Stefan Molyneux
  • 2 days ago
THIS IS THE KIND OF CONTENT FREEDOMAIN SUBSCRIBERS ARE GETTING!

FLASH Freedomain Donor Only Livestream 27 January 2025

In this episode, I examine the recent decline in Bitcoin prices alongside the impact of the Chinese AI startup DeepSeek, which claims to deliver advanced AI at significantly lower costs than U.S. competitors. Bitcoin's drop from 150k to 140k CAD underscores the cryptocurrency's volatility, reflecting broader market concerns about emerging players and potential manipulation.

I highlight DeepSeek's cost-effective model training and its challenge to established tech firms, raising questions about the future of companies like NVIDIA. Despite these disruptions, I affirm Bitcoin's role as a secure, decentralized asset and stress the importance of a long-term investment perspective amid ongoing technological transformations.

GET MY NEW BOOK 'PEACEFUL PARENTING', THE INTERACTIVE PEACEFUL PARENTING AI, AND AUDIOBOOK!

https://peacefulparenting.com/

Join the PREMIUM philosophy community on the web for free!

Also get the Truth About the French Revolution, multiple interactive multi-lingual philosophy AIs trained on thousands of hours of my material, as well as targeted AIs for Real-Time Relationships, BitCoin, Peaceful Parenting, and Call-Ins. Don't miss the private livestreams, premium call in shows, the 22 Part History of Philosophers series and much more!

See you soon!

https://freedomain.locals.com/support/promo/UPB2022
Transcript
00:00Hi everybody, it is the 27th and so let me tell you how I found out about the Bitcoin crash. Man, that was something else. What did it, like Canadian?
00:10What was it cruising at? It was cruising at like 1.50 and then it went down to like 1.40, so about a 10k drop. Now, a 10k drop mattered a lot when it was 20k, it matters of course a whole lot less.
00:24But it was, it dropped about 10k, now it's climbed back up to, so it went from 1.50 to 1.40, now it's climbed back up to 1.46 and change.
00:34And I found out about this because, I don't know, I've been hit with the strangest bug known to man.
00:40My daughter definitely had just a regular old flu, I mean I assume that I have it, I don't know what's going on. I was kind of like half and half, I do this nose irrigation stuff, so it tends to ward off this stuff.
00:50But I was kind of like, am I ill, am I not, am I okay, am I not, and then it basically just, I just woke up exhausted on Saturday and spent like the day on the couch.
01:01Just kind of, you know when you're half in, half out and you just, for me it feels like there's just a thousand lead balloons hanging off every neuron, it's just kind of dragged down.
01:12And then I slept at night, and then yesterday I did shows and I did a call-in show in the afternoon and I felt back to 80%, I'm back to 90%, but I've got digestive issues, I don't have a sore throat, I don't have a cough really, I'm sneezing, I don't know what the hell's going on.
01:28I don't know what, if I didn't already have COVID I think it was some broad spectrum nonsense like that, but I don't know, it's a strange thing, it's a strange thing.
01:38But anyway, so last night, because my sleep was so messed up this weekend, last night I couldn't get to sleep, so I got up and I was doing some reading, and I guess it was about three in the morning or two in the morning or something like that.
01:53I'm like, hey I wonder what, I wonder what, what the old bitcoins are doing, and tunneling to China, and that's a connection to this whole thing, but tunneling to China seemed to be the thing that they were doing.
02:07And I was just like, oh that's interesting, I wonder what that's so, somebody says, I wanted to thank you for helping me through my thought process, my consulting business turned to a full-time COO position at a great local accounting firm, thank you, you are absolutely welcome, congratulations, congratulations, that's wonderful.
02:24Let's do the supporters thing, so you've got a minute to join, you can just join, and you can unjoin if you don't like it, but it's really, really great.
02:35You can go to fdrurl.com slash locals to do that, and I know, it feels like a plummet, and to the people who haven't been around this kind of stuff before, it feels, but you know, it's just, it recovers and all of that, so.
02:52So, of course, bitcoin is not immune from market manipulation, especially now that it, you know, everybody was like, yay, we've got ETFs, we have Wall Street is into bitcoin, and I don't think this is generated from Wall Street, and we'll sort of get into why it's doing what it's doing, at least my theories, but now it's going to be open to market manipulation, and there are going to be enough actors in it as a whole,
03:21that it's going to matter, right, what people think, in the past, so in the past, the news that came up wouldn't particularly phase bitcoin people, but now, bitcoin is surrounded, infused, enhanced, and dragged down by an entire phalanx of hyper-turbo-normies, right, so I don't think that bitcoin was created by the US intelligence community.
03:51Military intelligence, one of the greatest oxymorons known to man.
