Post Awam9AfwRM3dJkIT9k by lffontenelle@mastodon.social
(DIR) More posts by lffontenelle@mastodon.social
(DIR) Post #Awa0cmQSMwmwzo6gEa by futurebird@sauropods.win
2025-07-28T03:41:11Z
1 likes, 0 repeats
Today my husband asked me what I thought about using an AI to pick stocks because that's what his friends are working about and he wanted to know "what the computer people are saying on your hacker network" (that is what he calls Mastodon)I showed him all the talk about the bubble. "OMG they all HATE it.""I'm certain the stock picks would sound reasonable and look nice, but maybe some of the tickers might not exist."
(DIR) Post #Awa0fxMMApYxm8hgh6 by futurebird@sauropods.win
2025-07-28T03:41:45Z
0 likes, 0 repeats
I thought everyone was talking about the very obvious "new bubble" but... it seems not.
(DIR) Post #Awa0qby62vuacJR99M by delProfundo@aus.social
2025-07-28T03:43:37Z
0 likes, 0 repeats
@futurebird 😂 yes please sell me two shares of ‘pickle in a toaster dot com’ Mr guessing machine that is even less suitable for this task than any of the others it can’t do 😂😂😂😂😂😂😂😂🤬😡
(DIR) Post #Awa0svTEVamngI6X2G by randomgeek@masto.hackers.town
2025-07-28T03:44:04Z
0 likes, 0 repeats
@futurebird at my extremely nontechnical new workplace, they're disappointed they can't use ChatGPT to write their reports.I try to reassure them they can make up plausible-sounding nonsense on their own just fine, and it'll cost the state less.
(DIR) Post #Awa0yJAfokOgdRzQTg by delProfundo@aus.social
2025-07-28T03:45:02Z
0 likes, 0 repeats
@futurebird but seriously the answer is ‘anything reliant on math or logic is not a suitable use case for LLM’s if you want more than a 5% success rate’
(DIR) Post #Awa11gnuPKxwnexD9M by andytiedye@sfba.social
2025-07-28T03:45:38Z
0 likes, 0 repeats
Computers certainly can produce βμ||śɧ¡ŧ at superhuman speeds.@futurebird
(DIR) Post #Awa1N9yDAD41L30jUu by BurritoSommelier@techhub.social
2025-07-28T03:49:32Z
0 likes, 0 repeats
@futurebird well worth the read: https://www.wheresyoured.at/the-haters-gui/
(DIR) Post #Awa1jkEF2Hk4ABHIqu by futurebird@sauropods.win
2025-07-28T03:53:32Z
0 likes, 0 repeats
@apLundell I don't think any of them know?I think someone at work is being seduced by some marketing and he wanted to know what to say about it.Everyone has tried to make programs to trade, the problem is these programs keep re-shaping the market.But in practice as retail traders they aren't going to be making day trader moves or using day trader money.
(DIR) Post #Awa1nra1e5gS7aQni4 by PizzaDemon@mastodon.online
2025-07-28T03:54:20Z
0 likes, 0 repeats
@futurebird I'm a hacker because whenever remotely plausible I say, "I'm in!"
(DIR) Post #Awa1wcmY29LEpfSy7E by ohmu@social.seattle.wa.us
2025-07-28T03:55:54Z
0 likes, 0 repeats
@futurebird I have noticed from peaking outside that it is here I find my cabal who find all [gestures].this about AI to be risible in the same way I do. Out there, accounts in their tens of millions are grabbing with both hands.
(DIR) Post #Awa2MXQSZRWZkWUWIK by indigoparadox@mastodon.social
2025-07-28T04:00:37Z
0 likes, 0 repeats
@futurebird I was going to joke about going all-in on $YEET, but I fact-checked to see if that existed and it seemed to be a crypto exchange... wich still seems appropriate, in this case.
