that we can adjust when healthy market conditions return. And of course, we control what we “feed” the Perceptron in terms of data inputs.
When programming artificial intelligence, we are designing for some kind of highly probabilistic outcome. In this case of course, the Perceptron becomes confident in trading patterns of digital assets and tells us when the conditions are ripe for a sharp move higher within a specific window.
But there is no way to know exactly why the Perceptron came to that conclusion. Some like to refer to AI as being a black box – we know what goes into the box and we know what comes out of the box, but we can’t see what’s going on inside.
The tradeoff that I mentioned earlier is an interesting one. If we design an AI for explainability, so that we know how it has come to a certain conclusion, the results will be far worse; more akin to a statistical analysis which isn’t AI at all. But if we design for optimal performance, which is my goal with the Perceptron, then we have to give up on explainability.
This is a trap that hedge funds that use AI fall into. They need explainability, otherwise they won’t allocate their limited partner’s capital. And that results in a less performant system. We don’t have those constraints. We can simply design the best system for optimal performance.
I’ll share a recent example of the Perceptron in action…
In March of this year, the Perceptron gave us a strong signal for the digital asset Loom (LOOM). Here’s a visual to show us what the Perceptron was “seeing” at the time.
In blue, we will see the price of Loom trading sideways. But let’s pay close attention to the data in orange. This is the “score” that the Perceptron is applying to that asset. We can think of this as the neural network’s “conviction” in the trade. That sharp uptick in the Perceptron score was our signal to buy. We entered our position in LOOM on March 16 at a price of $0.075.
Shortly after we received this reading, the entire digital asset market experienced a vicious drawdown. Virtually every digital asset sold off, Loom included.
At that point, most traders would have been tempted to sell. But the Perceptron saw something that nobody else saw. It maintained a strong signal for LOOM, which surprised us. But we knew that the neural network continued to see the high probability of a sharp move higher in the coming weeks.
And after some patience, we were rewarded.
In September, Loom spiked seemingly out of nowhere. It was the larger move the Perceptron had been predicting. This gave us the opportunity to lock in an 81.7% return on the trade by exiting the position at $0.137 on September 12.
And we should pay attention again to the data in orange. Notice how the Perceptron’s ranking for the asset drops sharply after the big move. I remember it well. It happened at 5 AM that morning and we sent out our alert a few hours later. When this happens, it is a clear signal to sell.
It meant that not only did the Perceptron no longer believe that LOOM would go any higher, but it was also actually predicting a sharp drop in LOOM. It’s a good thing that we listened to the signal. As predicted, LOOM fell sharply in the days that followed after we exited the trade.
We didn’t need to understand how the Perceptron came to its buy and sell signals. But we have confidence in how we designed it, the data that we feed it, and what the Perceptron is optimized to do.
It’s capable of synthesizing far more data and information than we could ever hope to do ourselves, which is what makes it so powerful. But we simply can’t know exactly what it’s thinking. And I’m perfectly fine with that.
That’s why your question is so useful Graham. It highlights why the adoption of this technology has been held back in traditional finance. Money managers need to “understand” how these neural networks “think.”
And that’s simply not possible. As I shared with readers when we launched the Perceptron, there is a sort of “magic” to these machines.
The risk we take holding through acquisitions…
Today, I have a technical question regarding a company takeover. I will be using the example of LogicBio. You have recommended selling the stock, because we are very close to the acquisition price of $2.07. Now, my question is what happens if someone doesn’t sell the assets? Will the stocks be automatically liquidated for the price of $2.07 and the money deposited into the trading account, or will all get lost?
– Paul P.
Hi Paul, thanks for your question. To catch readers up, last week we got news that a small-capitalization biotechnology company called LogicBio Therapeutics (LOGC) had entered into an acquisition agreement with biopharma giant AstraZeneca for $2.07 per share.
But what was really interesting was the premium that AstraZeneca had to pay in order to get the deal done. Prior to the announcement, shares were trading around $0.30. That means AstraZeneca had to make an offer that was 644% higher than where the stock was trading.
Acquisition offers will always happen at a premium to the current market price. But for an acquisition to happen almost 8X higher than the market value is very unusual. And here’s the thing; at $2.07 per share, LOGC is still wildly undervalued. AstraZeneca is getting one heck of a deal.
And this tells me that the current valuation for early stage biotech is beyond irrational. And I wouldn’t be surprised if the number of acquisitions for these small biotech companies picks up. There are simply too many great companies trading too cheaply.
As for your specific question, Paul…
Yes, if an investor decides to hold shares through the acquisition, the investor will receive $2.07 a share for every share owned. Investors don’t need to do anything.
