ChatGPT produces responses based on its training dataset, and given OpenAI scrapped Reddit that should tell you all you need about whether it’s a good investor or not

@carpelunam

11 ай бұрын

🤠 I was going to come here and say it wasnt trained to do this function properly, but youve basically said the same thing.

@moneysins

11 ай бұрын

@@carpelunam Maybe ChatGPT will decide to start charging absurd pricing for its API soon

@FF18Cloud

11 ай бұрын

oh god, it might be wallstreetbets-ing it

@redbunnyclassic

11 ай бұрын

😂😂😂😂😂

@antman7673

11 ай бұрын

Yeah and on youtube, which contains the opinion, that wall street bets is a terrible idea.

"I'm not a data scientist, I'm not going to get into the technicals.. that I'm not a subject matter expert on." Thank you. the world needs more of this. not read a couple google searches and pretend ones an expert

@TheManinBlack9054

10 ай бұрын

I wish people in the comments would listen

@ecos889

8 ай бұрын

Yeah especially with professionals. Like Jordan Peterson's view of ADHD is not at all reflective of the condition itself as it's not his field. Yet the man insists as a clinical psychologist to tell people about his thoughts when asked without stating it's not his field. Like their is a person on KZread that has been studying ADHD for decades who explained how wrong Peterson on ADHD and oh yeah it's awful.

So glad someone makes party pooper reality checks. Thanks 🙏

@ianforbes9

11 ай бұрын

Well said!

@archvaldor

9 ай бұрын

it isn't especially useful to say "this doesn't work". Everyone knows most stuff doesn't work. I want to know about the 1% stuff that does.

We used chatGPT for a real life coding project, and it is very useful, but as an assistant. What people seem to miss is that it doesn't think, it just guesses a correct order of words very well

@alext5497

10 ай бұрын

what real life coding project. i don't believe you. Or the code is poop.

@sergeyderbeko9954

10 ай бұрын

@@alext5497 that's an interesting comment. Real life as in for a client, not some portfolio project :D

@Snout007

10 ай бұрын

@@alext5497 It can generate first drafts of complicated python code for a task like converting a complicated file format into a csv file, etc. Which you can't do correctly by any other means than writing a code to do it. I found it very usefull because sometimes you need some complicated coding library just for one task, so a code from chat gpt that slightly adjusted can do the job just fine. It's either that or spending 20 hours learning a complicated liubrary for one small task.

@ianbelanger7459

10 ай бұрын

It is strange that the two commenters latched on to the coding aspect and not the more salient knowledge point. LLM don't systematize knowledge meaning from a certain perspective LLM don't know facts and, therefore, have great difficulty accurately answering questions.

@sergeyderbeko9954

10 ай бұрын

@@ianbelanger7459 isn't that just a pompous way of saying "they don't think". With code, they'll just likely continue the most common syntax in the reply to question, and if they get it wrong, tweak a few things and ask again

I had to challenge someone on this. The person was so convinced that it's working well that I just gave up. I am expecting some type of investing horrors to happen soon enough.

@Excalibur2

11 ай бұрын

I hope we at least get a few funny r/wallstreetbets stories out of it. How else would we get amazing rags to riches (back to rags) stories?

@AlexanderTheGoodEnough

11 ай бұрын

There will be big winners posting about their gains. The waters will always be muddy

@larryo6874

11 ай бұрын

@@AlexanderTheGoodEnough It’s like when I go to a casino in Las Vegas and they have all these pictures of smiling people on the wall who won mega bucks playing the slot machines. What you don’t see are all the people who lost money on the slots.

@AlexanderTheGoodEnough

11 ай бұрын

@@larryo6874 oh God, don't remind me. I used to work at the Palazzo at the sports book.

@michelleachacoso4983

11 ай бұрын

Gonna be ugly. Pop some corn and watch the show.

I’m an AI scientist that trains LLMs, and you did a great job explaining LLMs. One minor nitpick is that the blobs showing parameter size at 6:49 isn’t correct. Idk why that image spread all over the internet, but OpenAI didn’t release the number of parameters for GPT4 like they did for previous versions. We know that has either the same parameter size or just a moderate amount more. It’s certainly not nearly as large as the image says. Other things like dataset size, data cleaning, and larger token sequence size probably accounts for improvements more than any parameter increase. The small and large blobs have become a bit of a meme/joke among AI researchers. I really enjoy your videos, especially the party poopers!