(DIR) Post #Awa3TxCnDlxBgVCNXM by moira@mastodon.murkworks.net
2025-07-28T04:13:10Z
0 likes, 0 repeats
@futurebird they never know. doesn't even help to tell them, most of the time.
(DIR) Post #Awa3ySwvlHSUpCKQUK by knwmoon@mastodon.online
2025-07-28T04:18:40Z
0 likes, 1 repeats
@futurebird I’m super not looking forward to the inevitable avalanche of “nobody saw this coming!” opinion pieces 😫
(DIR) Post #Awa45jNP5K15ncAvE8 by sabik@rants.au
2025-07-28T04:19:55Z
0 likes, 0 repeats
@futurebird Hmm, if you could work out what stock advice ChatGPT tends to give, you could maybe trade against it?
(DIR) Post #Awa4NS0MPXah8OqXOC by farah@beige.party
2025-07-28T04:23:11Z
0 likes, 0 repeats
@futurebird Umm what bubble?
(DIR) Post #Awa6EzJrNYJllA3kky by grumpasaurus@infosec.exchange
2025-07-28T04:44:04Z
0 likes, 0 repeats
@futurebird this video comes to mindhttps://youtu.be/USKD3vPD6ZA?si=4CJX0a8nKYes54oF
(DIR) Post #Awa6OJXfYonl5MxwWm by log@mastodon.sdf.org
2025-07-28T04:45:41Z
0 likes, 0 repeats
@futurebird My working hypothesis on picking stocks is that the amount of loose money chasing after good investments drowns any signal from business fundamentals in the noise of people desperate to outrun inflation. There are so many investors smarter and more skilled, and so many dangerously stupid but rich, that the only way for ordinary folk to win is to buy no-fee whole-market index funds and hold forever. AI would be trained on the noise, and would only amplify it.
(DIR) Post #Awa7H1nvD9Gk02eomu by cy@fedicy.us.to
2025-07-28T04:55:16Z
0 likes, 0 repeats
The stock market is a rich man's game. The rich can use financial trading to anonymize themselves, so they don't get hung for their crimes. And the rich can manipulate the market itself, driving stock values up by buying, driving them down by selling. Not to say all the insider trading dirty money can buy.Anyone who's not ultra-wealthy will most likely get fleeced, because every stock they have will be devalued by wealthy investors dumping it, and every stock they buy will be less valuable than the one a wealthy investor decides to pump. Every action on the stock market feeds your money toward richer people than you, and never vice versa.
(DIR) Post #Awa7OEBPaL42Q34sK0 by futurebird@sauropods.win
2025-07-28T04:56:59Z
0 likes, 0 repeats
@farah The AI bubble. Companies who have enjoyed growth because of AI are overpriced.
(DIR) Post #Awa7Zzx8JpOblA0mPo by cinebox@masto.hackers.town
2025-07-28T04:59:04Z
0 likes, 0 repeats
@futurebird I bet it would do about as well as this guy's fish, which is actually pretty good https://www.youtube.com/watch?v=USKD3vPD6ZA...which says more about investors than about the LLM/fish
(DIR) Post #Awa7bZHMSiDDbzDnua by n_dimension@infosec.exchange
2025-07-28T04:59:23Z
0 likes, 0 repeats
@futurebird So you can guess market movements...with 62% accuracy (which is awesome BTW).Jim Simmons, mathematician cracked the algo in 1982.His hedge fun, Medallion is closed to investors now.What we know of his algo, is that it uses many, many data points and signals, a massive amount.In theory, #AI could possibly stumble upon a successful algo. But so far, every hedge fund algo has resisted efforts to back-fit market data successfully.Hedge funds make money by (as the name suggests) hedging their bets. They make money when the markets go up, they make money when the markets go up.But most of all, they make money by cheating. They steal shares, fake them, and cheat. They pay fines, but the profits are bigger than fines.Market makers especially, because they can read the orders (amongst other shenanigans)#finance #markets
(DIR) Post #Awa8448bK87pQViG4u by xinit@mastodon.coffee
2025-07-28T05:04:29Z
0 likes, 0 repeats
@futurebird Using a model trained with historical market trends, company information and performance, it's possible that this might work okay. Nobody's going to share this model. You'd have to make your own. Good luck.Asking a general purpose language model? Forget it.