Worth mentioning is that acquisitions can come in many flavors. There are all-cash deals, cash-plus-equity deals, or all-equity offerings. That means that instead of cash, investors might receive some ratio of shares in the acquiring company. Or they might receive a mix of shares and cash.
When I recommended selling LogicBio, shares traded at $2.03. From my view, it just didn’t make sense to continue to hold the stock to squeeze out a few more pennies.
That said, we did receive several emails from subscribers who said they would continue to hold shares through the acquisition to catch the extra upside. There’s absolutely nothing wrong with subscribers making their own decisions based on the research. I encourage everyone to do so. My goal is to help inform my subscribers, but our own investment decisions are ultimately ours.
But there’s another reason I’ll often recommend selling before an acquisition is officially complete. There’s always the possibility that the acquisition will meet its own challenges. That’s the situation we had earlier this year with Activision (ATVI) in the pages of The Near Future Report.
Activision was one of our portfolio companies. And in January, news broke that the business had agreed to be acquired by Microsoft for $95 a share. We held on for a bit longer to see if any competing offers would materialize. And by April, we had sold the position for around $80 per share.
The reason we did that is because the Activision/Microsoft deal was being heavily scrutinized by regulatory agencies, specifically the FTC. In the event the deal fell through, shares of Activision would have sunk. And I didn’t want subscribers exposed to any downside.
And while I ultimately expect the deal to be approved, ATVI has been volatile as the deal works its way through regulatory approval. That’s the risk we take when holding through acquisitions. If the deal falls through, even if it’s just delayed, the share price can sink quickly. That’s a big reason why I usually prefer to sell acquired company stocks before the deal is complete.
With all that said, I don’t expect the LogicBio deal to fall apart. And there aren’t any regulatory concerns that we need to worry about.
Let’s conclude with some of the recent feedback we received on the topic of life living on one of Saturn’s moons…
This is a very interesting question! I’m not a scientist but I have been reading Science Fiction for a very long time. So, flights of the imagination are in my experience, and it’s amazing how many have come true, though different from the authors’ perspective, at the time the author wrote the story. A lot of your recommendations for evolving technologies that are coming true now, remind me of Sci-Fi stories I have read in the past.
So, can Enceladia support life? It seems quite possible to me. If, under the ice cap, there is an air pocket trapped with the liquid water, that would boost the possibility of this idea working. Algae could probably live there, and once that happens; more complex life forms evolve. Possible crustaceans, worms, exist here on Earth where no sunshine penetrates deep enough to make a difference. The other problem is, the under ice volcanic eruptions that probably happen. Is the ice porous enough to vent those eruptions?
– Jerry F.
Ever since seeing the Webb pictures I have been convinced that there must be life on some of those planets, moons or even entities that we may not even be able to see. It is the next giant evolutionary phase, and there is no reason to assume that life is the same as ours. I wish I was much younger. Keep up your good work.
– Tom C.
Hi Jeff, Great articles and insights on both Musk and Enceladus. However, you spoiled it for me at the end with your comment regarding whether the seafood is any good there. I trust you are okay, then, if some aliens arrive on Earth in order to sample the local, bipedal Smorgasbord, with the help of a large book entitled “To Serve Man”? (I think you are old enough to get this Twilight Zone reference.)
– Dr. Q.
Re Enceladian sushi, see Guardians, which is the third chapter in the marvelous book Tuf Voyaging, by George RR Martin. This chapter was apparently written, or at least copyrighted, before chapters 1 and 2.
– Spencer M.
Hello – I’d love to taste some Enceladian sushi. Or, at least, see some astronauts try it first. Nice story. Please keep up the good work!
– Paige D.
For any readers who missed the story, on Wednesday, we looked at some new research surrounding Enceladus, one of the moons orbiting Saturn. As it turns out, the moon has all the basic building blocks for life.
Temperatures on its surface are around negative 230 degrees Fahrenheit. So, we shouldn’t expect any life on the surface. But we also learned that Enceladus has oceans under its ice, and a warm core, and most certainly hydrothermal vents.
That raises the question of aquatic life on Enceladus… and if the seafood is any good.
Thanks for all the fun comments on Enceladus. Dr. Q – hilarious. Spencer – most people don’t know that Martin wrote science fiction before he ever wrote his first book in the Game of Thrones.
I had fun with this topic. In all seriousness, sushi is my favorite food. I was spoiled living in Japan. I had the absolute best sushi on this planet for so many years. There are a few great spots in New York City that come close. But outside of that, nothing comes close to Tokyo. But who knows what kind of seafood is out there???
Thanks to subscribers for being so engaged with our topics.
That’s all the time we have this week. Remember, if you’d like me to answer one of your questions, send me a note by writing to [email protected]. I’ll do my best to get to it in a future mailbag.
Editor, The Bleeding Edge