Re: p2
Definitely a bit tedious, I had to "play" a whole session to spot bugs that I had. It took me far longer than average. I had buggy dissepearing boxes because of update order, I would reccomend a basic test case of pushing a line/pyramid of boxes in every direction.
zogwarg
joined 1 year ago
Updated Reasoning
Ok it probably works because it isn't bang center but a bit up of center, most other steps most be half half noise vertically, and the reason it doesn;t minimize on an earlier horizontal step (where every step is mostly half half), is because the middle points on the trunk, that don't contribute to the overall product therefore minimizing it even lower.
Day 14, got very lucky on this one, but too tired to think about why part 2 still worked.
spoiler
#!/usr/bin/env jq -n -R -f
# Board size # Our list of robots positions and speed #
[101,103] as [$W,$H] | [ inputs | [scan("-?\\d+")|tonumber] ] |
# Making the assumption that the easter egg occurs when #
# When the quandrant product is minimized #
def sig:
reduce .[] as [$x,$y] ([];
if $x < ($W/2|floor) and $y < ($H/2|floor) then
.[0] += 1
elif $x < ($W/2|floor) and $y > ($H/2|floor) then
.[1] += 1
elif $x > ($W/2|floor) and $y < ($H/2|floor) then
.[2] += 1
elif $x > ($W/2|floor) and $y > ($H/2|floor) then
.[3] += 1
end
) | .[0] * .[1] * .[2] * .[3];
# Only checking for up to W * H seconds #
# There might be more clever things to do, to first check #
# vertical and horizontal alignement separately #
reduce range($W*$H) as $s ({ b: ., bmin: ., min: sig, smin: 0};
.b |= (map(.[2:4] as $v | .[0:2] |= (
[.,[$W,$H],$v] | transpose | map(add)
| .[0] %= $W | .[1] %= $H
)))
| (.b|sig) as $sig |
if $sig < .min then
.min = $sig | .bmin = .b | .smin = $s
end | debug($s)
)
| debug(
# Contrary to original hypothesis that the easter egg #
# happens in one of the quandrants, it occurs almost bang #
# in the center, but this is still somehow the min product #
reduce .bmin[] as [$x,$y] ([range($H)| [range($W)| " "]];
.[$y][$x] = "█"
) |
.[] | add
)
| .smin + 1 # Our easter egg step
And a bonus tree:
I liked day 13, a bit easy but in the right way.
Edit:
Spoilers
Although saying "minimum" was a bit evil when all of the systems had exactly 1 solution (not necessarily in ℕ^2),
I wonder if it's puzzle trickiness, anti-LLM (and unfortunate non comp-sci souls) trickiness or if the puzzle was maybe scaled down from a version where there are more solutions.
re:followup
If you somehow wanted your whole final array it would also require over 1 Peta byte ^^, memoization definetely reccomended.
Day 11
Some hacking required to make JQ work on part 2 for this one.
Part 1, bruteforce blessedly short
#!/usr/bin/env jq -n -f
last(limit(1+25;
[inputs] | recurse(map(
if . == 0 then 1 elif (tostring | length%2 == 1) then .*2024 else
tostring | .[:length/2], .[length/2:] | tonumber
end
))
)|length)
Part 2, some assembly required, batteries not included
#!/usr/bin/env jq -n -f
reduce (inputs|[.,0]) as [$n,$d] ({}; debug({$n,$d,result}) |
def next($n;$d): # Get next # n: number, d: depth #
if $d == 75 then 1
elif $n == 0 then [1 ,($d+1)]
elif ($n|tostring|length%2) == 1 then [($n * 2024),($d+1)]
else # Two new numbers when number of digits is even #
$n|tostring| .[0:length/2], .[length/2:] | [tonumber,$d+1]
end;
# Push onto call stack #
.call = [[$n,$d,[next($n;$d)]], "break"] |
last(label $out | foreach range(1e9) as $_ (.;
# until/while will blow up recursion #
# Using last-foreach-break pattern #
if .call[0] == "break" then break $out
elif
all( # If all next calls are memoized #
.call[0][2][] as $next
| .memo["\($next)"] or ($next|type=="number"); .
)
then
.memo["\(.call[0][0:2])"] = ([ # #
.call[0][2][] as $next # Memoize result #
| .memo["\($next)"] // $next # #
] | add ) | .call = .call[1:] # Pop call stack #
else
# Push non-memoized results onto call stack #
reduce .call[0][2][] as [$n,$d] (.;
.call = [[$n,$d, [next($n;$d)]]] + .call
)
end
))
# Output final sum from items at depth 0
| .result = .result + .memo["\([$n,0])"]
) | .result
I remember being quite ticked off by her takes about free will, and specifically severly misrepresenting compatibilism and calling philosphers stupid for coming up with the idea.
One look day 9 and I had to leave it until after work (puzzles unlock at 2PM for me), it wasn't THAT hard, but I had to leave it until later.
re:10
Mwahaha I'm just lazy and did are "unique" (single word dropped for part 2) of start/end pairs.