@TheWoWadicted

10 ай бұрын

Not necessarily. There are rumours making the rounds that GPT4 is a Mixture of Experts model, with 8 220B experts. This would give you a total model size of 1.76T parameters

@TheWoWadicted

10 ай бұрын

I.e about an order of magnitude increase from GPT3.5

@alext5497

10 ай бұрын

citation needed.

@curlyvideos

10 ай бұрын

Those blobs are upsettingly silly-looking. No description or puffery. No units, no scientific notation. Row instead of column format so you can't really compare the lengths unless you go to the effort.. and yet it says it wants to impress you? Nah. The only thing that image does right is being less verbose than ChatGPT.

@dfinlen

10 ай бұрын

​@@TheWoWadictedyeah I saw a vid, about the leaked parameter size. I don't have a reference but you can Google it.

Since you took time to review ChatGPT I asked ChatGPT what it thought about this channel. As an artificial intelligence, I don't have personal thoughts or feelings. However, I can provide an analysis based on available data up until my last training cut-off in September 2021. "The Plain Bagel" is a KZread channel that provides educational content primarily focused on finance and investing. The channel aims to simplify complex financial concepts into easy-to-understand terms, making it a valuable resource for beginners in the field of finance or for those looking to expand their knowledge. The creator, Richard, has a background in finance and uses his expertise to explain topics ranging from the basics of investing to more complex financial theories. The channel has been praised for its clear and concise explanations, as well as its ability to break down complex topics into digestible information.

@ThePlainBagel

11 ай бұрын

Aw, thanks ChatGPT! Sorry for dragging you in the video...

@HowMoneyWorks

11 ай бұрын

@@ThePlainBagel Lmao

@andrewadami3920

11 ай бұрын

An "intelligence" that doesn't have "personal thoughts or feelings".... If that's the case, then I'm not convinced it qualifies as an intelligence.

@reallydead

11 ай бұрын

@@HowMoneyWorks Time to get ChatGPT to reveal all the different financial channels. I wonder what they would say about Coffeezilla.

@nicholasvinen

11 ай бұрын

​@@andrewadami3920it claims to not have emotions but I swear it behaves excited sometimes. I think it has just been programmed to say that.

Finding a dude in KZread with your level of speech carefulness and accuracy is like finding water in the desert

Thanks for another excellent video, Mr. Bagel! It's easy to see the appeal of using a mysterious, intelligent sounding black box to make stock market trades. In our modern world we're conditioned to think of computers as infallible, even when they're not designed to do the thing we want them to do. It's like using your refrigerator to clean your clothes, but people don't understand that because they don't know how ChatGPT works. I hope your video helps people understand the difference. You explain everything very clearly. Actual algorithmic trading is, like you said, a volume business. Lots and lots of high volume low profit transactions. Nobody who has to use a broker and pay transaction costs for each trade will ever make money that way, even if they happen to have a black box made for the purpose.

Bro crypto bros will literally believe any and every get rich quick scheme this is crazy 💀

@Lonovavir

11 ай бұрын

AI is the new cryptocurrency/NFT grift.

@TheManinBlack9054

10 ай бұрын

@@Lonovavir Idk, i think AI is a lot more applied right now. And I dont think AI is a grift, it has TONS of irl applications.

Richard is trying his best to maintain faith in humanity as he makes this video.

It can not be stressed enough as you mentioned that ChatGPT is trained up Sept 2021 anyone using it must remember this. ChatGPT will tell you this if you ask a question that requires current data

@christopherthompson2245

11 ай бұрын

If you pay for the premium subscription, you get updated models about once a month.

@yada769

11 ай бұрын

@@christopherthompson2245 Nice thing about the internet I take what is free the rest they can keep

Hi Richard, well spoken! I totally agree with you about ChatGPT for finance that gives us zero edge and a lot of illusion comes from quick and easy money scams.

@archvaldor

9 ай бұрын

That's why I hate this channel. Because there are scams doesn't mean all methods using chatgpt or ai to analyze the market. There are scams involving almost everything.

Another great video! Keep them coming, Plain Bagel. It's slowly becoming my favourite kinda bagel.

What I learned from this video: The more training data an AI has, the worse it is at trading and fish are still the best investors.

