Frank Ricard

737 posts

Frank Ricard banner
Frank Ricard

Frank Ricard

@NovusOrderM

Katılım Mayıs 2011
595 Takip Edilen458 Takipçiler
Frank Ricard
Frank Ricard@NovusOrderM·
I wonder if we get to a point where AI is structured like humans in that they are awake (inference) and asleep (training). Humans, when awake, are constantly solving problems and learning new things. These are stored in short-term memory. Then, when humans sleep, the hippocampus activates and sends signals to the neocortex - converting short-term memory into long-term memory. I can't help but notice the similarity between inference and training. Do we get to a point where AIs are generating a ton of tokens during inference and storing that in short-term memory (daytime hours). Then, when power is cheaper at night, the server racks "sleep". The cluster goes offline, replaying the day's interactions, generating synthetic past memories and algorithms to protect its core knowledge and gently update its global weights. This is how I imagine RSI might work.
English
0
0
0
17
Frank Ricard
Frank Ricard@NovusOrderM·
Five stages of grief. The value of CRM's historical data decreases over time, which has always been the case. Data collected 20 years ago is less valuable than data collected 6 months ago - anecdotes aside. CRM's value proposition is that the data generated in the future goes up in value while also offsetting the declining value of historical data (more data collected + better connection to historical data). Agentic AI will compress the value of future data that CRM collects. Agentic AI will also be better at distilling said future data. This applies pressure to both sides of CRM's value proposition. Not to mention the implementation dip. Once agentic AI is plug and play, the implementation dip is flattened, which is yet again an attack on traditional CRMs.
Frank Ricard tweet media
Maddy A@its_maddy_a

“I think we are getting brainwashed.” @Benioff said this on @theallinpod. “We’re using $300M of @AnthropicAI this year… the vast majority of those tokens don’t need to go to Anthropic.” Some tasks need @claudeai . Some need @OpenAI . Most need smaller, cheaper, faster models like @ZeroGPU_AI @Benioff believes in what we do - @salesforcevc should take a look. zerogpu.ai

English
0
0
0
43
Frank Ricard
Frank Ricard@NovusOrderM·
@TMTLongShort I know the ai-2027 article has been kind of dragged over the last year, and the doomerism "it's going to kill us all" is still way overblown IMO, but every so often I go back and reference it, and boy has it been pretty accurate thus far.
English
0
0
5
1K
Just Another Pod Guy
Just Another Pod Guy@TMTLongShort·
The consumer is going to act so fucking odd when they realize ASI is real and the tech bros found a way to build god. Econ textbooks and macro models will have to be thrown out and re-written from first principles. You. Are. Not. Ready. Travelmaxx while you can. Soon Americans will be persona non grata if we gate-keep access as much as I think we will. The end-state is abundance but first we have to drain the world of commodities and destroy international service sectors to pay for the infrastructure ramp. You. Are. Not. Ready. 🫡
English
53
27
574
76K
Frank Ricard retweetledi
Andrej Karpathy
Andrej Karpathy@karpathy·
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
7.1K
9.8K
128.4K
18.4M
Frank Ricard retweetledi
Liz Thomas
Liz Thomas@LizThomasStrat·
The explosion of agentic AI and compute shortages are pushing up prices: Average LLM token costs are now $2.12/mil tokens,+12% this week alone and +65% since end of Feb.
Liz Thomas tweet media
English
38
190
1.1K
635.1K
Unfiltered Artist
Unfiltered Artist@EmpireEnjoyer3·
@Danvir47 The war in Iran was a good idea. The execution was horrendous. Trump listened to the Turks and Qataris on not using the Kurds to invade Iran in the first two weeks. This is where it all went south.
English
4
0
2
274
Unfiltered Artist
Unfiltered Artist@EmpireEnjoyer3·
TRUMP DID THIS BY NOT FINISHING THE JOB. - He caved to Turkey and Qatar on cancelling the Kurdish invasion. - He refused to arm Iranian citizens. - He caved to Iran and cancelled power plant and bridge day. - He continues to stall to this day. He will go down in history as the president who oversaw America’s defeat in Iran.
OSINTdefender@sentdefender

According to the New York Times, citing U.S. officials, Iranian forces exercised a level of battlefield adaptability, not previously attributed to their military. Per the report, possibly by liaising with their Russian counterparts, Iranian forces were able to learn U.S. flight tactics and better employ air defenses against U.S. aircraft, as officials told the New York Times that they believed American air tactics had become somewhat predictable.

