What Did You Buy Today?

[fenders53]

(05-04-2021, 01:54 PM)Otter Wrote:

(05-04-2021, 12:40 PM)fenders53 Wrote:

(05-04-2021, 12:33 PM)Otter Wrote:

(05-04-2021, 12:29 PM)fenders53 Wrote:

(05-04-2021, 12:27 PM)Otter Wrote: Bought the PLTR dip today. LOL.

The everlasting dip.  I may do the same but a put sell in case there is more falling knife action. 

I remember when the Seeking Alpha comments section was mocking all the bagholders who bought SQ at $100 in 2019. Those are some happy bagholders today. I should have averaged up on my position, with hindsight being 20/20.

I think PLTR has more potential upside within its business area at higher margins/ROI over time than SQ does.

PLTR is the position I have the least understanding of in my port.  It's an extreme violation of my investing rules but I have to have some spec to keep it fun.  I put some trust in your opinion though.  I keep selling PLTR puts every few weeks and make income.  Eventually I may be sucked into a few hundred shares and I'll ride it out with you.

The fun part of this is that how AI actually arrives at its decisions is beyond human understanding. If human minds could replicate those "thought" processes, we wouldn't need AI. What we do know is that special purpose AIs can now regularly wipe the floor with Go Grand Masters:

https://www.theguardian.com/technology/2...-lee-sedol

The number of potential move combinations in a single game of Go is more than the total number of atoms in the Universe:

https://en.wikipedia.org/wiki/Go_and_mathematics

The ability of a special purpose AI to defeat the world's best Go players was specifically referenced by Ray Kurzweil as one of his yardsticks for AI achieving commercial feasibility when he wrote The Age of Intelligent Machines in 1990. The book itself is a great read, and Kurzweil's ability to relatively accurately predict the next thirty years of computing history was visionary. There is a reason Larry Page personally hired Kurzweil to come work for Google to "bring natural language understanding to Google" in the area of machine learning/AI. No one can explain the thought process underlying Alpha Go's ability to wipe the floor with the world's best Go players. There are programmers who created the algorithms, but the machine uses those to teach itself, and its method of thinking is entirely alien to us. If we could understand its thought processes, it wouldn't be able to play Go better than we can. 

A Special Purpose AI should not be confused with an Artificial General Intelligence, which is the sort of superintelligence commonly portrayed in movies. We are probably still a couple decades away from that. However, special purpose AIs can solve complicated problems far faster than humans alone. This is true even in the arena of traditional ERP software. ERP is all about maximizing efficiency in a business to increase earnings. 

There is a network effect to Special Purpose AIs. The more data they have access to that is relevant to their special purpose (like a move-by move history of all recorded professional Go matches), the better they are at spotting patterns and fulfilling their special purpose (whether winning a game of Go or optimizing a company's supply chain, overhead, etc.). There are other Special Purpose AIs out there in the world of ERP, but most have to be fed structured data (as opposed to unstructured data):

https://learn.g2.com/structured-vs-unstructured-data

When Palantir was founded with seed money from the CIA, its entire purpose was to solve the hard computing problem of dealing with unstructured data, and using it to provide meaningful information to decision-makers. The company has excelled at that in the public sector, and is the preferred contractor for the U.S. Government in this space. It has its hooks into government data from the most sensitive national security sectors, to nuclear safety, to healthcare and other sectors. The company has been training its AI on the world's largest data sets for almost two decades now, and PLTR has benefited massively from that. The technology is now at a point where they feel comfortable deploying in a commercial space. 

Initial build out of capability in an industry (like resource extraction with BP and Rio Tinto) still requires PLTR to send its engineers out to a customer in that industry, perform "discovery" (learn where all the data is, the organization's requirements for securing and handling data, etc.), and then tailor its AI to access and process that data. However, once they get a footprint in an industry, each successive deployment within that industry can be done for the next company on a faster and faster timeline. Data analytics has been around for a while, but the power of PLTR's software to grab unstructured data, analyze it, and provide a meaningful interpretation that can optimize an organization and increase EPS is something new. There are a lot of companies now trying to copy what PLTR does, but first mover advantage and network effects in a field like this are meaningful. Once Google had the best search algorithm, it drove more people to use Google, which improved at a faster rate than competitors because it had access to more data to keep optimizing the algorithm.

So, neither you nor anyone else will ever understand how exactly PLTR's software arrives at its suggestions for optimizing businesses or governmental actions. If people could do that, we wouldn't need AI. That part will always remain a black box to human understanding. What we can understand, though, is that the software works (Demo Day and Double Click presentations are full of examples), and how the network effect and first mover advantage benefit PLTR. They have had nearly two decades of taxpayer money to develop the black box with one of the most security-conscious and demanding users of data that exists. Other players in the space have a nearly insurmountable mountain to climb in an attempt to compete. It's kind of similar to how SpaceX has rendered just about every other rocket manufacturer's products obsolete with Falcon 9 and Falcon Heavy. Competitors are looking at 5+ years of R&D on an ambitious timeline, just to get to where SpaceX is today.

The scope of the potential market is also as big as the entire economy. If the black box can optimize operations and drive meaningful increases to EPS, it is valuable, and any company seeking to maximize shareholder value would be negligent for failing to use the service that can do that, so long as the cost of the service is less than the benefits it provides. Margins are notoriously fat for these sorts of services. No one is going to want to use the second-best option. When I want to find something on the internet, I use Google. 

This whole industry is in its infancy, as computing power has only recently crossed the threshold in the business world where special purpose AIs can take on hard, unstructured optimization problems like PLTR does, and provide useful results. The industry itself will be as ubiquitous and boring as search, operating systems, or cloud services within 20 years. Maybe another horse wins the race to dominate the market, but the three-letter agencies that matter have placed their bets on PLTR. I'm willing to make my bet alongside them. 

I know just enough to believe it's not going to zero, and will be OK or much better if I keep the faith.