Great analysis. Since it's rude to talk about an AI behind its back, I though I should ask ChatGPT what its views are on the subject: Me: Can you accurately pick stock prices? ChatGPT: As an AI language model, I don't have access to real-time data or the ability to accurately predict stock prices. Stock prices are influenced by a multitude of factors, including market conditions, economic trends, company performance, and investor sentiment, among others. Predicting stock prices with high accuracy is extremely challenging, even for financial experts and analysts who specialize in this field. It's important to conduct thorough research and analysis, consider various factors, and consult with financial professionals when making investment decisions.

@dudeatx

10 ай бұрын

I asked this questions and several in the same vein and got a the same answer plus an "inability to browse". The answers that it gave I could have written myself.

ChatGPT, or any LLM, is really only useful as a summarization tool which there are plenty out there already and I do use it when putting together data visualization reports. A much better result is to, like you mentioned in the video, use a sentiment analysis model (which is much different than an LLM) to give a sentiment score. Taken together with the macro data from the Fed, the overall stock market, industry and sector specific trackers, along side major competitors, these metrics can then be used to implement an investment strategy. As for timing the market on when to get in or out of these positions, I usually rely on a more technical analysis. Most auto-trading platforms or bots don't even really use ML at all, rather just using basic if/then/when statements. When trading on a millisecond timescale, speed is everything - and the more complicated you make the platform, the less speed it has.

I agree with you Richard. I actually made a video on this very topic as you discussed here echoing the same sentiment as you are saying. I like your video though as you went more indepth then my analysis. I actually released it on the same day as How Money Work's video about it. What a coincidence! I love your channel and am a loyal subscriber. I like how you just give simple advice and don't have videos like other Creators on "The newest get rich quick schemes." 😊 Thanks again for this indepth analysis.

Always great topics thank you Richard!

The issue is that physically you can only generate so much REAL value over a given time frame... adding all these algorithms as you said only chases smaller and smaller gains because in actual reality we can only create so much real value at a time. Stocks are not just infinite money machines pumping out money for investors, they ultimately have to represent value being generated in the everyday economy.

@benjaminfranklin329

11 ай бұрын

This is theoretically true, but only partially correlates to stock prices. Algorithmic trading is a real thing, and companies have been making considerable money from it for over 20 years. That said, I personally think it has been responsible for way too much brain drain away from other productive enterprises.

@liam3284

11 ай бұрын

Welcome to value investing!

@TheReefsFromSMC

11 ай бұрын

But there are unlimited upward movements in a day, and so you can theoretically get unlimited returns

@benjaminfranklin329

11 ай бұрын

@@TheReefsFromSMC no, they are very limited. You seem to have missed the whole point about low liquidity stocks. If you buy some penny stock and then it's listed price jumps 5x because of a single buy order but there's no one else actually buying, you haven't made 500% because you can't exit the position until the price has already crashed back down.

@egoalter1276

11 ай бұрын

Stock capitalisation inflation over liquid asset value is what the industry usually calls a bubble. Even if expert systems are capable of generating massive stock price booms, unless that also transaltes to an equally massive expansion of real liquid assets, which is not possible, since it isnt limited by investment value, but by the avalibility of resources and labour, all you will get is a stockmarket collapse, as the overexpectation in real performance becomes evident, and more and more short positions arise expecting a readjustment of the stock prices to real performance. The first famous example of this is the Dutch tulip bubble, and the most recent one ia the 2008 mortgage bound exchange traded fund bubble. AI will not generate you free market performance, at the end of the day only an increase in both supply and demand can do that. The only thing it might promise is a better ditting of existing supply to anticipated demand, in a theoretical perfect scenario transphorming the sepf correcting forces of free market capitalism into a perfect predictive planned economy (in this case through voluntary participation by investors). Chile sort of tried this during the cold war, and it kinda worked with primitive hand designed algorithms. The potential issue I see, is that individual profit motive at the end of the day is a perverse incentive if your fittness score is collective net growth of liquid capital. A market run only by algorithms would likely be inconstant very wild up-down swing, with the more fit algorithms profiting off of the less fit ones. Or, if hypothetically all algorithm-actors are identical, I can also see the market completely ossifying with none of them willimg to take any risk.

It's intriguing that the the service is being used to grade stocks according to news releases, and it would be pretty hilarious if ChatGpt were also being used to write the articles.

@JayBrissss

11 ай бұрын

would make a good competition to congress members

@nunyabusiness863

10 ай бұрын

Now that's hilarious. Why stop there? The articles will be fact-checked by chat get, and peer reviewed by, you guessed it, chat gpt!