(DIR) Post #Awa8I8PQEvhwjCI0MS by bweller@mstdn.social
2025-07-28T05:07:02Z
0 likes, 0 repeats
@futurebird It does seem like the hype is driven by people who could not invent or build it themselves from first principles, much less recognize the limitations of any particular approach, and I suspect that also applies to the CEOs of AI companies.@apLundell
(DIR) Post #Awa9Fh7Mpwa7SdFER6 by Jdreben@mastodon.world
2025-07-28T05:17:47Z
0 likes, 0 repeats
@futurebird the best way to use language algorithms is to quickly access information you can verify yourself. Too often I hear people desiring to delegate their own decision making. It is an illusion that it is an advisor! I have had this same conversation with people. I find it interesting how people are wanting to trust it when they should be imo using it to come to decisions that they trust themselves
(DIR) Post #AwaAHOaIoX9aVw9bIO by H4Heights@mstdn.social
2025-07-28T05:29:18Z
0 likes, 0 repeats
@futurebird I can’t wait to see what havoc AI is going to cause as functions of government are put into the hands of psychotic bots. Real people will die. That’s a “when”, not an “if”.
(DIR) Post #AwaCGPt8gH3EB0cq36 by mina@berlin.social
2025-07-28T05:51:33Z
0 likes, 0 repeats
@futurebird I worked for more than a decade in algorithmic trading, and I can honestly say: Instead of using fancy computer models (doesn't matter whether you understand their Maths or whether it's a black box), it's cheaper and equally good to throw darts at the stock markets page of the Financial Times.And to your husband: I don't know a single person who really understands the LLM technology and does not hate the fact that it appears everywhere.1/2
(DIR) Post #AwaCfVq5h1A1s9ANKC by david_chisnall@infosec.exchange
2025-07-28T05:56:06Z
0 likes, 0 repeats
@futurebird I would expect LLMs to be uniquely bad at stock picks. The stock market is all about other people’s expectations. The value of a stock now is the consensus view of what the value of the stock will be worth in the future. It’s an n-step game of figuring out the move everyone else is going to make. You want to buy before everyone else buys and sell before everyone else sells.In contrast, LLMs are failing indicators. They are trained on a pile of data and will make predictions based on that, which will always lag the market.That said, if a lot of people use LLMs for investing, they will probably get the same recommendations. If a lot of them then buy, the price will go up. And they will then say ‘wow, LLMs are amazing at stock picks’ and invest more. You can probably use that information to fleece a lot of idiots, if you’re careful.
(DIR) Post #AwaDFxetjp3mZtJLe4 by Iveyline@mastodon.social
2025-07-28T06:02:41Z
0 likes, 0 repeats
@futurebird Fools gold?
(DIR) Post #AwaDfFU8b48B6aKtZA by EugestShirley@m.ai6yr.org
2025-07-28T06:07:14Z
0 likes, 0 repeats
@futurebird OMG, what great idea! In the1990s I used to get similar stock picks from random email tipsters.Pump and dump spam.
