Investing General Discussion

  • Guest, it's time once again for the massively important and exciting FoH Asshat Tournament!



    Go here and give us your nominations!
    Who's been the biggest Asshat in the last year? Give us your worst ones!

Unidin

Molten Core Raider
842
490
More interesting data being floated around. Student loan payments resuming soon isn't going to help these stats.

View attachment 487116
I'm curious on the updated repayment rules and how they'll effect lower income borrowers, or those with limited savings. The base for not paying for a single person household is going up to 225% of the poverty line ($33k per year). It also is lowering the payment to 5% of disposable income for the income above that. Median household income in the US is about 70k, so many people will pay nothing or close to nothing in student loans.
 

Sanrith Descartes

You have insufficient privileges to reply here.
<Gold Donor>
45,092
122,689
I'm curious on the updated repayment rules and how they'll effect lower income borrowers, or those with limited savings. The base for not paying for a single person household is going up to 225% of the poverty line ($33k per year). It also is lowering the payment to 5% of disposable income for the income above that. Median household income in the US is about 70k, so many people will pay nothing or close to nothing in student loans.
I could be 100% wrong on this, but i was reading the site and it made it seem like if I swapped my loan plan to this new income repayment plan then it seemed like it started the repayment clock over again to 20 years. I figure I am going to wait until the courts have their say with all this and see where it goes from there.
 

Unidin

Molten Core Raider
842
490
I could be 100% wrong on this, but i was reading the site and it made it seem like if I swapped my loan plan to this new income repayment plan then it seemed like it started the repayment clock over again to 20 years. I figure I am going to wait until the courts have their say with all this and see where it goes from there.
Even if it's so, paying zero for 20 years is going to be the case for a large portion of borrowers.

Planet Money did an episode on it, and it seems like it's possible it will wipe out more debt than the $10k/$20k plan.

 

Blazin

Creative Title
<Nazi Janitors>
7,076
36,515
Blazin Blazin @Jysin so I know you look at different indicators, are you seeing a continued move on SPY down to the 100-DMA? I see a spot of possible resistance at $431, but really nothing else but air down to $428 and the 100.
let's use SPX in this case so we dont look like total rubes, 4300 and gap near 4220 are key levels. I think the current down is losing momentum. We may have struck a high on the yields in the short term, bears had nothing today. We had some froth come out of the market. 4-5% pull back is all it takes for people to lose their shit and be primed to be wrong yet again. I am not seeing the sky is falling stuff, credit spreads total yawn fest.

My presumed next move is that we are green next week, I'd say 70/30 split as we sit today. Does china destabilize further do investors give a shit? Tune in next week and we'll find out!

I put more hay in the barn today which is more satisfying than buying JEPI
 
  • 2Worf
  • 1Mic Drop
Reactions: 2 users

Sanrith Descartes

You have insufficient privileges to reply here.
<Gold Donor>
45,092
122,689
Even if it's so, paying zero for 20 years is going to be the case for a large portion of borrowers.

Planet Money did an episode on it, and it seems like it's possible it will wipe out more debt than the $10k/$20k plan.


But because they are doing this under executive fiat, it could be changed by a new administration.
 

Unidin

Molten Core Raider
842
490
But because they are doing this under executive fiat, it could be changed by a new administration.
True. But I don't think that one of these student repayment plans has ever been repealed. Not saying it wouldn't happen, but once programs go into place, it's rare they get pulled back.
 
  • 1Like
Reactions: 1 user

Sludig

Potato del Grande
10,039
10,709
"the fed was expecting a hard landing but are managing this to a soft landing... all fed data is better than expected"

For my supposedly libertarian not just a closeted Colorado Liberal, guy in my discord sure has a rosy take on why the voters are still going Biden.


Not the impression I get here, but I'm not good at pulling links out of my ass on why the fed data has been massaged and generally shit. I don't know how he overlooks inflation still awful, but my impression was sure we had a month or two of good markets but they are dropping hard again and I thought it sounded like earnings and many other important metrics are all pretty awful.

