"The game called for picking six numbers from 1 and 54. For a pro gambler, some sets of numbers—such as 1,2,3,4,5,6—aren’t worth picking because so many other players choose them, which would split the pot. Marantelli’s operation bought 99.3% of the possibilities."

Wouldn't this be 54x53x52x51x50x49 = 18,595,558,800 therefore 99.3% is 18,465,389,888.4 or basically over 18 billion dollars in tickets (at $1 per ticket)?

I also could not work this out.

I think there are a bewildering number of games on the Texas lottery [1] but none seem to give the odds listed. But even more confusingly the Texas lottery itself lists the chance of a jackpot as 1 in 25.8mil [2]. So I’m not sure what we’re missing?!

The 1:26mil number must be about correct as google says the whole lottery grosses about 1.2bn a year. So 100m a month. So about 12.5mil tickets sold per draw if they are twice weekly. So a roughly 50% chance of a jackpot. If it were as our numbers show it would take about 16 years for a jackpot to occur even if that were the only game people played.

1: https://en.m.wikipedia.org/wiki/Texas_Lottery

2: https://www.texaslottery.com/export/sites/lottery/Documents/...

I believe the game in the article is no longer offered. I think it was a regional game with 3 numbers like lucky 7s or whatever.

I suspect you have to choose the right numbers but not the right order. That makes the numbers work out.

Order doesn't matter, you need to divide by 6!

Where the final '!' means 'factorial' not just an emphatic 6.

"6 factorial" is 6 x 5 x 4 x 3 x 2 x 1 = 720.

There is a massive difference between "capable of generating music, sounds, and speech using text and audio inputs it’s never been trained on" and "sounds never heard before." This is simultaneously nonsense and obvious truth. Create a new color you've never seen before.

I really, really wish I could copy comments on mobile. I'm in my doctors office waiting and I want to throw this into rot13 so i can read it

Im really curious, what prevents you from copying comments on mobile?

why are you fealung with cache at all in a wrb context? that's what CDNs are for.

React, like Haskell creates new problems you need sophisticated solutions for.

CDN is for content. Useless for say a Google search type problem. So I call an API endpoint for data in my component. But what if I get rendered 600 times? Well luckily I have memoized the call! As long as it was under 2Mb that is.

This describes T-Mobile incredibly well...

I would highly recommend Thinking Fast and Slow by Daniel Kahneman. It comes on normal book format so you can read it.

Adversarial game playing as a way of training AI is basically the plot of War Games.

And also the breakthrough that let AlphaGo and AlphaStar make the leaps that they did.

The trouble is that those board games don't translate well to other domains. But if the game space can operate through the realm of language and semantics, then the hope is that we can tap into the adversarial growth curve, but for LLMs.

Up until now, everything that we've done has just been imitation learning (even RLHF is only a poor approximation "true" RL).

I don't want to be this guy, but can we get this whole thing removed? the lack of integrity here seems anathema to HN

If it turns out it really was written by AI and these errors are as dumb as they look the whole domain should be blocked and the numerous past submissions from there removed.

if we're going with hot takes, I've watched a lot of DefCon vids and many presenters come off as outlandish arrogant. not simply smug, more "I am levitating above the normies."

That's not specific to presenters; there's a lot of insecurity (no pun intended) on parade in this industry. The sort of people who can bridge air-gapped networks using bubblegum and popsicle sticks tend not to minor in human relations.

Just read it as showmanship. They're trying to be over the top for the sake of performance.

Nobody should be surprised by this. Speed and Quality are orthogonal after all. Why do it right when you can do it twice?