English
27
9
91
10.1K
Frank Ricard
Frank Ricard@NovusOrderM·
@greenbreadmaker @hamids If memory is able to scale to allow Micron to 10x, then that means power is able to scale, CoWoS is able to scale, etc. Which would mean every company involved in this scale up would be re-rated. The question isn’t demand anymore. It’s capacity and capital.
English
0
0
2
73
liello
liello@greenbreadmaker·
@hamids A 10x for $MU would make it the biggest company in the world by a trillion or two… I have high hopes as well but may be very difficult. I hope I’m wrong. Would be curious to see why you specifically think it could 10x by 2030? Obviously not a guarantee or prediction.
English
3
0
1
293
Hamid
Hamid@hamids·
According to $NVDA CEO/Founder Jensen: "we are at the beginning of a decade-long build-out" when responding to the cyclical question for memory & $MU specifically. So turns out all the "it's cyclical" bears about $MU might be right that it's cyclical! The only problem is that this cycle might last another 10 years according to the 1 person who knows the most about AI demand! Could you imagine 10 more years of growth for $MU!? This video is worth watching:
The Future Investors@ftr_investors

Very interesting conversation where Jensen Huang and Michael Dell explain how AI is structurally changing the memory industry 💾👇 $NVDA $DELL $MU $HXSCL

English
22
42
463
78.1K
Frank Ricard
Frank Ricard@NovusOrderM·
- Kill Iran's export revenue permanently. Destroy Kharg Island's loading infrastructure. Kill the terminals, the tank farm, the loading platforms. Not degrade, destroy completely. - Russia. Use the same tool that redirected Russian tankers from China, but at scale. Entice Russia with a structured sanction relief package. Russia floods the market with oil to capitalize on $130+ prices. - Saudi/UAE Quid Pro Quo. Max output via Red Sea. - Domestic Production Sprint. Emergency executive orders, expedited permitting/drilling, temporary waiver of certain EPA regs, and federal land lease accelerated. US producers don't need much encouragement at $130 oil prices. - SPR. Released surgically during price spike moments. - Hormuz Convoy. By July, oil prices will be elevated but managed. Begin a slow Hormuz escort through cleared lanes & surveilled lanes. - IRGC will mine and harass as much as possible, and there's a serious risk of tankers being hit or (more devastatingly) a US Navy ship. My guess is some form of these strategies (and more that I'm too dumb to recognize). Higher oil prices through August/September at least, spiking early and settling down by early Fall (unless something devastating happens, which is a real risk). Base case goal: - Iran decimated - Oil in the $80-90 range by October. Gas $3.25-$3.75. Both falling going into the midterms. The trendline in prices is more important than the absolute number. Show Oil/Gas prices falling in September/October and you might just preserve a win at midterms. Best case goal: - Iran capitulates by July/August under US terms. I think this is very unlikely. I believe the last 6 weeks were spent planning and preparing for this. China is also still a wildcard.
Open Source Intel@Osint613

BREAKING: 🔴 U.S. President Donald Trump: "I have instructed Secretary of War, Pete Hegseth, The Chairman of The Joint Chiefs of Staff, General Daniel Caine, and The United States Military, that we will NOT be doing the scheduled attack of Iran tomorrow, but have further instructed them to be prepared to go forward with a full, large scale assault of Iran, on a moment’s notice, in the event that an acceptable Deal is not reached. Thank you for your attention to this matter! President DONALD J. TRUMP."

English
0
0
0
55
Frank Ricard
Frank Ricard@NovusOrderM·
There's something innate in humans, where most believe there's something unworldly about our brains. Where "true intelligence" is whatever humans can do that machines can't. Applying this thinking to AI will continue to result in moving the goalposts. In the 1980s, people said a machine could never beat a human at Chess. A decade ago, people said a machine could never create art, write poetry, or understand natural language context. The brain is a giant biological computer, trained on hundreds of millions of years of evolution. Evolution that eventually resulted in the modern brain, where curiosity and problem-solving reign supreme. Compounding the last few hundred thousand years of curiosity and problem-solving, the net result is the brain we all have today. Natural selection operates as a highly inefficient but incredibly rigorous optimization algorithm. The loss function in AI is a more efficient Darwinian-like innovation. It allows AI to minimize errors and improve over millions of iterations, similar to how biological brains evolved through the ultimate loss function: survival and reproduction. One thing I am certain of: The more capable AI becomes, the further the goalposts will move. But eventually, we will run out of field.
Big Brain AI@realBigBrainAI