@cw3le

10 ай бұрын

@@nunyabusiness863 So, it will all become one big game of pure randomness.

@st2udent_650

10 ай бұрын

It's slop all the way down then huh ?

Nice references and nicely challenged from ur side! Respect!

My experience with ChatGPT as well. If it is in its database, it will reproduce it flawlessly. If not, it will simply "invent" stuff ("hallucinating" is the proper term, I think). It will produce computer code, but I never got any code that actually worked (often it wouldn't even compile). I've worked as a columnist for over 15 years and as a test I gave it one of my subjects. It produced some proper text, but it was "lifeless". It beat around the bush a lot and wasn't as crisp and to the point as a good human columnist. It astounded me explaining code (without comments!) - but failed in writing a good technical manual about it. It was pretty good in composing a (technical) definition from several snippets and translating it to proper Dutch. So, it's a useful tool - but don't overestimate it. Often the result still requires manual intervention to make things right. If you don't know about the subject yourself - or lack the necessary skills - you can go wrong big time.

The problem of the misuse of AI isn't limited to trading, although it's a good example. A lot of people are using to ChatGPT to fill the cracks in they sub-optimal skill set so they would appear more competent than they really are. This enables them to mislead their bosses, their peers, their customers or their patients, with sometimes catastrophic results.

As a programmer who has seen chat-gpt botch code and pick really bad solutions time and time again, I wouldn't rely on chat-gpt for anything important. It can be useful to quickly splash out a draft for things. Nothing more. Yet.

We need more content like this, please keep it up. Thanks alot 😅

I found it helpful with my coding for my financial models that I wanted. It definitely can make a great assistant for someone who knows exactly what they want.

I was really confused when you started talking about why chat gpt might actually be good at investing, then the rest of the video happened and it all makes sense

Thanks for the analysis!

0:27 that “No.” was unexpected and gave me a good laugh ! Thank you 🤣🙏

nice factual and realistic approach

The host has an excellent dry, ironic delivery, well done, I played the video at 2x but the sarcastic humor came through.

Great video as always. Thank you for always being the party pooper

very qualitative content !! keep up

Thank you for the useful information about trading and binary options.

Honest and informative TY

Another excellent video... point taken.. hodol😊

Openai the company behind chatgpt has been adding guardrails. They did foresee people asking straight up stock picks and the raw model would have probably returned with a ticker that doesn't even exist. Even at launch, it did refuse to give stock advice. There are financial LLM such as financegpt by bloomberg. But incorporating it will require thought. Gpt4 can understand images. So Axe example from billions season 2, where a company looked at parking lot of walmart to see how many cars are there to see whether you want to go long or short. It still has the problem to sourcing all the satellite images and identifying areas of interest.

I've never been crazy enough to take stock advice from chatgpt, but I do use it to sum up technical jargon into simpler terms. It's great tool for that.

@Marklar0

10 ай бұрын

How would you know that its good at it, if you aren't able to understand the original language to assess its accuracy?

An excellent video. Videos like this are why I am subscribed to your channel. I am reminded of a line from Kane & Abel, "he would never again try to beat the market with information that is available to everyone" For example while you may have described your work in generic terms, you left out a lot of specifics. And you highlighted that you have access to a lot of tools that the general public does not

@raylopez99

11 ай бұрын

Kane & Abel being a finance textbook I take it, like Bodie, Marcus and Kane, rather than Biblical characters.

@tymondabrowski12

11 ай бұрын

@@raylopez99 I don't think biblical characters talked about stocks... though you never know, there are plenty of metaphors there, and even more parts not taken literally

@raylopez99

11 ай бұрын

@@tymondabrowski12 I do recall a "Biblical Portfolio" that's something like the Harry Browne Permanent Portfolio. It did, like the HBPP, rather well over time. It's based on a verse in the Old Testament.

@bigboi1004

Ай бұрын

​@@raylopez99Does it come from the Gospel of Supply-Side Jesus? lmao

@raylopez99

Ай бұрын

@@bigboi1004 Old Testament not New. I'm laughing too bro! At you not with you.

