There is a three year old somewhere right now squatting in a cold puddle with mud up his forearms and snot running into his mouth, screaming at nothing. he will bite his sister in twenty minutes and cry for an hour after because she bit him back. he will lose his favourite toy in some wet grass and grieve it like a death. by nighttime he will have lived more honestly than most grown men manage in thirty years. and somewhere in a clean room with expensive candles and a view of some mountain, a grown adult who paid four thousand to sit on a cushion for a week is begging his own nervous system to produce what that kid got for free by falling face first into a thorn bush
Chebyshev feels irrelevant for detecting cheating.
Suppose honest students have score differences drawn from a symmetric distribution centered at 0 called D1.
Cheating students have differences drawn from another symmetric distribution centered at a mean > 0 called D2.
The observed data is then a mixture
D ~ P * D2+ (1-P) * D1
where P is the fraction cheating (which itself can be a random variable)
Chebyshev holds for D1 and the mixture. And given its a very loose upper bound, it cant tell the difference between D1 and the mixture for small P.
The actual question is: “Am I sampling from a single distribution or a mixture?”
Chebychev's inequality is just an upper bound on the probability of deviation from the mean. It says that far outliers MAY exist with a certain probablility. NOT that they WILL exist as this guy is claiming.
He is a well-known mathematician btw and I am sure he knows this and is making a bad faith argument.
If in the history of the semi final system across tournaments we see the top 2 reaching the finals significantly less often than 81%, that actually shows that there is some structural advantage due to the ground change or rest days between eliminator and Q2, where the losing team of Q1 is favored.
If we assume playoffs are random, then the probability that Q1 is the same as the final is 50%.
But Q1 = finals for 90% of the time, indicating the top 2 teams are inherently better than 3 and 4.
So even if we do a semi final the chances of the top 2 going into final will be considerably high.
For a basic calcuation, if Q1 = finals 90% of the time, then that means the losing team of Q1 beats the winner of the eliminator with a 90-10 advantage.
For simplicity assume both top 2 teams have a 90-10 advantage over 3 and 4.
Then in a semi final system, the final will be the top 2 teams with a 81% chance.
In all seriousness, one of the best ways to build and maintain your memetic resistance is to make sure that at any given time, you’re reading something from before 1900. It doesn’t much matter what it is.
Governments controlling capital flows by restricting the harder form of money (gold in this case) usually leads to the debasement of the softer money (INR in this case).
Probably because it was already happening, and this just accelerates it.
A similar analogy - a bank asks its customers to not withdraw cash for some time and keep it in deposits. This just makes customers aware of fractional reserve banking, and probably increases the chance of a bank run.
Whether AI "replaces humans" or "makes humans irreplaceable" in the labor market won't be a product decision. It will be a fact about the nature of reality that we all discover together.
coke zero caffeine zero sugar zero calories is the even more technodharmic drink of choice bcs it is entirely synthetic, it has no taste analog in nature. it is a testament to the spark of divinity in man, evidence that we can prescind from our base conditions to mirror our creator in infinite creativity
Computer use agents is probably where this is most obvious. Claude can pretty much do any kind of semi deterministic workflow involving interacting with sheets, docs and browser windows, as long as the demonstration from the human sufficiently breaks the workflow into subtasks.
However every time I try this, Claude does finish the task and learn how to do it, but it takes 2x more time than me and wastes too much compute. I would rather use that compute at more intelligent tasks like coding or math.
Many people think that when AI can do a task better than us, it will outcompete humans.
But token costs are not trivial, and, for many types of tasks, even skilled labor tasks, humans are much cheaper than AI. Increasing efficiency & compute supply will alter those calculations
If a VC wants to take equity in your business and thinks it can have a venture portfolio worthy return in a relatively short span of time, you should be allowed to call your ice cream business a startup.
There is a platform which offers access to secondary markets for US startups.
Have seen low demand for Perplexity shares, and every month the minimum investment amount keeps decreasing to lure more retail investors.
People leaving regular companies: Time for a change! Excited for my next chapter!
People leaving AI companies: I have gazed into the endless night and there are shapes out there. We must be kind to one another. I am moving on to study philosophy.