Posting here, not pol, because it's still a numbers question and I don't need the more purely ideological responses of Araysar etc
 

Creslin

Trakanon Raider
2,505
1,151
let's use SPX in this case so we dont look like total rubes, 4300 and gap near 4220 are key levels. I think the current down is losing momentum. We may have struck a high on the yields in the short term, bears had nothing today. We had some froth come out of the market. 4-5% pull back is all it takes for people to lose their shit and be primed to be wrong yet again. I am not seeing the sky is falling stuff, credit spreads total yawn fest.

My presumed next move is that we are green next week, I'd say 70/30 split as we sit today. Does china destabilize further do investors give a shit? Tune in next week and we'll find out!

I put more hay in the barn today which is more satisfying than buying JEPI
We are still in a rally focused on very few mega caps atleast in tech. I think nvda guidance could be a key to whether the ai frenzy gets some more fuel or not.
 

Blazin

Creative Title
<Nazi Janitors>
7,076
36,515
We are still in a rally focused on very few mega caps atleast in tech.
I disagree, significant broadening out over months. Apple and Msft with good pullbacks and people still just keep sticking to this line. Over at trillion has come out of big tech and market is a couple percent off recent high
 
  • 1Like
  • 1Solidarity
Reactions: 1 users

Sanrith Descartes

You have insufficient privileges to reply here.
<Gold Donor>
45,092
122,689
"the fed was expecting a hard landing but are managing this to a soft landing... all fed data is better than expected"

For my supposedly libertarian not just a closeted Colorado Liberal, guy in my discord sure has a rosy take on why the voters are still going Biden.


Not the impression I get here, but I'm not good at pulling links out of my ass on why the fed data has been massaged and generally shit. I don't know how he overlooks inflation still awful, but my impression was sure we had a month or two of good markets but they are dropping hard again and I thought it sounded like earnings and many other important metrics are all pretty awful.

Posting here, not pol, because it's still a numbers question and I don't need the more purely ideological responses of Araysar etc
Earnings were generally strong in Q2. This is in my opinion shows support for a continued bull scenario.

"Using data from the August 11th S&P 500 Earnings Scorecard, 91.2% of companies have reported results (90.8% of aggregate earnings). Of the 456 companies that have reported, 78.7% reported earnings above analyst expectations, which surpasses the prior four-quarter average of 73.4% and well-above the long-term average of 66.4%. This was the highest earnings surprise rate since 2021 Q3.

At a sector level, eight sectors posted an earnings beat rate above both its prior four-quarter average (Energy, Industrials, and Real Estate did worse relative to their sector averages)."

 
  • 2Like
Reactions: 1 users

Sludig

Potato del Grande
10,039
10,709
Earnings were generally strong in Q2. This is in my opinion shows support for a continued bull scenario.

"Using data from the August 11th S&P 500 Earnings Scorecard, 91.2% of companies have reported results (90.8% of aggregate earnings). Of the 456 companies that have reported, 78.7% reported earnings above analyst expectations, which surpasses the prior four-quarter average of 73.4% and well-above the long-term average of 66.4%. This was the highest earnings surprise rate since 2021 Q3.

At a sector level, eight sectors posted an earnings beat rate above both its prior four-quarter average (Energy, Industrials, and Real Estate did worse relative to their sector averages)."

Then does that not actually basically support the soft landing celebratory? I was going of most of the doomsday laughing at the Fed in this thread over last however long
 

Big Phoenix

Pronouns: zie/zhem/zer
<Gold Donor>
46,649
99,451
We are still in a rally focused on very few mega caps atleast in tech. I think nvda guidance could be a key to whether the ai frenzy gets some more fuel or not.
Speaking of NVDA, Intel is laying off GPU/AI workers;



Back when TSMC reported their q2 earning they said they dont see "ai demand" significantly effecting their revenue this year.

AMD didnt report anything of note in regards to AI other than their MI300 gpu is coming sometime late 2023.

