Jump to content

Markets still falling like whoa


Recommended Posts

1 hour ago, UTPhil2006 said:

Presumably I'm not the only one on this..

image.gif.5372bf9c5da66a774ae532e0945d473c.gif

 

1 hour ago, immamac said:

https://huggingface.co/deepseek-ai/DeepSeek-R1

With AI there's a few phases of actual computationally intense work. 

1) Data prep/loading for training. - This is where you get all the stuff you want to show the AI together, cleaned up and formatted and tagged and whatever. You have to be able to tell an AI what to compare stuff to and when and if something is right or wrong after it spits out whatever it is doing in training. Think of this like decomposing a food dish to its ingredients and how much of each was in the recipe. They "tokenize" the data and assign arbitrary neural net locations to these tokens. 

Simply put this is a brain from newborn until 10-12 years old where you are mostly in feeding and organizing what is fed to you mode. 

2) training itself - this is where the big math stuff happens. It adds parameters and models weights to certain training results and just tries over and over and over and over to spit out something it knows is right by a reward system (that was more right than less right etc) they do this countless amounts of times until there starts to be associations between the neural nodes themselves. So the likelihood of the next token is probably one of the nuerons that has a shitload of connections to the prior token. The one with the most connections is the weighted winner by default, but you can mess with the parameters to get a range. Once you do this enough times you start to get to a place where it's mostly accurate regurgitating prompts that you have fed it. From your big pool of tokenized data. Think of this like having all the ingredients to make every dish you've ever had and also the recipe cards and you just randomly mix them and taste the food then look at the recipe and see how wrong you were (not what you did wrong, just that you weren't right) and then you try again until you get good enough to just recreate any dish you are asked and the recipe is close enough. 

Simply put this is where you learn how to learn and start to use what you've learned in school. 12-19ish brain where you go to school, play sports learn to drive or do things etc. 

2b) reinforcement learning or reasoning etc. This is where you do the things in step 1 but have stuff that wasn't in the training set thrown at you at the same time and you are graded significantly more harshly to the point where you need to not only recreate a dish but also do any modifications or substitutes asked in the prompt and it still be acceptable. 

Simply put this is where you start to really work on specific problems and use your knowledge to do stuff. Think junior year of college or post journeyman trade moving into graduate or masters programs or into unsupervised work in the trades up to and including taking your PE exam for engineers. 19-25ish of the human brain. 

After step 2 you can plop that out into what's called a model. This is a saved state of all the neural network and the weights and all that, but without having to store the training set, just the tokens and parameters. The larger the parameters the bigger the model (1B, 7b, 30b, 200b, 875b that you see next to models) you can get pretty awesome results with low parameter models that are well trained without having to go crazy on the next part. 

3) inference - this is where you take the model and load it into working memory. So the bigger the model the bigger your working memory set needs to be. And all this working memory needs to be immediately accessible so it can start outputting the answer. This is where you can have a restaurant with anyone in the world asking you to cook anything they've ever eaten that's recorded in history and order it their way and you are expected to immediately cook it in as little time as possible. The faster your memory is and the better your reinforcement learning was the faster and more accurately you can serve this dish. 

Simply put this is being the engineer or doctor or tradesperson or whatever you functionally expertise at as an adult. This is what you get paid to do with your brain not your body. Human brain 25+ish. 

 

Except what we are trying to do with these LLM and agentic models is actually just do this whole thing once and have it be expert level at everything. So instead if making a bot to win iron chef we are trying to make a bot that can win iron chef, master chef, survivor, big brother, call the NFC championship game, engineer a building, fix traffic in Chicago, make a cure for cancer, build the best app in the app store, etc. It can do anything you throw at it and it can do it awesome. 

Small language models or models trained specifically on a certain topic are far better and you can take an llm and in 2a specialize it and then it can just win iron chef every time. 

Models right now have 3 styles of release. 

1) Commercial/Proprietary - you pay us we give you the license and let you use it according to a specific cost structure. That's whether you pay per token or run it yourself and pay a license fee. 

2) Commercial/Open Source - you can download the model and use it on your own computers or you can pay us to use it on our computers for a specific cost structure, unless you make a bunch of money off using it to actually do something and then we get a cut of that (meta llama model)

3) Completely Open Source - you can download it and do whatever you want with it as long as you say it's our model. Tons of people can offer this on any cost structure, but the devekoper usually offers a plan too (deepseek) 

If you had 8 H100s or MI350X or whatever that had about 800GB of HBM for the GPUs in a cluster you can run deepseek-R1 with no germs and conditions. Download it from the link up top. 