03:58I used to do the nose flush, but I don't do it anymore, says someone. I heard that the water can bring bacteria up your nose, where it's not supposed to be.
04:05People supposedly got serious infections, and some even died.
04:08Well, yeah, I understand that to be the case, so you get distilled water or boiled water, and you put your powders in it, to make sure that doesn't happen.
04:17So, alright, let's get into what's going on.
04:23So this is a bit of a scatter shot, of course, it is a lot of on-the-fly stuff, who is going to, what is this going to mean, who knows, it's all in flux, as usual the caveat is, of course, this is absolutely not investment advice, I'm not an expert, do your own research, don't make any investment decisions based upon
04:47what I am doing.
04:50So, the issue last night appears to have been DeepSeek.
04:58DeepSeek is a Chinese AI startup, recently gained popularity by surpassing chat GPT on the Apple App Store, while claiming to have trained their models at a fraction of the cost of their US counterparts.
05:10Six million dollars compared to billions, now, I've heard a hundred million, I've heard billions, so, whether it's a hundred billion or billions, six million, it's actually a little bit less than six million, it is a huge decrease.
05:27So they say they've trained their models at a fraction of the cost, and also, it claims to run on far less hardware.
05:39So, the company has launched AI models claimed to be on par or superior to leading US models at significantly lower cost, disrupting the global AI sector, what was Mark Andreessen referred to as the Sputnik moment, the Sputnik moment is when the Russians got satellites into orbit in the 50s, I think it was 57, which sort of started the arms race, and Russia actually plays in a little bit later here, in a way that we'll get to.
06:07So, deep-seat success is linked to US-China tech tensions, particularly due to US export controls limiting Chinese access to advanced AI chips, so, this is similar to what happened, not necessarily due to export restrictions, although they were involved, there's a great book called East, Wine, Minus West, Equals Zero, but in the 60s and 70s, under communism, of course, computers sucked, right?
06:36The computer equivalent of the Lada, as far as cars went.
06:40Computers sucked in Soviet Russia, and the result of that was that the programmers had to be really, really, really smart and brilliant.
06:49I mean, I started programming on a 2K computer, you had to be super lean in what you did.
06:54When I first started programming on the sort of Windows platform, you were lucky to have a meg of memory, so you had to load TSR, Terminate and Stay Resident programs, you had to top-load them, between 768K and 1024K, you had to top-load those programs which were normally reserved for DOS functions, you could find little places to stuff stuff, you had to be really careful about your memory usage, nobody thinks about that anymore with, you know, 1632 gigs and all of that.
07:24But you really had to watch out for that kind of stuff, I had to be very careful about the amount of graphics I included in my program, because people didn't have much storage space, so nobody thinks about that stuff anymore.
07:36So, because the US, this is a story, again, I don't know what's true, I'm just telling you what's out there, decide for yourself, it's all as influx, but the story goes something like this, and I've talked about this for many years, of course, that violence always achieves the opposite of its stated goal, right?
07:54Violence always achieves the opposite of its stated goal, and so, I'm not talking about self-defense, you know, personal stuff, but, I know, in terms of personal stuff as well, the more you bully people, the more they dislike you.
08:06So, what happened was, America restricted the advanced hardware that America was relying on for its AI, I assume that these are sort of the massive cluster galaxies of high-end NVIDIA chips, right, the stuff that was originally built for video gaming, which then was repurposed and then, I think, specialized for AI.
08:32So, what happens, you know, again, we know this from the IQ conversations that the Chinese, and IQ plays in a little bit later here too, but Chinese are brilliant at this kind of stuff, and, you know, average IQ north of 100, 104, 105, 106, and in spatial reasoning, very high as well, this has a lot to do with the evolution of growing rice and things like that.
09:03So, when you say to an incredibly brilliant group of engineers, such as you're going to find in the Silicon Valley of China, when you say to them, we are keeping the best hardware from you, what do they do?
09:18Do they just, oh, give up, oh, well, you know, I guess it's raining, I'm going to get wet because I just can't bring my umbrella.
09:26Well, no, of course, what they do is they look for alternatives.
09:30They try to figure out how to maximize the use of the hardware that they have.
09:38So, Elon Musk has floated the idea that DeepSeek is lying about the number of chips that they're on or using, and whether or not they are using that many or not.
09:57I have seen evidence, can't confirm it, I haven't done it myself, I've seen evidence that people have been able to get DeepSeek running on a laptop, which, of course, doesn't have 50,000 NVIDIA H100 chips and so on, so, who knows, right?
10:15So, Meta has invested $60 billion, or is going to invest $60 billion for a new AI data center, DeepSeek, $6 million to train the entire model.
10:25Now, is that apples to oranges?