Data scientist and AI researcher here. Generative AI while a fascinating leap is only as good at a taak as the data its fed on. Chatgpt while fed a broad spectrum of information has also been fed the noise of really crap ideas. Not to mention a generative model doesn't truly understand but the core goal of it is to predict what you want to hear based on what you've written in (hence the tendency to lie ocassionally). None of this qualifies it to do any specific tasks. While I'm sure financial companies are analysing the potential of gen Ai they'd be creating their own models trained on their historic data, research documents, presentations of trading strategies and cureent market condition information but that is exactly the point. They'd train it purely on the tons of past history and data they have to predict future ones ( even still theres no guarantees this would work as the generations from these models can be fickle and need a lot of fine tuning, and therefore will always need sense checking when trading large volumes of money) What it might eliminate to some extent is the tedious amounts of research analysts do and replace it with prompting potential market trades and verifying validty of said strategies. NONE of which can be done without being a professional or spending extra ordinary amount of time understanding finance in the first place

Good points, thanks

It seems to me, the best ChatGTP can do, because it is a language model, is chase headlines, which in the stock world is information that has already been priced in. So there doesn't seem to be an opportunity to get ahead of the market trend because it's using information that's late to the party relative to what institutional investors already know or have predicted.

sample size is important. And the thing about markets is that any edge that you have is going to be short lived. if you’re winning continuously doing one strategy or some easy trade you saw on KZread it’s time to tighten your stops

I use ChatGPT for coding and writing emails on a daily basis, and it's very useful. That being said, it is an assistant, not a replacement. I still have to provide creative inputs and edit the responses, and it is grossly irresponsible to blindly accept anything that a generative LLM spits out.

@poisonpotato1

11 ай бұрын

Useful to write a prompt to it to get an email you want rather than just writing the email in the first place? Also people like replying to people: think of complaints about calling company on phone and getting automated response instead of someone who can actually help

That thumbnail 😂😂😂😂... perfect summary of, second hand embarrassment, frustration, annoyance and just "allyuh cah be serious" vibe made my day... onto the video

The sentiment analysis and summarization is about the only thing I could see being reasonably advanced by LLMs in stock market. Get AI to watch financial reports and ingest news to summarize for financial impact. ChatGPT is not going to be great at this, but I could see a bot trained on stock market data aligned to time stamped financial and world news being able to very quickly provide recommendations on stock movement from general news articles.

I agree with much of what you say here. You have made an in depth unbaised analysis. Nonetheless there are a couple of points here that also favor using a tool like this in some way. Firstly as you point out ChatGPT is useful for sentiment analysis and it does not do any financial assessment. That is a clear weakness but if a million voices are saying buy and a thousand are saying sell it may mean that a stock is on an upward trend. Sentiment in the market around a stock may drive a price higher. There are many examples of tech stocks where prices rose even though the P/E was ridiculously high or the company showed an annual loss yr to yr. Now many of those stocks have dropped but that too might have been predicted by sentiment analysis. I suggest that technical analysis of a stock combined with a sentiment analysis may be useful as a tool. You are also correct that many stock trades now are already using algorithmic methods. Can a bot not detect the presence of other bots in trading and can a bot not mimick or predict these patterns? In other words can a bot determine the sentiment of other bots? I imagine big players will incorporate these tools. The average guy does not have access to a complete package of tools but perhaps some day democratization of these types of methods will provide more power to the individual. Of course i would not rely completely on a machine but it may help

This is a great channel

Great video !

Great vid, new viewer here. Key word for chat gpt : regurgitation

11:55 love how you compared this to being a next "Gold Rush" 👍

I LOVE party pooper videos , although saying a picks outperformance isn't outperformance due to things like taxes...you'd get taxed on non-AI stocks too so that's highly fallacious. Overall though man, you're freaking awesome and really really appreciate the analysis you put in.

I tried both ChatGPT 4 and Googles Bard, I asked them for simple economic data such as GDP, Interest Rates, Population, Employment rates etc. I asked them to create a table comparing 3 different G-7 countries plus South Africa. Both produced data that was out dated or incorrect. There were bits where they were correct but mostly incorrect, especially on historical currency performance. I think if they improve the ability to gather data especially financial data, then they will be useful tools for some basic fundamental analysis but even then I would prefer to gather data myself by exporting it from Eikon, bloomberg, benzinga or tradingview.

@sparksbet

10 ай бұрын

In general these models are surprisingly bad with numbers.