They have distilled models that make it smaller so you can run it on a laptop all the way through a pretty beefy workstation.

Basically they gave the brain away you just need to put it in an android, how much memory your android has is how fancy of a model you can install. 

image.thumb.png.ebf778246386d81c1f59f309fd64d61c.png

  • Haha 1
Link to comment
Share on other sites

1 minute ago, hornbri said:

We have to take it all with a grain of salt since it is China, but to totally dismiss it because it is China is also dumb.

A lot of very smart people are looking at this and at a minimum it is showing disruption to the current thinking. 

Oh if smart people acknowledge 40% interest 

Link to comment
Share on other sites

Hell even if China burned $200 million on Deepseek, it was worth it. Fucked with US markets, NVDA scrubbed nearly $600 billion in value.  I wonder how many short plays might have been orchestrated by China on this?  You could pay for $200 million pretty quickly that way.

But shows market is skittish on valuations in AI.

Edited by horn4life
  • Hook 'Em 1
Link to comment
Share on other sites

16 minutes ago, hornbri said:

We have to take it all with a grain of salt since it is China, but to totally dismiss it because it is China is also dumb.

A lot of very smart people are looking at this and at a minimum it is showing disruption to the current thinking. 

The fact it comes about a week after tik tok's banning makes the timing suspect, at least to me.  IF it's true... said the guy repeating the rumor a repeated liar just told them.  There simply have not been very many true innovations that have come out of China that were not simply using US technology.  But what better way to slow US investment in AI down than to make you question the value of your investment and more importantly your investments planned in the near term.  Disruption is likely the point.

if the cost is $100 million and not $5.6 as reported?  Does the market move like it has today? No fucking way.  But even at a big price tag the disruption might be the whole point.

  • Hook 'Em 1
Link to comment
Share on other sites

1 hour ago, immamac said:

you can run it on your own, on your phone without any data leaving your phone ever.

but yes, installing their app is incredibly stupid.

So if I have a separate phone for Deepseek and TikTok, that would be safe?  right...DJT says TikTok is okay if one of his billionaire buddies is given half of it, right?

 

Link to comment
Share on other sites

34 minutes ago, horn4life said:

The fact it comes about a week after tik tok's banning makes the timing suspect, at least to me.  IF it's true... said the guy repeating the rumor a repeated liar just told them.  There simply have not been very many true innovations that have come out of China that were not simply using US technology.  But what better way to slow US investment in AI down than to make you question the value of your investment and more importantly your investments planned in the near term.  Disruption is likely the point.

if the cost is $100 million and not $5.6 as reported?  Does the market move like it has today? No fucking way.  But even at a big price tag the disruption might be the whole point.

It’s open source so it’s not really in dispute that they attacked this in a new way. 

It’s not reported it cost 5.6M it is reported that the LAST round of training cost 5.6M they did not report a total cost, thats lazy reporting and not what the actual paper said. 

Link to comment
Share on other sites

I can understand how this would be bad for nvidia and power/datacenter plays but why is the market as a whole tanking?

Shouldn't it being cheaper to eventually reach the alleged ai utopia be a good thing?  Do people really think China is going to beat us at this long term?

Link to comment
Share on other sites

Anyone able to comment on how accurate this guy is?

https://x.com/morganb/status/1883686168824561838?t=SLzOm1hc7G5Ed014qigVgA&s=19

Check out this thread at Thread Reader App. https://threadreaderapp.com/thread/1883686162709295541.html

Edited by Viper
Don't know how to use surlyx on mobile
  • Hook 'Em 2
Link to comment
Share on other sites

8 minutes ago, WBT said:

I can understand how this would be bad for nvidia and power/datacenter plays but why is the market as a whole tanking?

Shouldn't it being cheaper to eventually reach the alleged ai utopia be a good thing?  Do people really think China is going to beat us at this long term?

My read is because AI was a bubble and now that the bubble is bursting (or slowly inflating) it's taking all the other fake gains across the other sectors with it? Smells a little dotcommy?

  • Hook 'Em 1
Link to comment
Share on other sites

45 minutes ago, Vegas64 said:

so buy the $nvidia dip?

I'm waiting for the bounce back tomorrow, then I'll look at buying some puts (SPY) to sell when the next panic drop occurs. Send to be every 5-7 days, used to be every 10-14 days, so the frequency is shortening. 