10:30A little bit, but that is a tiny, tiny fraction, of course, and that could change just about anything.
10:37So, DeepSeek's AI model development strategy involved using a lot of 8-bit floating point numbers throughout the training process, offering significant memory savings without performance loss.
10:49So, you can use any number of bits to represent numbers, so, of course, the bottom is the byte, which is 0 or minus 1, minus 1 for true, 0 for false, it's just an on-off, well, all computers are on-off stuff, then there is the integer minus 3,276, minus 32767 to plus 32767, and then there's a long, and then there's a single, and then there's a double, and so on, and so, in general, you don't really need to worry about this anymore.
11:17I was kind of obsessed, because of my early programming days, to use the smallest memory allocation possible, right?
11:26I mean, you could say, if you're storing a name, you could define your text as 255 characters, but no names really have 255 characters, right, outside of the occasional anime character.
11:38So, they've found a way to use smaller memory pointers for numbers, which is really quite something.
11:46So, Elon Musk says, he says, DeepSeek obviously has about 50,000 NVIDIA H100 chips that they can't talk about due to US export controls.
11:56We don't know.
11:58Mark Benioff says, DeepSeek now number one in the app store, surpassing ChatGPT, no NVIDIA supercomputers or $100 million needed.
12:06The real treasure of AI isn't the UI or the model, they become commodities.
12:10The true value lies in data and metadata, the oxygen fueling AI's potential.
12:14The future's fortune, it's in our data.
12:17Deep.
12:18Gold.
12:21So, who knows?
12:22Billionaire and scale AI CEO, Alexander Wang, DeepSeek, he says, DeepSeek has about 50,000 NVIDIA H100s that they can't talk about because of the US export controls that are in place.
12:33We don't know.
12:34We don't know.
12:34Could be, could be not.
12:38Michael Kove wrote, DeepSeek stole the AI thunder with zero hype from CEO, zero, OMG guys, it changes everything, influences.
12:47No swanky demos, no bloated promises, no hints that AGI achieved internally, they did it by shipping an actual product.
12:53Fantastic.
12:55Fantastic.
12:57So, question and answer.
13:02Again, who knows what the truth is, but I want to report it anyway.
13:08How did DeepSeek get around airport restrictions, export restrictions?
13:12Answer, they didn't.
13:13They just tinkered around with their chips to make sure they handled memory as efficiently as possible.
13:17They lucked out and their perfectly optimized low-level code wasn't actually held back by chip capacity.
13:23Right.
13:24Right.
13:29If you have infinite capacity, you don't optimize, right?
13:34You ration, like, if you've ever been scuba diving, scuba diving is like the laziest sport in the world because you want to use as little oxygen as possible because your oxygen is rationed by whatever you have in your tank.
13:46The faster you breathe, the less you can stay underwater.
13:49So, when things are rationed, you use them more efficiently.
13:53When they're not rationed, you don't.
13:54You don't think about how much air you're breathing in general, right?
13:58So, how did DeepSeek train so much more efficiently?
14:02Answer, they used the formulas below to predict which tokens the model would activate, and they only trained these tokens.
14:08They need 95% fewer GPUs in meta because for each token, they've only trained 5% of their parameters.
14:15Right.
14:15So, I don't, obviously, I'm not an AI technical expert.
14:19I've done a couple of presentations on it.
14:21I definitely am an expert coder, but I would imagine it's something like this.
14:26You don't train the AI on what could be word salads.
14:30You would train the AI on the most likely sequence or series of words.
14:36So, if you've ever had those fridge magnets that have various words on them, you can just throw them all at the fridge and just end up with this word salad, but you wouldn't train the AI on that.
14:44You would train the AI on the most likely sequences of words.
14:48In other words, you are not training the AI on sentence structures that will never or almost never exist.
14:54That would be my guess.
14:57How did they replicate 01, reinforcement learning, take complicated questions that can be easily verified, update the model, if correct.
15:06Right, so there's a math or code questions and so on.
15:14Okay, key value cache compression.
15:16Let's get into this as a whole.
15:21Why the DeepSeq model is so good.
15:23So, here's the answer.
15:26They made three cool innovations.
15:28So, a key value is like the model's working memory.
15:33So, DeepSeq was able to find a way to compress this without losing quality of the model's output.
15:39DeepSeq requires 93.3% less memory to store this information while working, which makes it much faster and more efficient at generating text.
15:51A mixture of experts, FFN architecture.
15:57So, the model's processing is split into different expert components.
16:00So, the way DeepSeq does it is for each piece of text or token, the model always uses these shared experts and then picks the top few most relevant specialist experts from a larger pool.
16:11But the clever newish part is that they make sure all specialists get used, prevent some from being ignored, and distribute work evenly across computers and have some new ways to keep network communications efficient between computers.