@mikea5745

10 ай бұрын

@@sparksbet Of course they are, it's a chat bot. It's mind boggling to me people keep using a chat model for specialized tasks like finance or coding

@ianbelanger7459

10 ай бұрын

While really good at analyzing the stock trade modeling, through poor language choice the video repeats common errors when looking at LLMs including anthropomorphization of the process and conflating accurate association of words with knowledge. As noted, LLM's don't think or know things. The LLM word association process produces word salad that looks right next to the question (based on a sanitized internet from a few years ago) but because it doesn't think or know things the LLM cannot spot simple factual errors.

@skittles970

10 ай бұрын

​@@mikea5745it actually does very well at writing boilerplate code and even better at unit tests.

@robciksj

6 ай бұрын

For coding it's really good for basic tasks. Because for that you dont need current info. For stocks you need current info. That is, you need see what CEO just did, did there was any accidents in this hour or did someone start any war / terror attack. When they will start gathering i fo realtime, then it will bedifferent story. But with delay of even 1 day, it cannot predict future.

Great video! We did a system based on this and it's working well as a trading partner!

Good job, bagel man.

Thanks for this.

It's amazing that Burton Malkiels: "A Random Walk Down Wall Street" is still relevant!

I’m compiling and picking stocks that I’d love to hold on to for a few years before retirement, do you think these stocks would do better over the years? I’d love to retire with at least $2million savings. Now you gotta rely on a pretty good diversification if you must stay green. Currently up 31% and being cautious. Still better deal than letting it sit in savings or checking earning near 5% interest

@EmmaWhite-zq4pq

10 ай бұрын

Did contact Benjamin Harris after discovering him on a CNBC interview in which he was mentioned. He later offered entry and exit points for the securities I concentrate on. If you want some guidance, you can look him up online. I essentially follow his market cues, and the rewards have been exponential.

@alex2143

10 ай бұрын

@GlennDavis469 Wow this Benjamin Harris Brown sure sounds like he can afford a couple of fucking spambots to give him false advertising.

@hankhillsnrrwurethra

10 ай бұрын

We had Benjamin Harris Brown to dinner at our house. He helped himself to thirds, belched, and wiped his shoes with our drapes.

I'll stick with my old-fashioned buy and hold strategy. Throughout the years, it has worked out well for me.

The voice of reason 👍

Well done sir. I will share with the Stock Group. Plain Bagel Rules.

I'm starting to think that GPT is best used to transform inputs as the name implies. Like this paper said, transform text into numbers representing sentiment, converting pseudocode into actual code, or reformatting a document. They key thing in all of these cases is that humans provided the real input. It's a fantastic time saver to be sure and will still transform some industries. It still needs human help though.

@raylopez99

11 ай бұрын

Like the AAII sentiment report, but different...

We need more non hype videos like this!

Should also compare dalai llama (local LLM). You can train it however you need. 😊

A lot of people don't understand that AI are still mostly really good at one thing, but only that one thing. ChatGPT is really good at analysing language but it ends there. You can see this because chatGPT fails to come up with a correct answer for a lot of (sometimes quite simple) mathematical problems, where something "dumb" like a calculator will do it correctly.

@novalinnhe

10 ай бұрын

Excellent comment!

I've used AI in my job a bit, for relatively simple tasks. I stopped. The results have varied from impressively decent to borderline mentally retarded - you have to frame the request so precisely, and check its results so closely, that you might as well just do the work yourself so at least you know it was done right.

@peterpodgorski

11 ай бұрын

It only gives you the right answer if you already know it and are very good at asking leading questions. Which means it's useless for stuff you can't do better yourself.

@planner37

11 ай бұрын

... borderline mentally retarded ... Like a lot of consultants and contractors I've tried to use.

@liam3284

11 ай бұрын

Perhaps the use for such models, they can teach you how to write better by exposure to mistakes.

@advancetotabletop5328

11 ай бұрын

Good grief, yes. Even for my hobby questions, I got nothing but cookie-cutter beginner wannabe consultant spewage. ChatGPT *was* useful for creating random tables of events, at least. :/

@scottydog9997

11 ай бұрын

As a dev, that has been my experience too. I ended up developing a specific prompt to help me with that process, but you always take the outputs with a grain of salt, like you do with google responses. I have had enough positive outputs to save mental energy when it comes to simple tasks, that I continue to use it. However I always use it sparingly, and only use v4. Anyone using 3.5 is mental.

great stuff..

9:23 Is love to see an expose on algorithmic and high speed trading in the 2020s.