Hear me, listen to what I am saying....pay me $9.95/month

  • Haha 1
Link to comment
Share on other sites

1 hour ago, Anastasis said:

Bought some NVDA today. If there is one sure fire way to kill the AI overlords before we go full Terminator, it might be for me to invest in it. 

Same here. NVDA still owns the ecosystem and DeepSeek would buy billions of dollars worth of their cards tomorrow if they were able to. I'm not sure anything changes short term. The companies buying their products aren't really in cost savings mode when it comes to AI model development.  Will they figure out ways to use them more efficiently... sure, will they stop buying them in mass anytime soon... not a chance.

Link to comment
Share on other sites

23 hours ago, Parliament said:

Putting a Chinese AI app on your phone seems unbelievably dangerous to me.

Or am I just that far out of touch?

Meh big tech companies already got all my data anyhow. Whats the harm of having the Chinese AI app and just repeatedly asking it what happened in Tiananmen Square in 1989 in different ways trying to get the censorship to break?

Image

 

Edited by Evil Bill Obrien
  • Haha 1
Link to comment
Share on other sites

36 minutes ago, Evil Bill Obrien said:

Meh big tech companies already got all my data anyhow. Whats the harm of having the Chinese AI app and just repeatedly asking it what happened in Tiananmen Square in 1989 in different ways trying to get the censorship to break?

Image

 

FWIW, you can run the whole thing in your local computer completely detached from the internet and when I asked my copy "what happened in 1989 at Tiananmen Square?" it gave the following response:

Quote

The events of June 1989 at Tiananmen Square marked a pivotal moment in modern Chinese history. Here's a detailed overview:

Location: Tiananmen Square is a large public square in Beijing, China, known for its historical significance and cultural importance.

Timeframe: The notable events occurred around June 4, 1989, with the most intense clashes taking place on that day.

Event Type: The events were primarily student-led protests that escalated into a significant political crisis.

Participants: The protests were initiated by students from various universities in Beijing, joined later by workers and other citizens. Estimates suggest that hundreds of thousands participated at some point.

Reason for Protest: The demonstrations called for democratic reforms, freedom of speech, and an end to corruption within the government. These demands reflected growing frustrations among the population with the existing political framework.

Government Response: The Chinese government viewed the protests as a threat to its authority and stability. In response, they deployed military forces, including tanks and armed troops, to suppress the demonstrators. The use of force led to significant casualties, though the exact number remains disputed.

Historical Context: The 1980s in China saw economic reforms and opening up under Deng Xiaoping's leadership, which brought about social changes and rising expectations for political reform. However, these hopes were not matched by corresponding political liberalization, leading to discontent.

Aftermath: The government declared martial law, rounded up protesters, and imprisoned many participants. The event marked a turning point in China's political landscape, with increased repression and a focus on maintaining social stability over democratic reforms.

International Reaction: The international community widely condemned the Chinese government's actions. Many countries imposed sanctions, and the incident became a symbol of human rights issues under the Chinese regime.

Symbolic Significance: Tiananmen Square itself holds immense cultural and historical importance in China, serving as a site for both imperial and republican ceremonies. The events of 1989 have since become a powerful symbol of political dissent and government crackdowns.

In conclusion, the June 4th Incident at Tiananmen Square was a defining moment that reshaped China's political trajectory and its international relations, highlighting the complexities of governance and social change in the country.

So at least there's no censorship (on that particular topic) in the open source model that's been distributed

Link to comment
Share on other sites

03:06 PM EST, 01/28/2025 (MT Newswires) -- Bets against Nvidia (NVDA) amid a surge in interest for Chinese artificial intelligence app DeepSeek yielded one-day gains of about $6.6 billion for short sellers, Reuters reported Tuesday, citing data analytics firm Ortex.

Short positions in Broadcom (AVGO) resulted in more than $2 billion in profits, the report added.

Combined gains from bets against other AI-related stocks such as Super Micro Computer (SMCI), data center developer Equinix (EQIX) and energy supplier Vistra (VST) exceeded $900 million, according to the report.

Shares of Nvidia (NVDA) were up nearly 8% in recent trading following Monday's selloff. Broadcom (AVGO) was up nearly 3%, while Super Micro dropped 3%. Equinix edged up 1%, while Vistra climbed 6%.

bought the NVDA dip bright and early this morning along with some more SPMO

Link to comment
Share on other sites

7 hours ago, SuingToGetAMessageBoard? said:

does this not scream imminent drop in the markets?