Is NVDA going to meet, beat or fail its guidance come Wednesday?
 
  • 1Like
Reactions: 1 user

Palum

what Suineg set it to
27,181
42,825
Speaking of NVDA, Intel is laying off GPU/AI workers;



Back when TSMC reported their q2 earning they said they dont see "ai demand" significantly effecting their revenue this year.

AMD didnt report anything of note in regards to AI other than their MI300 gpu is coming sometime late 2023.

Is NVDA going to meet, beat or fail its guidance come Wednesday?

Based on my own experiences, AI is massively over blown in market expectations this early not necessarily because of the technical limitations per se but because of the inability to bring it to market based on compliance, risk and politics. All AIs must be trained to be "inaccurately correct" to adhere to certain political norms and guidelines, and further the enterprise compliance components are way underdeveloped. It's very exciting for R&D and dumb marketing types can translate that excitement into hype but just look at MS. They acquired significant leads with OpenAI licenses but all their autopilot projects are mostly still invite only. I've been twiddling my thumbs waiting for approval from their ethics board on a sandbox.

My gut tells me there is going to be massive overinvestment and the bubble will pop eventually as the legal and practical questions catch up with the R&D investment pace right now. It's one thing to have a "toy" like ChatGPT, but it's another to have public facing expert systems and all that. Most orgs that weren't ready to set up a rules based chatbot that used just NLP to link KB articles in 2022 are nowhere near ready to deal with more advanced chatbots with generative AI output today.
 
  • 6Like
Reactions: 5 users

OU Ariakas

Diet Dr. Pepper Enjoyer
<Silver Donator>
7,428
20,991
Based on my own experiences, AI is massively over blown in market expectations this early not necessarily because of the technical limitations per se but because of the inability to bring it to market based on compliance, risk and politics. All AIs must be trained to be "inaccurately correct" to adhere to certain political norms and guidelines, and further the enterprise compliance components are way underdeveloped. It's very exciting for R&D and dumb marketing types can translate that excitement into hype but just look at MS. They acquired significant leads with OpenAI licenses but all their autopilot projects are mostly still invite only. I've been twiddling my thumbs waiting for approval from their ethics board on a sandbox.

My gut tells me there is going to be massive overinvestment and the bubble will pop eventually as the legal and practical questions catch up with the R&D investment pace right now. It's one thing to have a "toy" like ChatGPT, but it's another to have public facing expert systems and all that. Most orgs that weren't ready to set up a rules based chatbot that used just NLP to link KB articles in 2022 are nowhere near ready to deal with more advanced chatbots with generative AI output today.

If this is anything like my entire career in tech sales then AI conversations are going to force the bottom 99% of companies to realize that they never adopted the last two trends, data mining and process automation. Then, they are going to start doing those things because they are the "cheap" ones compared to AI. If anyone thinks I am being too critical about how advanced most companies are, please look at simple things like SaaS adoption or even older trends like basic virtualization.
 
  • 3Like
Reactions: 2 users

Mist

REEEEeyore
<Gold Donor>
31,365
23,787
Based on my own experiences, AI is massively over blown in market expectations this early not necessarily because of the technical limitations per se but because of the inability to bring it to market based on compliance, risk and politics. All AIs must be trained to be "inaccurately correct" to adhere to certain political norms and guidelines, and further the enterprise compliance components are way underdeveloped. It's very exciting for R&D and dumb marketing types can translate that excitement into hype but just look at MS. They acquired significant leads with OpenAI licenses but all their autopilot projects are mostly still invite only. I've been twiddling my thumbs waiting for approval from their ethics board on a sandbox.

My gut tells me there is going to be massive overinvestment and the bubble will pop eventually as the legal and practical questions catch up with the R&D investment pace right now. It's one thing to have a "toy" like ChatGPT, but it's another to have public facing expert systems and all that. Most orgs that weren't ready to set up a rules based chatbot that used just NLP to link KB articles in 2022 are nowhere near ready to deal with more advanced chatbots with generative AI output today.
All this is saying is that AI models are all going to have to be retrained again and again and again until they get ones that are useful. Which means that GPU demand will remain extremely high if not continue to escalate. The transformer attention model has been proven to work, that the output scales pretty dramatically with the size and relevance of the dataset, the amount and quality of fine-tune layers applied on top, and the number of parameters in the model. The best way to build better AI models is to do more and more training runs with more and more GPUs.