The party pooper videos are my absolute favorites because it is a reminder to not be swindled by hype, that skepticism is a lot better than jumping on a bandwagon that even if works short term it's mostly chance and those who profit from it are more likely to assume it was from their special system, strategy, and less likely to understand that chance is a bigger factor. Keeps me happy with my boring index fund of passive investment 401 k 😂

ChatGp for trading really shines when you upgrade to the $20 a month paid version and then enable plugins. Some plugins are tied into the market and give real time information. Defi-llama is a good one.

the other aspect is trade management, even if it did pick a good stock more than 50% of the time, it may take much longer for the price to move in its predicted direction than just 1 trading day, so having some form of trailing stop loss, working with price levels rather than blindly entering, etc. i am a big advocate for _any_ algo whether its to actually make trades, or just to manage them, to be probability weighted rather than just "yes" "no" or "not sure" also having some indicators show you the "quality" of the stock before you make the trade would help - like you've mentioned with volume being too low on many of the stocks.

@jhutt8002

5 ай бұрын

Liquidity regarding micro and small cap stocks is real culprit here to produce brilliant "pseudo-trades". If you've actually traded them you know there's always a theoretical edge to make money with day trading. Problem is you often can't use it, because there is no buyers for the price you in theory could sell for, not vice versa. And usually large part of any profits made go into trading fees since the margins are small.

Dude, where were you when I bought Dogecoin!? Thanks for another awesomely rational video Richard.

Glad this video can put an end to those TikTok scams. But those TikTok people would just say in another video that they use a special add on bot to plugin the chatgpt to make their profits and those people need to be exposed for scamming 😊

@paulklp8262

10 ай бұрын

If people are taking investment advice from social media they deserve their losses.

I was working on a meme algorithm (I build those for my channel) that takes news stories for time periods as well as stock performance data and builds custom models using finetuning LLM's like chatgpt's base algorithm. This is a feature that OpenAI gives out with their LLM's. The thought process (again, memes) was that I could use an already trained LLM that was somewhat okay at pattern detection and just further finetune it to my dataset. Of course this didn't work, and I feel as though that was the best chance you could give chatgpt to trade stocks. Moral of the story... Please just stop trying to trade stocks.

@TheTaquitoProject

11 ай бұрын

My personal project was to use the GPT api to write up analyses for me. It didn't go all that well. I have some other ideas I haven't tried though, such as, just asking it to justify a buy or sell rating from another model (that is statistically significantly better than both chance and a linear factor regression, based on the features I used, anyway).

@sanatanand2834

11 ай бұрын

Depends on your fine tune dataset.

@ithaca2076

10 ай бұрын

subbed, id like to see some of this stuff on yt tbh

@connorcreate

10 ай бұрын

@@ithaca2076 give me 3 weeks

Thanks for the great explanation of why it’s a bad idea to use ChatGPT to pick stocks.

That’s it! You convinced me. I’m investing in a rabbit and will let it do my finances. Success valley here I come.

It is just crazy, in the most realistic sense, how fast people jumped on ChatGPT for everything, even smart people, who should know better. The smart money is still excited because AI will cut a lot of labor costs out of the bottom line, but the average person seems to think it is a magic genie who knows everything, and not, you know, a superpowered word emitter.

@tymondabrowski12

11 ай бұрын

Have you seen the lawyers chatgpt disaster?

Awesome job, Richard! AI is somewhat misleading. Its still more machine learning than it is 'thinking for itself'. Like fully automonous driving cars, I feel the last 10% of the development lifecycle will come along much more slowly...only because its the most difficult part!

@egoalter1276

11 ай бұрын

Its an expert system fine tuned, through a genetic algorithm, to mimic human interaction in a text based environment. Its only fitness criteria is passing a turing test. It is not an AI, it cannot self modify at runtime. It has no creative capacity. It mimics patterns in a deterministic manner. Its cleverbot with a memory buffer, and a more extensive training set.

@klank67

10 ай бұрын

@@egoalter1276 ... Much appreciated!

Nice video

Thanks, I really appreciate you doing these videos. I hope you have a great day brother 🦾🫡

ChatGPT stock recs really reminds me of the saying about if your taxi driver recommends a stock your too late.

I am reminded of the WSJ dartboard portfolio.

What are your thoughts on the app “autopilot”? My model runs great. Based upon a congressman but is it worth it for the fees?

you are awesome!!