Mathematician Terence Tao offers a counterintuitive take: AI doesn't look intelligent because our definition of intelligence was wrong all along. He argues that the entire history of AI has followed a predictable pattern: "The history of AI has been here's a task that only humans can do, like maybe it is read natural language or win at chess or solve a math problem, and then one by one someone finds some AI algorithm that also does that." But every time a machine cracks one of these "uniquely human" tasks, we move the goalposts. The solution never feels like real thinking: "You look at how it's done and it doesn't feel like intelligence. It's, oh, it was some trick. You just cobbled together these neural networks and you ran some algorithm, and we were looking for some elusive intelligent way of thinking, and we don't see it in the tools that actually solve our goals." Tao then flips the problem on its head. What if the issue isn't with the machines, but with us? "But maybe it's actually because intelligence is not what we think it is." He points to large language models as the clearest case. What they do sounds almost embarrassingly simple: "Large language models in particular become very successful, and a lot of what they're doing is just predicting the next token, clicking the next word in a sentence. And that doesn't sound like something which is intelligent." To show why this feels wrong, Tao draws a comparison to how we'd judge a human doing the same thing: "If you ask someone to improvise a speech and they have no preparation, and at every moment they're just saying the next word that comes to their mind, you don't think that this could actually work." And yet it works for LLMs. Which forces an uncomfortable possibility: "Maybe that's actually a lot of what humans do as well."

English
0
1
0
55
Osman Sarood
Osman Sarood@osarood·
@BBNFL1984 @amitisinvesting "It’s the biggest technological advancement of all time. We replace workers with compute." when did I say that is not the case? However, AI revolution doesn't need memory stocks to 10X .. especially ones which have no moat.
English
3
0
2
192
amit
amit@amitisinvesting·
$MU $818 to $668 who’s buying the dip?
amit tweet media
English
541
170
3.8K
820.1K
Frank Ricard retweetledi
Matt Shumer
Matt Shumer@mattshumer_·
I firmly believe that even the most optimistic people in AI are severely underestimating how big the market for inference is going to be.
English
85
24
759
106.2K
Frank Ricard
Frank Ricard@NovusOrderM·
@TMTLongShort And even then, under that hypothetical, AI’s themselves would be programmed to be curious for us. The ultimate objective function will be ”be curious, ask questions, find the solutions”.
English
0
0
2
95
A. H.
A. H.@CountDataLove·
@TheLongInvest $817B cap on a memory company. Bubble math is clear. This is why allocators are building complementary exposure outside US tech concentration.
English
1
0
0
366
The Long Investor
The Long Investor@TheLongInvest·
The comments below this post are worse than feared. I honestly thought you knew we were in a bubble and understood what parabolic moves mean when the market is in Wave 5. There is a 100% guarantee that all parabolic moves like on the $MU chart below will trigger a decline of at least -50%. That is a certainty. It can still run higher in the short term but the crash will come. Bookmark this because I will be referencing this a lot
The Long Investor@TheLongInvest

$MU Honestly, we are in living in a time where people think this is normal And FOMO'ing that they did not chase this. This is as parabolic as you'll see and this is a $817 BILLION MARKET CAP company Yes we are in a bubble, yes they will all collapse and yes this time IT IS NOT DIFFERENT.

English
103
8
416
131.2K
Colin
Colin@colin_gladman·
@TechSlayer528 @TheLongInvest And it’s the number 1 stock I buy puts on, so not sure what you’re getting at, but I don’t care.
English
1
0
3
167
Frank Ricard retweetledi
Dwarkesh Patel
Dwarkesh Patel@dwarkesh_sp·
Continual learning sometimes gets discussed as if the goal is to dissolve the context/weights distinction. Let the model just keep accumulating, fine-tuning itself on the fly. @karpathy points out, though, that this isn't how humans do it. Our working memory gets wiped regularly. What we actually have is a consolidation process (sleep) that distills stuff into the brain, in a weird and lossy way. This is very different from how people sometimes talk about continual learning. It's not obvious it's something you can get for free from doing long enough RL loops.
English
39
52
988
58.3K
Just Another Pod Guy
Just Another Pod Guy@TMTLongShort·
Fuck it. Out of BTC 🫡 (Paper hands. Hope I have to buy back in 10x higher) I’m now 70/30 Gold/Cash
English
56
4
312
40.6K
Frank Ricard
Frank Ricard@NovusOrderM·
This. The PwC study surprises me, but it also tells me a lot about another bottleneck. It’s the consumers. My firm (PERE) just had our all hands meeting last week. At the end, we went around and talked about what we’re using AI for (our firm is basically all AI maxi’s at this point). Two folks (me included) had actual numbers we could point to. We saved in total, in Q1 2026, about 50% in legal costs YoY. The real ROI comes when the consumers catch up. And they will catch up, because this is the type of technology that will kill you if you don’t adopt it.
English
0
0
1
135
Sidika Kelly | Ops → Web3
This AI ROI debate reminds me of when supermarkets introduced automated ordering systems. At first, everyone asked: “Where is the ROI?” The benefits sounded theoretical: better availability, less spoilage, lower inventory, improved forecasting. But the real gains only appeared after workflows, staff behavior, and operations adapted around the technology. AI may be going through the exact same phase right now.
English
1
0
3
1.3K