Yes, at some point, one of these hardware wizards is likely to come up with a kernel, language and/or library that can improve efficiency by 10-100x, maybe more... but I don't see that causing people to stop training models.

Also you keep saying things like "All AIs must be trained to be "inaccurately correct" to adhere to certain political norms and guidelines" blah blah when you actually get this for effect for free at the RLHF layer. It is not something that requires additional coding or significant GPU cycles.

I do agree that the legal implications could possibly be massive. OpenAI may need to totally scrap their current model given the fact that it appears to be trained on large amounts of pirated and/or copywritten media. But that just means they will need to do MOAR training runs on a cleaner dataset.

Ultimately, I think we'll see enterprises spinning up GPU-heavy cloud instances and using the open-source models and fine-tuning or performing knowledgebase embedding using their existing internal KBs and data lakes. I'm a pretty lousy Python coder and with a few Youtube tutorials I've already been able to demonstrate internally how we can do this with our own internal Elastic data lakes.
 

Palum

what Suineg set it to
27,181
42,825
All this is saying is that AI models are all going to have to be retrained again and again and again until they get ones that are useful. Which means that GPU demand will remain extremely high if not continue to escalate. The transformer attention model has been proven to work, that the output scales pretty dramatically with the size and relevance of the dataset, the amount and quality of fine-tune layers applied on top, and the number of parameters in the model. The best way to build better AI models is to do more and more training runs with more and more GPUs.

Yes, at some point, one of these hardware wizards is likely to come up with a kernel, language and/or library that can improve efficiency by 10-100x, maybe more... but I don't see that causing people to stop training models.

Also you keep saying things like "All AIs must be trained to be "inaccurately correct" to adhere to certain political norms and guidelines" blah blah when you actually get this for effect for free at the RLHF layer. It is not something that requires additional coding or significant GPU cycles.

I do agree that the legal implications could possibly be massive. OpenAI may need to totally scrap their current model given the fact that it appears to be trained on large amounts of pirated and/or copywritten media. But that just means they will need to do MOAR training runs on a cleaner dataset.

Ultimately, I think we'll see enterprises spinning up GPU-heavy cloud instances and using the open-source models and fine-tuning or performing knowledgebase embedding using their existing internal KBs and data lakes. I'm a pretty lousy Python coder and with a few Youtube tutorials I've already been able to demonstrate internally how we can do this with our own internal Elastic data lakes.
The point is that at some point, there has to be a profit incentive when there's not 0% funny money all the time. The brakes will be pumped when revenue isn't coming. I don't know if that's now, but my perception is that it's going to be a lot easier to take existing models and make new products and the gold rush will quiet down to normal R&D, licensing and marketing mediocre products. That doesn't mean there's no upside to nVidia, just that I think this is still speculation by people who promised an unconstrained vision rather than reality.
 

Tide27

Bronze Baronet of the Realm
1,867
8,685
Was talking to some new friends we made the other day, and they bought their house earlier this year I think. I asked them if they didn't mind me asking how bad their interest rate was. It was like 6.5% or something.

A little over a year earlier and we locked in 2.75%. That'd make our monthly payment more than 50% more than it currently is. Insane.
I feel like some people figured they had no choice at this time. The nature of my job and trying to progress upwards meant, in theory, that renting was the best option.

However, the rent on the home we were living in went from $1300 to $2200 over the course of 2 years...and the entire area followed suit.

Housing prices started to get so outrageous that we couldn't even purchase a used home because they were selling for 50k + above ask, so we had to build. I think when we got the clear to build, interest was around 4%, by the time we were allowed to lock in the interest rate...it had jumped to 6% :(