#!/usr/bin/env jq -n -R -f
([
inputs/ "" | map(tonumber? // -1) | to_entries
] | to_entries | map( # '.' = -1 for handling examples #
.key as $y | .value[]
| .key as $x | .value | { "\([$x,$y])":[[$x,$y],.] }
)|add) as $grid | # Get indexed grid #
[
($grid[]|select(last==0)) | [.] | # Start from every '0' head
recurse( #
.[-1][1] as $l | # Get altitude of current trail
( #
.[-1][0] #
| ( .[0] = (.[0] + (1,-1)) ), #
( .[1] = (.[1] + (1,-1)) ) #
) as $np | # Get all possible +1 steps
if $grid["\($np)"][1] != $l + 1 then
empty # Drop path if invalid
else #
. += [ $grid["\($np)"] ] # Build path if valid
end #
) | select(last[1]==9) # Only keep complete trails
| . |= [first,last] # Only Keep start/end
]
# Get score = sum of unique start/end pairs.
| group_by(first) | map(unique|length) | add
Day 9 discussion
Part two for me was also very slow until I, speed up the index search by providing a lower bound for the insertion. for every insertion of size "N",
I have an array lower = [null, 12, 36, ...], since from the left any time you find free space for a given size, the next time must be at an index at least one larger, which makes it close to being O(N) [assuming search for the next free space is more or less constant] instead of O(N^2), went from about 30s to 2s.
https://github.com/zogwarg/advent-of-code/blob/main/2024/jq/09-b.jq
Day 8
Al lot of grid index shuffling these past few days! Not too difficult yet though, will this year be gentler or much harsher later?
Part 2 code in JQ
#!/usr/bin/env jq -n -R -f
[ inputs / "" ] | [.,.[0]|length] as [$H,$W] |
#----- In bound selectors -----#
def x: select(. >= 0 and . < $W);
def y: select(. >= 0 and . < $H);
reduce (
[
to_entries[] | .key as $y | .value |
to_entries[] | .key as $x | .value |
[ [$x,$y],. ] | select(last!=".")
] | group_by(last)[] # Every antenna pair #
| combinations(2) | select(first < last)
) as [[[$ax,$ay]],[[$bx,$by]]] ({};
# Assign linear anti-nodes #
.[ range(-$H;$H) as $i | "\(
[($ax+$i*($ax-$bx)|x), ($ay+$i*($ay-$by)|y)] | select(length==2)
)"] = true
) | length
60
AGI Sparklings proponents rejoice! Finding a literal map(*) means LLMs have a world model.
(awful.systems)
Transcribed:
Max Tegmark (@tegmark):
No, LLM's aren't mere stochastic parrots: Llama-2 contains a detailed model of the world, quite literally! We even discover a "longitude neuron"Wes Gurnee (@wesg52):
Do language models have an internal world model? A sense of time? At multiple spatiotemporal scales?
In a new paper with @tegmark we provide evidence that they do by finding a literal map of the world inside the activations of Llama-2! [image with colorful dots on a map]
With this dastardly deliberate simplification of what it means to have a world model, we've been struck a mortal blow in our skepticism towards LLMs; we have no choice but to convert surely!
(*) Asterisk:
Not an actual literal map, what they really mean to say is that they've trained "linear probes" (it's own mini-model) on the activation layers, for a bunch of inputs, and minimizing loss for latitude and longitude (and/or time, blah blah).
And yes from the activations you can get a fuzzy distribution of lat,long on a map, and yes they've been able to isolated individual "neurons" that seem to correlate in activation with latitude and longitude. (frankly not being able to find one would have been surprising to me, this doesn't mean LLM's aren't just big statistical machines, in this case being trained with data containing literal lat,long tuples for cities in particular)
It's a neat visualization and result but it is sort of comically missing the point
Bonus sneers from @emilymbender:
- You know what's most striking about this graphic? It's not that mentions of people/cities/etc from different continents cluster together in terms of word co-occurrences. It's just how sparse the data from the Global South are. -- Also, no, that's not what "world model" means if you're talking about the relevance of world models to language understanding. (source)
- "We can overlay it on a map" != "world model" (source)
@ESYudkowsky: Remember when you were a kid and thought you might have psychic powers, so you dealt yourself face-down playing cards and tried to guess whether they were red or black, and recorded your accuracy rate over several batches of tries?
|
And then remember how you had absolutely no idea to do stats at that age, so you stayed confused for a while longer?
Apologies for the usage of the japanese; but it is a very apt description: https://en.wikipedia.org/wiki/Chūnibyō,
And of course no experiments whatsoever, the cost of the Manhattan project, the hundreds of thousands of employees were merely a "focusing" magick, a sacrifice to re-enforce the greater powers of our handful of esteemed and glorious thinking men, who wrought the power of destruction from the æther.
@ESYudkowsky: Yes, but because the first nuclear weapon makers knew what the duck they were doing - analytic precise prediction of desired outcomes and of each intervening step. AGI makers lack similar mastery or anything remotely close, and have a much harder problem; that's the big issue.
@EigenGender: seems pretty noteworthy that the first nuclear weapons were made under conditions where they couldn’t do any experiments and they involved a lot of math but still worked on the first try.
view more: next ›
EDIT: I have a sneaking suspicion that the computer will need to be re-used since the combo-operand 7 does not occur and is "reserved".
re p2
Also did this by hand to get my precious gold star, but then actually went back and implemented it Some JQ extension required: