[QUOTE=chalsall;476335]This is almost too funny to be true. But it appears to be. The submission to Slashdot was accepted, but they couldn't deal with the exclamation marks![/QUOTE]

Turns out the acceptance email was actually from an experimental AI "story picker" they are working on.

The submission is now on [URL="http://slashdot.org/"]SlashDot[/URL]! "News for Nerds, stuff that matters."

[QUOTE=CRGreathouse;476380]2^77232917-1 has 23249425 decimal digits. The expected number of instances of ten consecutive identical decimal digits is about (23249425-9)/10^9 = 0.023249416. The probability is then about 1 - exp(-0.023249416) = 0.0229812 or roughly 2.3%.[/QUOTE]

And don't forget why even this seemingly-low-odds figure is misleading - with a dataset like this, one can look for any number of 'interesting patterns'. Failed to find a run of 10 consecutive same digits? How about a run of n digits of Pi, etc. If I look for 50 different 2%-likely patterns before finding one, "ooh - only 2% odds" for the one-that-turned-up is clearly an underestimate, since we are considering a large candidate class.

[QUOTE=ewmayer;476431]And don't forget why even this seemingly-low-odds figure is misleading - with a dataset like this, one can look for any number of 'interesting patterns'. Failed to find a run of 10 consecutive same digits? How about a run of n digits of Pi, etc. If I look for 50 different 2%-likely patterns before finding one, "ooh - only 2% odds" for the one-that-turned-up is clearly an underestimate, since we are considering a large candidate class.[/QUOTE]

Right! Had I reported the odds for a string of 10 9s it would have looked ten times as rare, but I figured we'd accept any digit repeating that often. Including other patterns you might look for would improve the 'rareness' estimate further.

If you have a decent idea of how many patterns you're looking for you can use [url=https://en.wikipedia.org/wiki/Bonferroni_correction]Bonferroni[/url] or [url=https://en.wikipedia.org/wiki/%C5%A0id%C3%A1k_correction]Šidák[/url] correction.

[QUOTE=CRGreathouse;476434]Right! Had I reported the odds for a string of 10 9s it would have looked ten times as rare, but I figured we'd accept any digit repeating that often. Including other patterns you might look for would improve the 'rareness' estimate further.

If you have a decent idea of how many patterns you're looking for you can use [url=https://en.wikipedia.org/wiki/Bonferroni_correction]Bonferroni[/url] or [url=https://en.wikipedia.org/wiki/%C5%A0id%C3%A1k_correction]Šidák[/url] correction.[/QUOTE]

And yet a run of 10 consecutive "9" as one of 10 consecutive "0" in a decimal system is much more fitting than a run of ten consecutive "4" or "7"s.

Yes of course that having so many concievable 9 digit patters one can be tempted to trivialize its meaning. However I would be surprised if for instance the first 9 digits of the the Mersenne Prime would appear once again in the decimal expansion. (It doesn't)

The digit '2' also appears to be under represented.
Digit Counts
0|2,325,846
1|2,324,106
2|2,323,306
3|2,325,845
4|2,326,305
5|2,325,065
6|2,324,655
7|2,324,051
8|2,326,039
9|2,324,207

[url]http://www.mersenneforum.org/showthread.php?t=5414[/url]

[QUOTE=CRGreathouse;476380]2^77232917-1 has 23249425 decimal digits. The expected number of instances of ten consecutive identical decimal digits is about (23249425-9)/10^9 = 0.023249416. The probability is then about 1 - exp(-0.023249416) = 0.0229812 or roughly 2.3%.[/QUOTE]

FYI, this goes back to my hint here:
[URL="http://www.mersenneforum.org/showpost.php?p=475345&postcount=145"]My hints[/URL]

I didn't want to be too specific because some clever person might analyze all the potential candidates to see which one had 10 nines in a row somewhere, to narrow it down.

Of course, I think Mark's hints were a little too on the nose, giving the first digit as well as the last two. That and knowing the FFT sizes we were talking about, plus a couple other tidbits were basically a dead giveaway.