(DIR) Post #AwaFZB0sVhO9PGsm7U by ShadSterling@mastodon.social
2025-07-28T06:28:32Z
0 likes, 0 repeats
@futurebird @farah is that new? I thought we were several years in now.(I tried to search for when it started, but had a hard time finding anything that went back more than one year)
(DIR) Post #AwaGDxoaUBFELVaAZk by femme_mal@mstdn.social
2025-07-28T06:35:55Z
0 likes, 0 repeats
@futurebird If your hubby isn't familiar with the history of tulip mania, now's a good time for that lesson.https://ritholtz.com/2025/02/tulipmania/
(DIR) Post #AwaGgfW1bM2bzOaTlg by energisch_@troet.cafe
2025-07-28T06:41:06Z
0 likes, 0 repeats
@futurebird funny, :mastodon: as a hacker network :mastodance:
(DIR) Post #AwaIe6ZYkPT6wd9744 by nazokiyoubinbou@urusai.social
2025-07-28T07:03:05Z
0 likes, 0 repeats
@futurebird Oh man, there are just so many reasons this wouldn't work even if you specifically trained a model for stock picking somehow (which of course no one has.) Even with good training it still would effectively be picking randomly. Much simpler just to, well, pick randomly, lol.
(DIR) Post #AwaJo1wmtoVbRvJ7po by Joy_intl@mastodon.online
2025-07-28T07:16:03Z
0 likes, 0 repeats
@futurebird Just out of curiosity, did your husband ask *your* thoughts on this?I'm asking, bc I'm currently annoyed that I have been voicing my skepticism and negative view of AI to my hub, based on:- my experience using AI as a newbie webdev- reading people here on the hacker network ( love that)But hub kept talking positively until he watched a YouTube dude who said similar to my view, and thar seemed t give him pause. 🙄
(DIR) Post #AwaTRwbz3To64XgSeW by psentee@stolat.town
2025-07-28T09:04:06Z
0 likes, 0 repeats
@futurebird …for a moment i thought it’s about selecting stock photos. Like instead of „generate me a bullshit picture for the text” it was „pick a picture for this text from a stock photo catalog”. Which would make a surprising amount of sense as a use case for an LLM so of course it wasn’t right
(DIR) Post #AwaZGKxrHlibqplQsC by noodle@aus.social
2025-07-28T07:38:04Z
0 likes, 1 repeats
@david_chisnall @futurebird There was a defcon talk many moons ago about trading based on observing stock manipulation spam. In many ways analogous to what you're suggesting.
(DIR) Post #AwaaDcPE5AEN1vpgFU by BrKloeckner@piaille.fr
2025-07-28T10:19:57Z
0 likes, 0 repeats
@futurebird If ChatGPT were consistent, it could be worth to ask in order to gain information on what others do. An advice from it to buy some stock would result in a market value increase, which would likely burst later, opening the opportunity to bet against the stock. However ChatGPT probably goes all over the place and gives different advice to different people.
(DIR) Post #Awaf9toztAmDekvtFg by gbargoud@masto.nyc
2025-07-28T11:15:19Z
0 likes, 0 repeats
@futurebird Honestly probably about on par with all the other ways people pick stocks
(DIR) Post #AwagHXegHqClBBJcf2 by badrihippo@fosstodon.org
2025-07-28T11:27:54Z
0 likes, 0 repeats
@futurebird my reactions:1. Ouch. That's...good to know?2. We thought everyone knew about the bubble because we're in a bubble lol3. Hey maybe this is a good chance for us to go short some stocks 😁
(DIR) Post #Await4E2M0d6oisG3s by badri@snipetteville.in
2025-07-28T11:35:07.620745Z
0 likes, 1 repeats
@knwmoon @futurebird ooh maybe one of us can write one "nobody saw this coming" piece, as if it's in the future, but publish it right now! If my magazine was still running I'm pretty sure we'd have gone for something like this 😜
(DIR) Post #AwajZeLmQtSv1Eniee by mmby@mastodon.social
2025-07-28T12:04:48Z
0 likes, 0 repeats
@futurebird looking at the recent publications, invest in water purification and attached suppliers
(DIR) Post #Awal2Ye4p2uhg4hue8 by burnitdown@beige.party
2025-07-28T12:21:13Z
0 likes, 0 repeats
@futurebird i will laugh so hard at everyone who loses everything by doing that.
(DIR) Post #Awal3lzcgLiDMcMg8u by immersfer@mementomori.social
2025-07-28T12:21:26Z
0 likes, 0 repeats
@futurebird what, people are letting some language models pick stocks for them?
(DIR) Post #Awal7QpXJ3WUt4OKR6 by futurebird@sauropods.win
2025-07-28T12:22:09Z
0 likes, 0 repeats
@immersfer Yes. People are doing that.
(DIR) Post #Awal9T1y60KoI7B2J6 by immersfer@mementomori.social
2025-07-28T12:22:25Z
0 likes, 0 repeats
@futurebird 😬😬😬
(DIR) Post #AwalEDzLRyWJ5PDbfs by realn2s@infosec.exchange
2025-07-28T05:21:03Z
0 likes, 0 repeats
@knwmoon @futurebird And the "the models will improve" prelude 😬
(DIR) Post #AwalEF9f76Mohh3MOW by n_dimension@infosec.exchange
2025-07-28T05:24:52Z
0 likes, 0 repeats
@realn2s @knwmoon @futurebird They constantly are.You can't see it if you don't use them 🙄
(DIR) Post #AwalEG8zR5QJltkKzA by adrake@sfba.social
2025-07-28T06:18:37Z
0 likes, 0 repeats
@n_dimension @realn2s @knwmoon @futurebird I get plenty of evidence through my coworkers that they're still uselessly inaccurate without having to waste my own time pulling the slot machine lever.
(DIR) Post #AwalEGoSwsbHqX9SdM by n_dimension@infosec.exchange
2025-07-28T08:33:53Z
0 likes, 0 repeats
@adrake @realn2s @knwmoon @futurebird Is the problem with your cow-orkers unwilling to learn the basics of effective machine use (mitigated by WAAAAH HATE)?Or are is the problem the Engine.Im #Vibecoding as we speak.Its not random if you pull the rod the right way and bring about a happy ending 😅
(DIR) Post #AwalEHskyPcvA8AOxc by adrake@sfba.social
2025-07-28T10:37:25Z
0 likes, 0 repeats
@n_dimension @realn2s @knwmoon @futurebird I don't think so. They are broadly positive on AI (I'm by far the most negative outlier), but they don't seem to notice the time they waste on fixing its frequent mistakes and strange rabbitholes.
(DIR) Post #AwalEIotUG8C4RMpZw by ApostateEnglishman@mastodon.world
2025-07-28T12:22:59Z
0 likes, 1 repeats
@adrake @n_dimension @realn2s @knwmoon @futurebird Such is the power of the media. Most folks find it difficult to believe that they've been trapped in an echo chamber of hype and systematically misled. "What, *everything* I've read it about this thing is wrong? How absurd!".The evidence of their own senses won't necessarily convince them. They'll blame their own prompt-writing skills before they realise time spent "improving" those skills could have been spent doing the work themselves.
(DIR) Post #Awalki76yrPrmiLFOS by Emili@eldritch.cafe
2025-07-28T12:29:11Z
0 likes, 0 repeats
@futurebird ..... Are they talking about using machine learning algorithms trained on stock traiding or LLMs?
(DIR) Post #Awam9AfwRM3dJkIT9k by lffontenelle@mastodon.social
2025-07-28T09:36:37Z
0 likes, 0 repeats
@xinit @futurebird There are already sophisticated computer algorithms taking decisions about when to buy or sell stock. There were even blamed for how quickly stocks crashed in 2008. But they model stock prices, not human language.
(DIR) Post #Awam9C5VBo67hJGOdU by mmby@mastodon.social
2025-07-28T12:33:11Z
0 likes, 0 repeats
@lffontenelle @xinit @futurebird ask them whether they'd use an LLM to predict the weather (no cheating by querying proper models!)
(DIR) Post #Awam9CiqpVZbfLfoy8 by futurebird@sauropods.win
2025-07-28T12:33:38Z
0 likes, 0 repeats
@mmby @lffontenelle @xinit OK but isn't cheating the whole point of LLMs?