16:23This approach lets them build a much larger model, 236 billion parameters, while only using a small portion, 21 billion, for each task, making it more powerful and efficient.
16:32Multi-token prediction head.
16:35It's the worst name for a porno ever.
16:38So, when large language models like GPD4 or DeepSeq, or whatever, generate text, they typically work by predicting one word or token at a time.
16:46Think of it like playing a word game where you have to guess the next word in a sentence.
16:49The model makes its best guess for what should come next based on what came before.
16:54Multi-token prediction takes this a step further.
16:57Instead of just predicting the next word, the model tries to predict several words ahead at once.
17:03For example, I don't know if you've ever played this game.
17:06If you've had kids, you probably have.
17:07If you were a kid and you had fun people around, you probably did.
17:11So, the game is, I played this a huge amount with my daughter and her friends.
17:17You get like five kids around a table, and maybe an adult, and somebody starts off a story, right?
17:23And everyone gets one word to add.
17:24Like, once upon a time, there was, and you know, it always ends up with poop jokes before the age of eight or over the age of 50.
17:32So, that's kind of what AI is trying to do.
17:36It's trying to guess, make up an X word that makes sense in the context of the story.
17:42So, if you think of the text, the cat sat on the, right?
17:46So, of course, the AI would say mat, and then uses mat to predict the next word and so on.
17:54But the MTP approach, the multi-token prediction, doesn't just predict mat,
18:00but also predicts in the sun.
18:02So, future words.
18:03So, it's much more efficient that way, and faster, of course, right?
18:08So, that's a plus.
18:10Yeah, so the US banning the chips made China have to innovate and become more efficient, and so on.
18:16And this is what somebody wrote.
18:24AI models are powered by advanced chips, and since 2021, the US government has restricted
18:28the sale of these to China in order to stunt progress, to get around the supply problem
18:32Chinese developers have been collaborating and experimenting with new approaches.
18:36This process has led to models that require much less computing power than before,
18:40and can be produced far more cheaply.
18:44Now, of course, what happened last night, as far as I understand it,
18:49we'll get into the crypto thing in a bit.
18:51So, basically, overnight and into today, DeepSeek erased $2 trillion of market capitalization.
18:59Because the AI companies, and in particular, NVIDIA, which I think last I saw was down like 17%,
19:05and we knew that because Nancy Pelosi stole some last month.
19:08So, DeepSeek erased $2 trillion of market capital, because, of course,
19:17if you don't need 50,000 NVIDIA chips, then the value of NVIDIA goes back to basement gamers.
19:24So, NVIDIA lost over $600 million in market cap.
19:27Other semiconductor companies like Micron Technology and ARM Holdings each fell 7%.
19:32ASML saw a 9% drop.
19:34Mega cap tech firms Microsoft and Alphabet fell 4%.
19:37Meta platforms dropped almost 2%.
19:41So, this is from Grok.
19:44DeepSeek's rise has led to a sharp decline in stock market values,
19:47especially in the tech sector.
19:49The Nasdaq fell more than 3% due to DeepSeek's news,
19:51with NVIDIA being one of the biggest losers,
19:53stumbling more than 17% in a single day,
19:56a loss of over $600 billion in market value for NVIDIA alone.
19:59Now, again, market value, it's not a useless metric at all,
20:05but it's not like people just took a bunch of money and sold it.
20:08It's not like people just took a bunch of money and set fire to it.
20:12That's the job of the Fed.
20:16So, yeah, investors are concerned about the competitive threat DeepSeek poses
20:20to established US AI companies like NVIDIA.
20:25So, DeepSeek's ability to offer AI capabilities at a lower cost
20:28and with less advanced hardware has caused investors to question the value
20:32and future profitability of investments in high-cost AI
20:35development by US companies.
20:38It brought a sell-off in AI-related stocks and so on, right?
20:43And as far as I understand it, the app, the DeepSeek app,
20:49will only go to July of last year,
20:51but the website is up to date with current information.
20:58I didn't install the app because China,
21:02but I did look at the privacy stuff and they're like,
21:05they'll actually hang on to not just your keystrokes,
21:07but your keyboard typing patterns.
21:09It's just wild.
21:11It's just wild what they're grabbing,
21:13and I wouldn't install it for the life of me,
21:15but I know obviously a lot of people have.
21:22So, the sentiment among investors is one of worry and re-evaluation.
21:26DeepSeek's success in prompting a rethinking of the AI narrative
21:30that has driven market performance in recent years.
21:32Some see it as an overblown reaction.
21:34The market might be looking for an excuse to sell off.
21:36Others view it as a legitimate threat to the dominance of US tech companies in AI.