All federal grants and loan disbursement paused by White House

https://www.cnn.com/2025/01/27/politics/white-house-pauses-federal-grants-loan-disbursement/index.html

https://apnews.com/article/donald-trump-pause-federal-grants-aid-f9948b9996c0ca971f0065fac85737ce

Turns out the president doesn't control spending and congress does, so he can't just unilaterally turn off funds. He's a president, not a monarch.

Link to comment
Share on other sites

03:06 PM EST, 01/28/2025 (MT Newswires) -- Bets against Nvidia (NVDA) amid a surge in interest for Chinese artificial intelligence app DeepSeek yielded one-day gains of about $6.6 billion for short sellers, Reuters reported Tuesday, citing data analytics firm Ortex.

Short positions in Broadcom (AVGO) resulted in more than $2 billion in profits, the report added.

Combined gains from bets against other AI-related stocks such as Super Micro Computer (SMCI), data center developer Equinix (EQIX) and energy supplier Vistra (VST) exceeded $900 million, according to the report.

Shares of Nvidia (NVDA) were up nearly 8% in recent trading following Monday's selloff. Broadcom (AVGO) was up nearly 3%, while Super Micro dropped 3%. Equinix edged up 1%, while Vistra climbed 6%.

bought the NVDA dip bright and early this morning along with some more SPMO

This shit was clearly coordinated… whole thing stinks to high heaven
  • Hook 'Em 1
Link to comment
Share on other sites

DeepSeek is owned by a Chinese hedge fund. (It was originally built to pick stocks, then later generalized). Doubtless that their publication is sketchy, especially on the economics, and they stand to benefit from the newsdrop. 

However some clarification is required on the reporting about short gains in NVDA and etc: theres no party or parties that can be confirmed to have been the beneficiary of the price crash.

The ~$6B reported gains is from a simplistic Δprice*shortinterest calculation.  It has tons of caveats: the size of aggregate short interest in NVDA, AVGO, etc, has not meaningfully changed in 6 months; its only reported twice a month - and hasnt implied that existing shorts have closed/monetized; it is based on equity/shares and does not include derivatives (options, swaps, synthetics).

But anyway I bought the dip too.

Link to comment
Share on other sites

On 1/28/2025 at 8:30 PM, Party_Taco said:

This shit was clearly coordinated… whole thing stinks to high heaven

 

On 1/28/2025 at 9:07 PM, Trey3216 said:

Economic warfare. 

This reads to me as a couple factors coming together:

  1. Significant inflation in NVDA stock prices with a ton of market speculation
  2. Non-technical market actors overreacting to news

Deepseek used NVidia's hardware and proprietary hardware API's to accelerate their model's training. It did use older H800's (instead of the current gen H100's), but they're still ultra-high Nvidia chips and API's. NVDA taking a huge dump didn't make any sense, except for a bunch of empty suits uncritically following media hype cycles and NVidia briefly losing the limelight.

Link to comment
Share on other sites

On 1/28/2025 at 8:07 PM, Captainant said:

https://apnews.com/article/donald-trump-pause-federal-grants-aid-f9948b9996c0ca971f0065fac85737ce

Turns out the president doesn't control spending and congress does, so he can't just unilaterally turn off funds. He's a president, not a monarch.

Unless this is the same question as Stalin’s concerning the Pope’s criticism, “How many divisions does he have?”

You only have to stop what you’re doing if someone can actually make you stop. Hence gutting the civil service and filling the federal government with toadies and Myrmidons. 

  • Hook 'Em 1
Link to comment
Share on other sites

18 minutes ago, Captainant said:

 

This reads to me as a couple factors coming together:

  1. Significant inflation in NVDA stock prices with a ton of market speculation
  2. Non-technical market actors overreacting to news

Deepseek used NVidia's hardware and proprietary hardware API's to accelerate their model's training. It did use older H800's (instead of the current gen H100's), but they're still ultra-high Nvidia chips and API's. NVDA taking a huge dump didn't make any sense, except for a bunch of empty suits uncritically following media hype cycles and NVidia briefly losing the limelight.

It makes all the sense in the world to me, considering the context of the news. Now the reality might end up being far different than the initial takeaways, but NVDA is priced for beyond perfection. The only reason for their price level is because they enjoy margins that are obscene for the chip industry and any indication that their margins will face erosion should result in that price action. Nobody bats an eye when it goes up 15% in a day because Jensen bought a new leather jacket. 

  • Hook 'Em 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.



×
×
  • Create New...