(DIR) Post #AwanKA8mruo6nZatqS by n_dimension@infosec.exchange
2025-07-28T12:26:39Z
0 likes, 0 repeats
@ApostateEnglishman @adrake @realn2s @knwmoon @futurebird Remember when you sucked at finding stuff on the internet coz you could not google?That.
(DIR) Post #AwanKBYLcMqbB8YpKC by ApostateEnglishman@mastodon.world
2025-07-28T12:45:14Z
0 likes, 0 repeats
@n_dimension @adrake @realn2s @knwmoon @futurebird Yes! But at least Google's algorithms *did* improve enough to make it genuinely useful (back before Google enshittified them again when they realised it was more profitable if people spent longer searching). We've already passed "peak AI", because there's no way we can stop genAI models from polluting their own datasets (the internet!) with error-strewn slop.You can only feed a chicken for so long on its own eggs before the chicken dies.
(DIR) Post #AwanKCYjsOkqIdkeZc by futurebird@sauropods.win
2025-07-28T12:46:49Z
0 likes, 0 repeats
@ApostateEnglishman @n_dimension @adrake @realn2s @knwmoon Feeding a chicken its own eggs is being charitable. I think it's like feeding the chicken something else.
(DIR) Post #AwauzrI62dTiB7mHVg by ApostateEnglishman@mastodon.world
2025-07-28T14:12:48Z
0 likes, 0 repeats
@futurebird @n_dimension @adrake @realn2s @knwmoon Something from the cattle fields, perhaps? 🤔
(DIR) Post #AwcnPFvVAob0VTjERs by RVLara23@mastodon.social
2025-07-29T11:57:09Z
0 likes, 0 repeats
@futurebird We are so doomed.
(DIR) Post #AwctLfIzTO6hmhO0ga by aeisenberg@cosocial.ca
2025-07-29T13:03:44Z
0 likes, 1 repeats
@futurebird i recently mentored a bright high school who wanted to learn about LLMs. She came up with the idea of using LLMs to “predict” the future price of a stock based on past performance and some recent headlines.I was impressed that she built what she set out to. Of course, it didn’t work at all to actually predict anything, and I used this as a lesson on what LLMs are *not* good for.
(DIR) Post #AwcuG1H0z0bI8NKlhA by futurebird@sauropods.win
2025-07-29T13:13:55Z
0 likes, 0 repeats
@aeisenberg That sounds like such a fun project. Hey why not try to just... predict all the news?Predict the headlines and see what happens. It won't work. But the way that it fails could be very interesting.
(DIR) Post #AwcvNqQFJLCUkbwlbE by Homempovo@ursal.zone
2025-07-29T13:26:30Z
0 likes, 0 repeats
@futurebird @aeisenberg I just asked some LLM what's going on in the world, and it said something about "former" President Trump, so not a good start. Unless it is a prediction, in which case I hope it comes true soon.
(DIR) Post #Awcw6bk2jwtQFbYGDw by undead@masto.hackers.town
2025-07-29T13:34:38Z
0 likes, 0 repeats
@futurebirdBack in, I think the late 90s, some guy thought that market trading was a lot of BS and went out of his way to prove it. He hooked up a cactus to water and light and some sensors. It would make a signal, and based on that signal a stock trade would be executed. If it made money, the cactus got light and water. If it lost money, the cactus didn't get water or light.Guy beat the markets, and his point was that even a cactus was smarter and more rational than wall st.
(DIR) Post #AwcwHQqdaBrdeMnVA0 by hugoestr@functional.cafe
2025-07-29T13:36:35Z
0 likes, 1 repeats
@aeisenberg @futurebird When I was a kid I had a TI home computer. The BASIC interpreter would tell you if there was a syntax error or a wrong command. My cousin and I decided that one meant yes and the other no.We spent like an hour asking it if there were aliens on earth, if we were going to be rich, if we could get super powers from chemicals potions. We gasped so many times.It is cute when you are first graders doing this