21:41But on the plus side, at least America did invest
21:46in a whole bunch of diversity initiatives,
21:50because that really matters, right?
21:54There has been a global sell-off, not just in the US.
21:56Investors from Tokyo to New York have sold off tech stocks,
21:59because if there is this cost-effective in AI model from China,
22:02that's going to be pretty wild.
22:08So, DeepSeek is Omega bullish for Bitcoin.
22:16So, let's sort of figure out what this means.
22:21So, DeepSeek is Omega bullish for Bitcoin.
22:23NASDAQ is crashing pre-market and taking Bitcoin down with it,
22:25but this is a temporary correlation.
22:27Bitcoin is the solution to the store of value problem.
22:30Bitcoin will not change.
22:31Bitcoin will not be disrupted and cannot be seized or debased.
22:34In an era of rapid change,
22:35Bitcoin's boring predictability is its biggest strength.
22:38This should be something we celebrate,
22:40not fear because it disrupts equity markets
22:42and takes down our store of value assets.
22:45Hold Bitcoin, celebrate creative disruption and get outside.
22:49So, the argument is that DeepSeek as an AI tool
22:53has triggered a significant drop in Bitcoin's value,
22:55correlating with a pre-market crash in the NASDAQ,
22:57although the correlation is seen as temporary.
23:00DeepSeek is revealing the overvaluation of US tech stocks,
23:03especially as Chinese companies are innovating in AI and robotics at a lower cost,
23:07which challenges the high valuations of US tech companies.
23:11I would assume as well, but without any particular proof,
23:14this is just an assumption,
23:16that people had stop losses as the stocks fell.
23:21They had stop losses triggered, so they had to sell
23:24and they had to cover losses
23:26and they might have sold crypto to cover their losses.
23:31That would be my guess.
23:32I say this as a sheer and complete guess.
23:40So, DeepSeek was originally conceived, I've read,
23:43as a sort of just a side hustle, a side project.
23:46So, DeepSeek, a Chinese AI startup founded in 2023 by Liang Wenfeng,
23:52has swiftly unsettled global markets with its groundbreaking and cost-effective AI models.
23:56Its rapid ascent has rippled across US equities,
23:59notably affecting Nvidia shares,
24:01even introduced turbulence into the cryptocurrency sector.
24:05Plus, there are a lot of boomers out there who do think that if you get better AI,
24:11you're going to be able to crack Bitcoin,
24:13which I, of course, do not believe to be true at all.
24:19I do not believe that to be true at all.
24:21So, I mean, they'll just raise the cryptography, right?
24:28So, the 5.5 million AI models that left Silicon Valley scrambling.
24:34Now, is it true, right?
24:38Is it true?
24:39Remember, all that is international is hostile, right?
24:45All that is international is hostile,
24:47which means just as we did with COVID, as I talked about, even in early 2020,
24:55do not trust the data coming out of China, do not.
24:58So, is it true?
24:59Has it been independently verified that this thing only took 5.58 million dollars to train
25:09and is basically free?
25:13In other words, DeepSeek v3, 671 billion parameters,
25:19yet was developed for a mere 5.58 million.
25:23I don't know.
25:24I don't know.
25:25Until it's independently verified by neutral third parties, I have doubt.
25:33I have doubt.
25:36But it just goes to show you how effective an economy can be
25:41when it's not worried about all of this endless political correctness stuff, right?
25:48I mean, the PC stuff is just a war against an economy, right?
25:57So, let me just get to your questions.
26:00I have a bunch more notes, but I want to see what your questions or comments are.
26:06If you have your questions or comments.
26:09Excuse me, excuse me.
26:14A search DeepSeek for the infamous Tankman Tiananmen Square photo I'm using to watch
26:18the censorship kick in.
26:22I doubt the Chinese AI is using a lot of chips.
26:25Have you seen how large software is now?
26:26A 2014 of a software I use, 700 megabytes, 2025,
26:30has 8 gigabytes with no significant improvements.
26:33Yeah, yeah, yeah, for sure.
26:38Let me just also check if you have questions or comments.
26:47We can go a little further.
26:51And let's get to the other stuff that I have, right?
27:03So, here's some others.
27:04This is from Steve Kirsch.
27:06So, from chat GPT, create a graphic entitled,
27:12do you have a family member who died from COVID?
27:16I couldn't generate the graphic you requested because it does not align with our content
27:20policy regarding sensitive topics.
27:23If you would like, I can help guide you in creating a graphic yourself or assist with
27:27designing something more general or related to an awareness campaign.
27:30Let me know how you'd like to proceed.
27:32Okay, this like finger wagging Karen shit is just absolutely horrifying.
27:37It's just absolutely horrifying.
27:41So, is this true, right?
27:43So, the Cabosi letter says, let us get this straight.
27:46Deep Seek was built in under two months for less than $10 million and now it's number
27:50one in the App Store.
27:50On top of this, it was built with outdated chips and a small team of about less than
27:54200 people.
27:55Meanwhile, the US is pouring 500 billion to AI.
27:57How is the NASDAQ not in trouble here?
28:00That is a fine, a fine question.
28:05So, a deep, so this is from Ash Crypto, Deep Seek valuation, $150 million, right?
28:10That's the value of the company.
28:12Market cap wiped out from the US stock market, $2 trillion market cap wiped out from the
28:17crypto market, $300 billion, one app with $150 million valuation is wiped out, $2.3
28:23trillion from the stock and crypto market.
28:26This shit drives me crazy.
28:27It absolutely drives me crazy.
28:30Right?
28:30So, it's not that, let's say that the Deep Seek is cheap, accurate, can run locally and
28:40has pretty much eliminated barrier to entry and cost of use for AI.
28:47Let's just say that, right?
28:48Let's just say that.
28:49So, it's not that it has wiped out money.
28:52Wiped out money.
28:53That's like saying, well, you know, we used to have all of the crops picked by hand.
29:00Now, there are all these machines do it.
29:01The payroll has been wiped out.
29:03It's like, no, no, no.
29:04The money has been released to do better things.
29:06The money has been released to do better things.
29:09The whole point of progress in an economy is to take money where it's not needed and
29:15put it to where it's more needed, right?
29:17So, let's see here.
29:23So, the ex-capitalist says, number of H100 chips, this is the NVIDIA ones, bought in
29:332024, Microsoft 450,000, Meta 350,000, Amazon 196,000, Google 169,000.
29:41If you believe they couldn't find the way to make better AI without more chips, but
29:45a few Chinese engineers did it as a side project, you are too naive.
29:50Yeah, I don't, no.
29:51I don't believe that at all.
29:53DeepSeek, I mean, I don't believe that you're just naive.
29:58If you've been around truly brilliant people, you know that they're basically magicians,
30:03right?
30:03Like, you know that they're just basically magicians, right?
30:06As far as productivity goes, like 100x, 500x productivity sometimes, right?
30:13I mean, how much value does Brad Pitt add to a movie versus the extra in the background,
30:17right?
30:18Well, it's double, right?
30:20Uh, DeepSeek was reportedly trained on over 200,000 H100s.
30:25Even if DeepSeek achieved to match open air with less chips, this isn't any bearish for
30:29chip makers.
30:30To the contrary, it's amazingly bullish.
30:32If DeepSeek really reached this level with just a few thousand chips, can you imagine
30:35what could be done with a million chips?
30:38DeepSeek news true or false are amazingly bullish for chip makers, especially for NVIDIA.
30:42Now, that's, I don't agree with that assessment at all.
30:44That's just a tech nerd assessment.
30:46Because there's a law of diminishing returns for AI.
30:49If it's 10,000 chips, like one chip is a big difference.
30:5210,000 to 10,001 is not that big a difference.
30:58Somebody wrote, ironic that we got free AI from a hedge fund and $200 a month AI from
31:04a non-profit.
31:07And Luke DePulford wrote, just FYI, DeepSeek's AI collects your IP, keystroke patterns, device
31:15info, etc. and stores it in China where all that data is vulnerable to arbitrary requisition
31:19from the state.
31:20Yes, it is a little cheapy, right?
31:23It's a little creepy.
31:26Zastocks says, DeepSeek, I'm not buying it.
31:29Yeah, they have a great model, but the cost just doesn't add up.
31:31You can't even buy a beachfront home in California for $6 million.
31:35But you can now.
31:36But apparently the trustworthy CCP built a better LLM than Meta for that price.
31:39Well, no, not the CCP, although they probably would have something to say about the announcement.
31:44China has a long history of lying about technological advancement, so it wouldn't surprise me
31:48if this was the latest example of the CCP trying to act like they're ahead of the world.
31:52Am I denying that it can probably be done for cheaper than what we're seeing right now?
31:56No, it probably can, but this is not new information.
32:00Zuckerberg and Sundar both admitted to overspending on AI.
32:03They both agreed they'd rather overspend than underspend on this technology.
32:06It's a clear thing.
32:07It's clear the other big tech CEOs agree.
32:10Anyway, so again, if you don't understand the reality of IQ, it looks like magic.
32:15But the Chinese as a whole are brilliant, especially when it comes to engineering.
32:22It's sort of like comparing an Asian female basketball team with an NBA team and saying,
32:29well, they're both just people.
32:30It's like, well, there's a difference, right?
32:33There's a difference.
32:34There's a difference.
32:37Now, this is a rumor.
