Can You Use ChatGPT to Create a Portfolio of Stocks?

Rob from NC posted over 2 years ago:

AI has a very long way to go to become truly "intelligent."

Dennis from Washington posted over 2 years ago:

Something I learned programming software 50 years ago "Garbage in, garbage out" is still true today. The output of a program, even a generative AI program is no better than the quality of the input. Expecting a program to scrape unauthenticated and un-curated data randomly from the internet and getting something useful is a fool's errand. I have played with ChatGPT a bit and could only get meaningful results if I used very narrowly defined queries that would only reference authenticated sources. And even then, I was usually disappointed at the limited results. The version of ChatGPT that we are allowed to play with for free is just not sophisticated or "smart" enough to know useful data from garbage or fiction. So, I learned I cannot throw my stock analysis Excel spreadsheets away just yet, but maybe I can use them to train the software...

John L from NJ posted over 2 years ago:

I am most impressed by the smart people who come up with the nice sounding excuses for generative language models. Hallucinate is so much better sounding than "pathological liar". But it doesn't matter. Would you hire someone who suffers from hallucinations or is a pathological liar to invest your funds. I think AI is over hyped. Both as a technology and as an investment theme.

gregg w from california posted over 2 years ago:

SORRY - the comment system removes all line breaks! If you want to read a longer explanation of why ChatGPT is more like a parrot than an intelligent Excel spreadsheet, read https://pastebin.com/5ZE3JqgM The biggest danger to "AI" in general, and ChatGPT specifically, is how it's marketed and perceived. it's basically a deep fake of human intelligence as defined by the contents of electronic documents, but people don't realize how fake it is.

Barry from TX posted over 2 years ago:

ChatGPT is to "intelligence" as ... Crypto is to banking security ... and The Internet is to privacy. All their promises are GIGO FUBAR. Every great "invention" has been implemented as BOTH creative/progressive and destructive/repressive forces - fire, civilization, commerce, religion, printing, markets, political parties, nuclear fission/fusion, the internet, etc. There are no reasons to expect that AI will be different. A lot of people are going to get rich and even more will lose their fortunes by investing in AI. That is the nature of investment markets.

Barry from TX posted over 2 years ago:

Maybe I was too hard on ChatGPT. I should have redirected my trash-talking to the known deficiencies in current investing industry practices. In their Abstract, the authors of the background article say, "TRUSTING investment advice from Generative Pre-Trained Transformer (GPT) models is a challenge due to model "hallucinations", necessitating careful verification and validation of the output." That quibble makes it fair game to ask, what is the difference between this disclaimer of their experimental results and any pro forma investment industry disclaimers like (1) "Past performance is no guarantee of future results" (a sample from a broker I use) and (2) "Deeper analysis should always be completed before making investing decisions" (a sample from an AAII offering). Relative to claim #1, ChatGPT "hallucinations" sound very similar to the pro forma "results" backtesting and model portfolios produce -they are presented without any guarantee of a linkage to real-world results you should expect. Relative to claim #2, the ChatGPT data in the Figure in the AAII review from the study "advises" all investors to rely on their own "due diligence" process rather than implement a list of the outputs from any bot ... or ... from a fund portfolio manager. #2 sounds like a distinction without a difference. Trying to tell if you are interacting with a computer or a “sentient” person goes back 75 years to the birth of computing. 'The father of modern computing', Alan Turing begins his 1950 paper, "Computing Machinery and Intelligence," with the question, 'Can machines think?' Turing sidestepped the challenge of defining "thinking" and created a very practical, albeit subjective, test called the “Imitation Game,” now known as the Turing Test: A human interrogator – who communicates with a computer only by exchanging messages -- decides if they are interacting with a computer OR a human subject based on the replies the interrogator poses. By means of a series of Turing tests, a computer’s success at “thinking” can be measured by its probability of misidentifying the computer as a human subject. The advent of ChatGPT reignited the conversation about the likelihood that the components of the Turing Test had been met. The referenced article does not resolve Turing’s 75-year-old question. AAII member comments above, including mine, suggest that ChatGPT is closer to “a blind chimpanzee throwing darts in the dark” than it is to the typical active fund manager who cannot beat the market index his fund benchmarks.

Barry from TX posted over 2 years ago:

Charles, your prescient topic started an avalanche of data searches for me. So far I have placed the whole "AI is the next big thing" that will change the world in the same trash-talking bin as Blockchain and crypto. Unlike those "hallucinations" that have been reigned in somewhat by the SEC, Generative AI promises to be truly disruptive on a much larger human scale. It may be far off (?), but it is a mile-long freight train approaching an RR crossing in many neighborhoods. Here is a friendly link to an article in today's WSJ that has two charts that project the impact of Generative AI on (1) educational attainment and (2) employment (by industry). The article has additional data points from research by Goldman Sachs, Morgan Stanley, McKinsey, and MIT. https://www.wsj.com/articles/generative-ai-promises-an-economic-revolution-managing-the-disruption-will-be-crucial-b1c0f054?st=m2298ou73kmh5zc&reflink=desktopwebshare_permalink Net: An average of 40%-70% of various educational degrees will be impacted (McKinsey) and 60%-64% of jobs in most industries (Goldman Sachs). The projections are extrapolations and speculations, but they are data points from reputable sources outside the AI industry which is prone to its own human "hallucinations." If these data points are anywhere near accurate, the big questions are (1) What /where will other jobs for these millions of dislocated people come from? and (2) How can colleges justify the exorbitant fees they charge to educate (there is a separate WSJ article on the topic of rising costs of student debt) people for careers/ jobs that will not exist? And my favorite, (3) how can I as an investor be a willing "yet to be named co-conspirator" getaway driver in this "bank heist"? As prior comments point out, we should expect a lot of hype from the AI crowd. They tend to lean over the tips of their skis. And they are a kiddie slope crowd on a Triple Black Diamond trail.