In the future, the hints should be a little more subtle. :smile:

[QUOTE=Madpoo;476450]In the future, the hints should be a little more subtle. :smile:[/QUOTE]

Well, around the time my and Andreas' verify runs finished I gave away the game not once but twice in the span of a few hours - luckily it was the wee hours [US time] of the night between 30-31 Dec, and the giveaway was in form of a data-busy screengrab, so it appears no one (except fellow verifier ATH) noticed.

Here's what happened - airsquirrels made a post about timings on a superfast V100 GPU he had access to, but said something to the effect that he didn't have a handy summary he could post. But he had sent me a screengrab the day before, so I made a "This one?" reply-with-attachment - I was so focused on the timing data in said captcha that I completely overlooked the fact that his timing data were [b]for the exponent in question[/b]. Luckily, ATH (late morning/midday-ish of Dec 31 in Europe) spotted the leak (but only on his 2nd viewing of the post, a little over an hour after I made it) and scrubbed the attachment. I really wanted to show the fabulously fast V100 timings, so I spent a few minutes blacking out the offending data columns in the captcha (exponent in one col, 10k, 20k, etc- Res64s in another, the latter could be used in conjunction with a smallish possible-candidate set to do a bunch of shallow 10kiter runs and see which matched the corresponding Res64), first fwded the resulting image to ATH for his perusal, but on not hearing back from him for several hours and needing to get to bed, I reattached the edited image to the post - see below for that one.

Luckily I posted to the Hoff-admin-thread to effect "could one of my fellow mods eyeball the revised attachment, because this late at night I no longer trust myself", because indeed, the exponent was still on full glorious display. Batalov quickly saw my mod-thread post and the remaining data-leak and once again deleted the image. Not my finest 4-5 hours, I'm afraid - on waking ~4am (my time, US west coast) to check my just-completed DC run I saw notes from both Serge and Andreas to effect of "it's still leaky", as a result of which I replied something to effect of "guys, please don't ever trust me with the nuclear codes". So now our readers have both the official backstory re. verify runs and runup-to-the-official-announcement from Aaron and the untold scandalous sekrit story of how the earth was completely unprotected from the alien invasion fleet on 2 separate occasions that same night due-to (and from) me. :)

[QUOTE=ixfd64;476402][I]The Guardian[/I] just published an article: [url]https://theguardian.com/science/2018/jan/04/largest-prime-number-discovered-with-more-than-23m-digits[/url][/QUOTE]

[QUOTE]“I’m very surprised it was found this quickly; we expected it to take longer,” said Chris Caldwell, a professor of mathematics who runs a website on the largest prime numbers at the University of Tennessee at Martin. [B]“It’s like finding dead cats on the road. You don’t expect to find two so close to one another.”[/B][/QUOTE]

Nice going, Professor! That will clearly attract new dead cat hunters, for sure!
:tom::deadhorse:

[QUOTE=Batalov;476456]Nice going, Professor! That will clearly attract new dead cat hunters, for sure!
:tom::deadhorse:[/QUOTE]

Well if you ignore the less-than-optimal choice of metaphor, the underlying statement of opinion is extremely valid. This is surely an extremely unlikely find, at least in the context of the standard hypotheses about the Mersenne distribution.

[QUOTE=Madpoo;476450]FYI, this goes back to my hint here:
[URL="http://www.mersenneforum.org/showpost.php?p=475345&postcount=145"]My hints[/URL]

I didn't want to be too specific because some clever person might analyze all the potential candidates to see which one had 10 nines in a row somewhere, to narrow it down.

Of course, I think Mark's hints were a little too on the nose, giving the first digit as well as the last two. That and knowing the FFT sizes we were talking about, plus a couple other tidbits were basically a dead giveaway.

In the future, the hints should be a little more subtle. :smile:[/QUOTE]

Fair enough. The hint of the 7 didn't really reveal anything extra when I gave it, but made it easier (the gpuOwl usage was a huge hint). I posted the last two digits as the least revealing evidence I could think of showing I had figured it out. There was only a 1 in 40 chance of guessing the last two digits. Things like summing the digits or hashing the number or its decimal expansion are brute-forceable, so I didn't want to post those. I'll try to think of something better for next time.