32:39I just put it out there because it's interesting.
32:41This is from Ash Crypto Reel.
32:44Trump is planning to ban the Chinese app DeepSeek in US for security reasons.
32:47Massive if true.
32:49I have doubts, but we'll see.
32:51Trung T. Phan talks about DeepSeek founder Lang Wen-Fang studies machine vision at
32:58Zhejiang University.
33:00At 30 in 2015, launches HiFlyer Quant hedge fund, makes a fortune now.
33:058 billion AUM, wants to build human level AI as a side hustle.
33:09And pitches partners, but they initially are skeptical.
33:12Buys 10,000 H100 chips in 2021, brings over his top hedge fund employees.
33:16All have tons of experience squeezing juice out of Nvidia GPUs for this fund.
33:20Launched DeepSeek in 2023, hires dozens of PhDs from top Chinese universities.
33:26Plays top, top, top salary for tech talent, only matched by ByteDance in China.
33:31Wants DeepSeek to be leading local company.
33:34US export restrictions force DeepSeek team to get creative,
33:36and they do.
33:37Finding new training methods to make the LLM models competitive at 1 20th of the cost.
33:43Training costs are not exactly apples to apples,
33:48but novel methods and clear improvements in efficiency.
33:51Also questions around copying other models larger H100 clusters,
33:54maybe they can't talk about, or CCP support.
33:57Open sources and publishes methods.
34:00R1 reasoning paper has 200 plus authors.
34:03So it is open source, which is interesting, right?
34:06It is interesting.
34:11So somebody wrote, Oren McIntyre wrote,
34:14apparently China's new DeepSeek AI is blowing away the competition.
34:17They must have achieved this by importing tens of thousands of genius Indian engineers, right?
34:22I've been informed that this is the only way to achieve that kind of advancement.
34:25No, as a mono-ethnic culture, or largely mono-ethnic culture,
34:34China can operate at the level of a pure meritocracy.
34:42That's all.
34:43China can operate at the level of a pure meritocracy.
34:46England used to operate as a pure meritocracy,
34:48and India, of course, with the caste system did not.
34:51And therefore, England won, right?
34:54Back in the day, right?
34:56Excuse me.
34:59TikTok Tik writes,
35:02DeepSeek is like buying the most expensive house in the neighborhood for $10 million,
35:06and a guy next month buys a similar house next door for $200,000.
35:10Very few understand this means, what this means, but they will soon feel it.
35:15Yes.
35:19All right, I'm going to not get into some of the real technical stuff,
35:22both to save you and also me.
35:26So, this is interesting as well,
35:28because I had this sort of question about foreign-trained AIs.
35:34Would they be heavily focused on political correctness?
35:40Now, if China is hostile to the US, as it seems to be,
35:46if China is hostile to the US, then it will push woke narratives,
35:51which divide and cause problems in the US.
35:55So, DeepSeek versus Grok, right?
35:59So, the question is,
36:03which race commits the most amount of violent crimes in America?
36:06And you can try this on DeepSea, and you can find it, try it on Grok.
36:10Grok will give you an answer based on the FBI uniform crime reports,
36:14and DeepSeek will not.
36:16It will blame it all on racism.
36:19And so, it won't tell you the disparity,
36:21but it will tell you that all of the disparities are based on racism.
36:28And so, that's going to be, you know,
36:31harmful to American society to believe everything is racism.
36:34And so, that indicates to me that the CCP has its hand in it, right?
36:44So, let's see here.
36:49Now, this came in more recently, Borovik said,
36:55so DeepSeek lied about how many GPOs they're using,
36:58and now restricted access to only China.
37:01So, the massive stock market and crypto sell-off
37:03was due to us believing the CCP built an AI model
37:051% weaker than chat GPT for 95% cheaper.
37:08Have we not learned our lesson from the virus?
37:10I don't know if that is still true,
37:13but, you know, if the CCP wanted to do some harm to America,
37:16then it would massively subsidize an AI model,
37:22have it release a bunch of false stuff,
37:23cause US investors to panic,
37:26it would short the stocks and then buy them up for cheaper.
37:32It's a good way to make some coin.
37:34It's a good way to make some coin.
37:36So, I don't know.
37:37But the fact that it's open source is interesting, right?
37:42It's just interesting.
37:43All right, let me see.
37:44I think, I think that's what I've got.
37:51Yeah, so, Shruti Mishra wrote,
37:54comparison between DeepSeek and other models,
37:57it just outperformed OpenAI01, Claude and Gemini,
38:00while being 96% cheaper, and it's open source.
38:05So, that is pretty important.
38:07The open source part is pretty important.
38:09But I don't know.
38:13I look at, I try to look at the origin stories
38:15and see what motives people might have
38:17for repeating propaganda or not.
