So PLTR just had another AIP.com meeting, similar presentations. CTO describes running data through "K LLMs" to come up with analysis or solutions. This brings up something I am quite ignorant on, not even sure it is important to understand. How does a company like PLTR or EATON run their data through a LLM they have not developed? For example, if I want to ask chat GPT a question, I go to an open AI website, log in, ask the question, get an answer. But how would chat GPT analyze data about the Eaton supply chain? It seems to me that hosted LLM solutions like chatGPT do not have access to Eaton data. What about other LLMs? It seems like for business and government warfare or psyops purposes one would have to be in possession of an internet isolated LLM on ones own computers to run sensitive data through. Are such LLMs something one can buy or rent?
Hi nurse, if you try to understand how the software langchain works this will make a bit more sense. You can use langchain to chat with a pdf that you create or a research paper. What you are describing below is like langchain but scaled up massively. Personally, I love chatGPT but I also think AI is extremely overhyped right now. Like the example of Eaton supply chain data and a LLM just doesn't make sense to me. It is not as if companies were doing no data science before chatGPT came out. The idea a non-data scientist is going to find all these valuable insights with the help of language models that traditional data science somehow missed is just nonsense. The low hanging data fruit was picked 15 years ago. There is a lot of AI business performance art right now. Palantir will certainly gain sales from this performance art in the short term though. I have traded PLTR a number of times and for me it is neither cheap enough to buy or rich enough to short here although it is coming back up on short territory for me. I have been short much more often than I have been long. Given the growth rate and growth delta, it is hard to not see 34 billion as fully valued. This is a super meme stock in the sense of being retail and narrative driven. My bet though is that language models are extremely useful broadly and rather useless specifically. Palantir could be a good fit in the broad sense too but so can many other options. That is the real issue with any long term investment thesis with PLTR for me. Like Tableau does the research and then Microsoft undercuts and smashes them.
In the latest AIP.com, I recall Eaton saying they were using PLTR AI products. Maybe not chat GPT, but something. I do not well understand this stuff and am trying to understand more. Sometimes when folks try to understand PLTR products and applications, they seem to have biased confidence in their off the cuff opinion. Happened the other day asking for insight from old coworkers.
The culture of Palantir seems to be to use as much jargon as possible when describing the products. In a way that makes me believe they want the products to be hard to understand and that is used as a marketing strategy. My experience is that the more you understand the company, the less anything makes sense. For me, Palantir Metropolis is exemplary of the problem they face as a business. Metropolis was at the right place and time to catch the quant revolution in finance. In the end though it just ends up being better and/or cheaper to build things in house, not be vendor locked at a premium. I think AIP will get a much bigger bump from LLMs than anything Metropolis got from hedge funds but it won't be sustainable. It most likely is a really good solution if you want to get LLMs into production as fast as possible. You can pretty much interchange chatGPT if you see LLM. Palantir marketing though makes it sound like they have their own LLM or machine learning algorithms when they simply do not. Again, I think this is 100% intentional. The examples of AIP they show in the videos on the website are so poor. The idea a non-software engineer is going to build some LLM application with a Palantir no code solution and then gain all this value is just science fiction. It is a wonderful stock to trade though. I was a hoping for a bigger bump recently to get short but I think Q3 will be rather nice so I do want to have a long position over earnings.
I appreciate the PLTR discussions from all. For those that say things like "the more you understand the company, the less anything makes sense", how is it that so many large companies are using the products? How is it that HCA is incorporating their product into staffing solutions across a system that encompasses 90,000 nurses? I do not understand what Metropolis is, have heard of Gotham though. I'm not sure about AIP examples as being poor, will trust your opinion. But, again, lots of large companies giving real and demo mode examples of AIP adoption. The Palantir youtube channel has some good info, most of which is from their customers describing the implementation. I could crowd this forum with individual videos on a regular basis, but trust that those really interested can go find/watch/stay updated. https://www.youtube.com/@palantirtech/videos Karp can come off overly enthusiastic and perhaps sales and profits won't scale as quickly as some would hope for, but the bias seems to favor the bulls.
Palintir gets a lot of gov contracts. and once secrets are shared these contrcats become sticky. Palintir nabbed a British health software deal that apparently somehow is priced into the shares. Didn't like that. But it's on the business side I think where the real AI comes in.. I mean high tech spying yes all of that but what of just business Its potential with commercial customers--> is Palantir's key to long-term growth. Data is becoming crucial to competing in the private sector, and its U.S. commercial customer count grew 50% year over year in the first quarter and another 35% in second. It still has just 161 U.S. commercial customers, <-- as of Aug/ Palantir released its Artificial Intelligence Platform (AIP) and management reported unprecedented interest in the product on its second-quarter earnings call. It also said that new deals from the second quarter spanned 30 industries. Lets assume they double the customer count to 300+ and the AI gets a higher bill // ~stoney
I see now that it is not super easy to find information about Metropolis. Metropolis is basically their failed financial services platform from a long time ago. JP Morgan was a client I know but then they just built their own in house version and stopped using Palantir. If you research Metropolis you can see how contrary it is to their marketing, otherwise every hedge fund would have been using Palantir for the past 15 years. They are a heavyweight , premium, all in one solution so it makes sense if a large company that wants to beat the competition to using large language models then Palantir should get a boost. I am just really not sure how much value LLMs are going to deliver to the bottom line in the short/mid term but I could be unbelievably wrong on that. I got a low conviction short on this morning at $18.20 and moved a stop to break even. It will probably depend more what happens with the NASDAQ than anything right here. I do like though how much it didn't participate in the NASDAQ move up in the afternoon. The stock usually doesn't sit still so I doubt I will be in this very long either way. Youtube for me is pretty useless on this stock as everything is basically a contest of ultra bullish cheerleader content creators trying to make money posting videos so it doesn't even have value as a sentiment measure.
I had closed my short out last week but I think I missed the long here this morning. That drop to the $15.40 area this morning would have been the long for me with good support underneath and hold through earnings but I was busy with other things. Probably not a bad trade to buy here still and hold through earnings but that is just not how I trade. I will probably now wait for the next short as there is always this retail money that comes piling in massively late on basically any news.
This is a problem with discussing trading online. I didn't even think about getting long today. That post Monday was then, this is today. This is now though after the bell. I want NQ down Sunday night! To believe you are following a rule, is not the same as following a rule. What could apply to trading more than that semantically?