Note that there are a fair number of native speakers of English in Nigeria - more than in all but 3 or 4 US states.
In addition, "non-native" English speakers in India (and Nigeria?) typically study English from the first grade, and in many cases attended elementary schools where English was the language of instruction.
I think the differences between US English and both Indian and Nigerian English have more to do with divergent evolution of the educational systems. British English has a lot of differences, too, but we don't notice it as much unless we run across things like "whilst", probably because there's more media crossover. (if you find yourself reading Thomas the Tank Engine to kids it jumps out at you, though - the entire vocabulary for railroads evolved during a period when US and British English were diverging)
In section 2, 34% of cases are found to have "substantive" disagreements differing by 2 or more buckets - True + Misleading, Mostly True + False, or True + False.
This is probably a better measure than the headline one. It's still a concerning fraction, although some fraction is no doubt due to forcing "I don't know" cases to return an answer anyway.
It's quite interesting - this isn't ethernet as we know it. Instead of each NIC using its own free-running clock, all the physical layers are sync'ed to each other at layer 1. (note that gigabit ethernet, which is what it uses, sends data at all times - when idle it sends the idle symbol)
"I wonder if this is actually true in the long-term though. If they were to flood the market with lots of high capacity memory, then I think our programs would start using more memory too. As a result we might end up needing more memory faster compared to if they keep demand unmet."
It's a gambler's ruin problem. Future profits are worth zero if you go out of business first.
Jatin Malek on Twitter had perhaps the best explanation of the DRAM crunch:
"The reason why RAM has become four times more expensive is that a huge amount of RAM that has not yet been produced was purchased with non-existent money to be installed in GPUs that also have not yet been produced, in order to place them in data centers that have not yet been built, powered by infrastructure that may never appear, to satisfy demand that does not actually exist and to obtain profit that is mathematically impossible."
That's just a snarky way to describe all business investment that requires purchasing things made in the future. It's entirely normal.
You could literally rewrite the quote to be about iron and about building railroads for trains and passengers that don't exist yet. See how silly that would be?
Except the "profit that is mathematically impossible" part. That's just made up and false. It's entirely possible that we are actually underestimating demand, and there is going to be tons of profit. Nobody knows for sure, but profit is very, very, very possible.
> You could literally rewrite the quote to be about iron and about building railroads for trains and passengers that don't exist yet. See how silly that would be?
My point is, railroads turned into a real thing. The demand was real and the general large-scale investment was justified.
Just because some companies made bad decisions doesn't mean the railroad industry as a whole was some kind of mirage or mistake. Laying down tracks for trains and passengers that didn't exist yet is still necessary.
>Except the "profit that is mathematically impossible" part. That's just made up and false. It's entirely possible that we are actually underestimating demand, and there is going to be tons of profit.
JP Morgan says $650 billion in annual revenue required to deliver mere 10% return on AI buildout is equivalent to $35 payment from every iPhone user, or $180 from every Netflix subscriber 'in perpetuity.'
Very, very, very unlikely it makes profit, which why AI keeps getting overhyped by CEOs.
Altman said months ago that they are expecting around $65/user/month from ad-supported ChatGPT. A strong hint about where they see account prices in the future.
When you run the numbers, $65/mo turns AI investment into a a 5-7 ROI, which is totally within normal bounds.
Considering there are over a billion unique weekly active users for the major labs, and demand has been relentless, it's a pretty easy sell to get investors on board.
Those numbers sound... unrealistic to me. Just doing some napkin math: 65 $/user/month / 0.01 $/ad ~= 6500 ads/user/month, which is about an ad per minute if you assume someone is using the chat interface 4 hours a day including weekends. Maybe you see that behavior from "my GF/BF is AI" types but I'm also already assuming 0.01 $/ad which is super high to my understanding (if you work in adtech please correct me if I'm wrong). I don't forsee over 50% of your workday or leisure time spent in ChatGPT as likely, especially if the ad rate is well beyond YouTube's nigh-unusable amount it is now.
Did they consider that profits on the build out won't be uniform, i. e. there will be some companies that go under but the rest of them will capture the profit?
Some companies going under doesn't change anything about the market as a whole.
If the demand is real and the company just sucked, their users and infrastructure will end up at a competitor: the value for that one company is bigger, but the overall per-user bill remains about the same.
If the demand is fake the infrastructure will be sold off at a big loss, allowing new companies to enter the market with far smaller investment costs, allowing them to undercut the competition, driving down the price users expect to pay for compute, resulting in a race to the bottom between the remaining AI companies in an attempt to attract enough users that their hardware won't sit idle - which in turn makes it far less likely that they'll be able to hit those revenue figures. And a bunch of investors just lost a few billion dollars, of course.
Consider that a large majority of the revenue will come from businesses.
Even $100 per month per employee will likely turn out to be quite reasonable, if it can make employees more productive by several hundred dollars per month.
the big question is: who would pay for those services then?
I mean I love simplicity, and if economy could be simplified to a big money printer machine directly printing the money to a burner, then it would be so simple, that even a short context window could comprehend the economic cycle finally!
1. What is the performance of this implementation of an assembly server? The comment you replied to answered it: It's likely crap. Hand written assembly is almost always worse than compiler.
2. How much performance can a human still squeeze out at the assembly level? That question is different and remains unanswered. But in my experience the answer is the same.
Also note “Disks are like snowflakes - no two are alike”, Krevat, Tucker & Ganger, HotOS 2011. The number of tracks and bit density is not the same on different surfaces within the same disk, or across disks of the same model.
My only experience with Veracrypt is via a law firm I was consulting with, who used it to protect some files they were sharing with me. Law firm and their end client are both big, prestigious companies.
In addition, "non-native" English speakers in India (and Nigeria?) typically study English from the first grade, and in many cases attended elementary schools where English was the language of instruction.
I think the differences between US English and both Indian and Nigerian English have more to do with divergent evolution of the educational systems. British English has a lot of differences, too, but we don't notice it as much unless we run across things like "whilst", probably because there's more media crossover. (if you find yourself reading Thomas the Tank Engine to kids it jumps out at you, though - the entire vocabulary for railroads evolved during a period when US and British English were diverging)