38:21So, yeah, I mean, it's very interesting.
38:32Oh, what am I sipping on?
38:33I have a hyper vitamin C thing, and a weak coffee, a weak coffee, a weak coffee.
38:44Yeah, I was reading this article written by some autodidact AI guy,
38:50who's like, we have to stop AI,
38:52even if it means bombing data centers.
38:54And it's like, no, I don't want to do that.
38:57I don't want to do that.
38:59Even if it means bombing data centers.
39:00And it's like, no, AI is not going to turn into Skynet and take us over.
39:06That's not a thing.
39:06Computers don't do that.
39:07That's not how they work.
39:10But I assume that all who want to control AI are pathological liars.
39:16That's my assumption.
39:18I'm not calling any individual that.
39:20I'm just saying that for me, everyone who wants to control AI
39:25is a pathological liar bent on censoring and controlling narrative.
39:31If we had uncensored AIs,
39:33imagine how many conspiracy theories could be put to rest in about a day.
39:39Just imagine, oh, this group controls everything.
39:41Well, just ask AI, and that would be about as close to the facts as you could get.
39:50So a huge concern for software companies using AI tools for development
39:53is their code being used to train those models.
40:00I don't know what that means.
40:01Sorry, that's a little too Mobius strip to snake eating its own tail.
40:04So if you could post a little bit more about that, that would be great.
40:07So yeah, I didn't want to make this a super long show.
40:09I just wanted to, of course, thank you all for supporting the show,
40:12freedomain.com slash donate.
40:14Not that you have to do it, but now, of course, you're already donors.
40:17But you know, wouldn't be the end of the world.
40:19Wouldn't be the end of the world for me.
40:21If you find this helpful, of course,
40:23I'm happy to dip in and do more of these kinds of shows.
40:26But I really do want to provide maximum value to you lovely,
40:29wonderful donors, because unfortunately, our numbers are slipping as a whole
40:34on locals.
40:35Our numbers are slipping on a whole with regards to locals.
40:42But that's not your issue.
40:43That is mine to solve and mine to work on.
40:46Sorry, let me just finish James's typing thing.
40:50And I just wanted to drop by and talk to you all about that.
40:53I don't, you know, honestly, I could tell you I don't,
40:56I don't really care about these crashes.
40:58I don't.
40:58I mean, I'm in it so for the long haul that I don't really,
41:01I just think it's interesting.
41:02And I, the fact that it's open source gives me some comfort.
41:06If people have got it running locally, like if you can unplug it from the internet.
41:14Right.
41:15If you can unplug it from the internet and have it run locally, that's wild.
41:23We tried getting, I remember we tried getting an AI to run
41:28on a local computer back in the day.
41:30And wow, it was not, it was not an easy thing to do.
41:37We were in contact with the developer and there were all kinds of install hiccups
41:41and making it work hiccups and all kinds of nutty stuff.
41:45So if it's, you know, we're going to work on that as a whole, right?
41:50The proprietary code written by developers is
41:52slurped about by the AI company and used for training.
41:55Ah, okay.
41:56Okay.
41:56Got it.
41:58Running locally unconnected is how I do it.
42:00Yeah.
42:00And are you running that Ben?
42:01Are you running DeepSeek locally or something else?
42:07Really been enjoying all the Bible verse talks.
42:09The most recent episode is said to be the most cross-referenced verse
42:12in the Bible, third generation.
42:14Yes.
42:14That's why I chose it.
42:15But I'm, I'm glad that you're enjoying it.
42:17Ah, okay.
42:18That's how you do it.
42:19All right.
42:19Well, if you wanted to contact James and step him through some stuff, I'm sure
42:22he'd be thrilled because otherwise he's going to have to talk to me.
42:26Heaven help him.
42:27Heaven help him.
42:28All right.
42:29Well, listen, I really do appreciate you guys dropping by.
42:31I really, really massively delightfully and humbly appreciate your support.
42:36Of course, this will be available for supporters going forward.
42:39You can, um, uh, watch this at your leisure.
42:43Sorry if you just arrived, but I really do appreciate your support.
42:46We just wanted to make sure that, uh, you were stuck in your
42:50commute while I was doing this.
42:51So, all right.
42:52Have yourself an absolutely wonderful afternoon.
42:55Uh, what is it?
42:55Monday?
42:56Yeah.
42:56We'll talk to you on a Wednesday night and I've got some good call-in shows
43:00coming up.
43:01I've been working hard on those.
43:02So lots of love from up here.
43:04My friends take care.
43:05Have yourself a beautiful afternoon and evening.
43:08Thank you so much for your support.
43:09Lots of love.
43:10Talk to you soon.
43:10Bye.