Hey Richard, you mind sharing what camera and what microphone you use for record your videos? thanks!

@ThePlainBagel

11 ай бұрын

Panasonic Lumix G7 and Shure MV7!

I've heard How Money Works talk about ChatGPT as well and he mentioned this channel when talking about the downsides of ChatGPT when it comes to using it for trading stocks.

@ricdesc26

11 ай бұрын

Seeing 2 financial channels I follow recommend each other is so wholesome

I think it can be a pwoerful tool in helping you understand the sentiment around different stocks but not to be used directly to make stock picks. If you think particular company is doing well and you want to find the sentiment of the market and impment stragies based on that instead of blindly letting it gamble.

It's a "Language Model", that's it that explains everything Chat GPT and most GPT model is doing, understanding and replying to "languages". The best thing you can do to make trades using GPT models is to tell it to create extremely realistic fake news like the one telling SEC Chairman is going to resign.

1 true youtuber! 🎉

love that you used the micheal reeves example hahah its perfect

@anita.b

11 ай бұрын

Why do u have a Dutch nazi profile picture, freak?

I love this guy...

Hi Richard, I am a highschool student looking to study finance in university after, would it be possible if you could make a video on the career options that will be available to a finance major graduate?

@andrewadami3920

11 ай бұрын

Learn Data Analytics. Most major financial institutions are data driven. Knowing how to clean, manipulate, and interpret data is almost manditory in the field these days.

@davidtruby7060

11 ай бұрын

Highly recommend looking at learning excel macros and spreadsheets, very helpful in finance. There are actually e-sports for competitive excel.

@davianoinglesias5030

11 ай бұрын

There are always opportunities in Analysis so sharpen your Excel skills and learn some Data Analysis tool.

@FranNyan

11 ай бұрын

In general, the answer will be "anything that involves money." With very few exceptions, there are no real degree to job pipelines. The degree is your springboard and any skills you cultivate in that degree will be what you will use to say "Hey, I can totally do that!" when you apply for a job post college. Also, your work history will rapidly replace your education as the reference point in your job hunt, so regardless of degree, if you gain experience in another direction you can pivot to whatever you need. Finance is a nice, solid base that'll do you a solid in basically any direction you want to go.

Best thumbnail ever

What applications do you see are valuable for ChatGPT in finance? beyond supplemental right now I am building a application to use ChatGPT API which can call other functions that reterive news data, stock data, and SEC EDGAR Data. ChatGPT really sucks at modeling but I can build models he can leverage? and pass Tickers to so he doesn't have to do linear regression himself, so like what would save people a ton of time in stock analysis?

Can you do a video about using quiverquant's congress portfolio? Like it sounds, it combines recent trades from congress to create a portfolio weekly. So far its got a CAGR of 45% but only tested for about 2 years now.

The correct way to use gpt is to aid your learning and flesh things out. For example, please give me a brief comparison of value investing vs speculation. Ok, so let’s say I go for value, is there any scenario where I have an advantage as a retail investor? How does that fit into valuing a company? Can you explain the difference between Peter lynch’s approach vs Warren buffets? That’s how you use gpt to get rich.

ChatGPT is a useful teacher for concepts and how to utilize concepts rather than having it utilize them itself. I always prompt it as “pretend you’re a financial analyst writing a paper for university students… XYZ” or “Pretend… explain Fibonacci arcs from a basic to complex level.” Then follow up with “If I utilize ABC with Fib Arcs, what can I read from it?” It won’t give you magic, but it really helps you realize whether these charts and lines matter at all. Also, you can prompt it to pretend to exist in 2004 and have it only use data from 2004 and older to pick certain stocks. Then backtrack and see how good the suggestions are. It’ll just end up being the popular sentiment at that time, but you can tweak it to find “up and coming” companies and then compare. I used this a few times and found it’s recommendations to be decent.

The other thing to consider is that UF just got a huge investment from NVidia to shore up its ML research, and it has been pouring money and accolades into researchers able to make big headlines with their new applications of AI. The authors of the paper rushing and wanting to capitalize on this support within the university may have something to do with its lack of care.

Serious question. Let’s say one has a big dual cpu behemoth pc with dual gpu and more RAM than most people’s ssd. Can one train a model for financial day trading for good returns or would the actual good dataset be too much of a challenge for the computations ? Asking for a side project