Yeah. Hyperscalers who are building compute capacities became asset heavy industries. Today's Google, MSFT, META are completely different than 10 years ago and market has not repriced that yet. These are no longer asset light businesses.
Well, this sounds like a "no shit Sherlock" statement:
>>Finding 3: Natural "overthinking" increases incoherence more than reasoning budgets reduce it
We find that when models spontaneously reason longer on a problem (compared to their median), incoherence spikes dramatically. Meanwhile, deliberately increasing reasoning budgets through API settings provides only modest coherence improvements. The natural variation dominates.<<
Language models are probabilistic and not deterministic. Therefore incoherence _by definition_ increases as a response becomes lengthier. This is not true for humans, who tend to act/communicate deterministically. If I ask the human, to read a pdf and ask, is there a word of "paperclip" in the pdf? The human deterministically will provide a yes/no answer and no matter how many times we repeat the process, they will provide the same answer consistently (not due to autocorrelation, because this can be done across different humans). LMs will have a probabilistic response - dependent on the training itself: with a very well trained model we can get a 99% probabilistic outcome, which means out of 100 simulations, it will give you 1 time the wrong answer. We have no clue about the "probablistic" component for LMs, however, simulations could be done to research this. Also, I would be very curious about autocorrelation in models. If a human did a task and came to a conclusion "yes", then he will always respond with increasing amount of eyerolling to the same task: "yes".
Also, imagine the question: "is the sky blue?" answer1: "Yes." This has 0 incoherence. answer2: "Yes, but sometimes it looks like black, sometimes blue." While this answer seemingly has 0 incoherence, the probability of increased incoherence is larger than 0 given that answer generation itself is probabilistic. Answer generation by humans is not probabilistic.
Therefore, probability driven LMs (all LMs today are probability driven) will always exhibit higher incoherence than humans.
I wonder if anybody would disagree with the above.
My only issue is that i am not a developer, I am heavily reliant on Excel, i know it inside and out and just not sure whether OpenOffice supports excel files. In the past it barely did.
It is not the scale that matters here, in your example, but intent. With 1 joint, you want to smoke yourself. With 400, you very possibly want to sell it to others. Scale in itself doesnt matter, scale matters only as to the extent it changes what your intention may be.
If there is one exact sentence taken out of the book and not referenced in quotes and exact source, that triggers copyright laws. So model doesnt have to reproduce the entire book, it only required to reproduce one specific sentence (which may be a characteristic sentence to that author or to that book).
You can only read the book, if you purchased it. Even if you dont have the intent to reproduce it, you must purchase it. So, I guess NVDA should just purchase all those books, no?
I think this analysis has little to say.
What would be important to know how those $ are being spent, not where they are collected from. We do not know how those $ are being spent.
1. Doctors, Nurses, Administration (management and field administration), other. We need to know total employment and total salaries (including private practices).
2. OTC, prescription and hospital administered drugs (separated for acute, such as ER, and chronic, such as inpatient and elective surgery). We need to know how much is being spent on these, which is _potentially_ one of the culprits of large discrepancy between US healthcare vs European healthcare. What would be great to have these by large cohorts of population (<20; 20-65; 66-85; 85<) and maybe the top 5 buckets (i am guessing: cardiovascular - chronic; diabetes; accidents; hospice; dialysis)
3. Facility expenses (rent, maintenance, utilities, other contractor)
4. Other
Without these, very hard to opine reasonably on the state of affairs. And to be fair, I suspect there is a reason why proper expense breakdowns are not available.
There is no need. The personnel cost multiplier is 3.6x, the non-acute drug multiplier is 6.25x and the "something else" multiplier 77x. So we can debate that there are ~4x more people in the US and CoL is higher maybe by 0.5x turns, but there is no debate that the 77x shows something is awfully wrong.
My bet is the private insurance because we dont have transparent data on how much does the same procedure (broken down by personnel (doctor, nurse, admin), implied equipment amortization, rent, drugs) cost with and without insurance.
I could totally imagine that there are plus personnel expenses buried within the "something else", or acute drug prices, which are administered during an emergency at a hospital. But we don't know, because healthcare spending is a black box in the US.
They sign the purchase order on 1/1/26. AMD issues invoice to be paid in 30 days, that is 2/1/26. OpenAI triggers warrant and informs AMD on 1/2/26. OpenAI receives shares on 1/4/26. On 1/5/26 OpenAI and AMD announce the GPU purchase deal. On 1/30/26 OpenAI sells its shares in AMD. From proceeds, OpenAI pays AMD on 2/1/26. Thus, AMD financed OpenAI's GPU purchase via AMD's shares.
I did a calculation once. US spends $4.9T on healthcare: $2T on personnel, $500B on non-acute drugs (ie OTC + prescribed) and $2.4T on something else.
Germany spends $550B on healthcare: $430B on personnel, $80B on non-acute drugs and $31B on something else.
My guess is that the "something else", which is non transparent, is actually private insurance jacking prices up.
I mean, to be fair, Google's scam of how much GBs you have is very annoying and downright scandalous.
I had 16.5GB or so used up so it was flashing red. When paid for Gemini, my total space jumped to 2TB and my usage dropped to 12GB. Disgusting. So might as well switch to fastmail. Not sure.
Likely not true re adoption. According to McKinsey November 2024 12% of employees in the US used AI for >30% of their daily tasks. I saw another research early this summer, it said that 40% of employees use AI.
Adoption is already pretty relevant. The real question is: number of people x token requirement of their daily tasks equals how many tokens, and where are we there. Based on McK, we possibly around 17% unless remaining 50% of tasks requires just way more complexity, because then that would obviously mean the incremental tasks require maybe exponentially more tokens and then penetration will be indeed low. But for this we need to know total token need of daily tasks of average office worker.
I just heard a thesis that there is no bubble unless there is debt in it. Currently mostly internal funds were used for increasing capex. More recently started we seeing circularity (NVDA -> OpenAI -> MSFT -> NVDA), thus this is less relevant so far yet. Especially as around ~70% of data center is viewed to be GPU, so NVDA putting down $100B, that essentially funds "only" $140B of data center capex.
META is spending 45% of their _sales_ of capex. So I wonder when are they going to up their game with a little debt sprinkled on.
There is a book called "Talent is overrated" it essentially says, you need to 1) invest time, 2) do targeted practice, and 3) have a mentor, who helps you in targeted practice. Practice alone is not enough, it must be targeted at 1) what is relevant and/or 2) where your biggest weakness is at the moment.