dl
https://toobnix.org/w/9oA5WZEkbjKkfbsSxEoJ8v #archive
Okay I guess on the hour it's the Sunday-morning-in-Europe #lispyGopherClimate .
a. I will lightly go over my recent #commonLisp #symbolic #DL ( #ai!) algo / article.
b. Excitingly that gave me occasion to use
(this-function foo &rest keys &key &allow-other-keys)
which I used to pass data-like parameters along without cluttering up arguements (waters' functions of seven arguments). People had tried to explain it to me previously but now I underst
#lispyGopherClimate #lisp #technology #podcast #archive, #climate #haiku by @kentpitman
https://communitymedia.video/w/c3GdAXe7BQTbK3VrcXCm7E
& @ramin_hal9001
On the #climate I would like to talk about the company that found #curl and #openssl's #deeplearning many (10ish) 0-day vulns "using #ai ". (#llm s were involved).
This obviously relates to my #lisp #symbolic #DL https://screwlisp.small-web.org/conditions/symbolic-d-l/ (ffnn equiv). Thanks to everyone involved with that so far.
I implemented that using #commonLisp #condition handling viz KMP.
As someone who now has a feedforward neural network of a single hidden layer implementation (in #commonLisp) I learned a lot about what has a spot to be stuck into the algorithm.
https://screwlisp.small-web.org/fundamental/sharpsign-sharpsign-input-referential-ffnn-dl-data/
Something that turns out to be easy to stick in is making training data which copies data out of the current context or even to simply point the training data definition into the #deepLearning inference input context though I did not prove convergence for the latter case.
Yeesh, okay it is out and I can stop telling you I am going to write it.
https://screwlisp.small-web.org/complex/lisp-feedforward-deep-learning-example/
#deepLearning #symbolic #ffnn pure ansi #commonLisp
The example deliberately causes a hallucination programmed into the training data via simple #roc reasoning. To my knowledge the ROC interpretation of ffnn inference is original to me, here.
The hallucination is in both lockstep and single #neuralnet activations.
Be the first to complain my #DL works on subregions of jagged lists! #programming
My #dlroc #deepLearning #DL in #CL #commonLisp #McCLIM project is a single file defining a package now, which is to say easy to load and work with.
As long as you know you should use (load (compile-file #P"dl-roc.lisp")) in this case. Otherwise it should be compatible with all previous notes and examples.
It's Sunday morning in Europe! #archive 8UTC Sunday as always #lispyGopherClimate #podcast since 2022.
https://toobnix.org/w/pU6zu95YDdyGKqsxxRf3Vg #peertube live
#RSS (recent times) https://toobnix.org/feeds/videos.xml?accountId=580185
@vnikolov 's Quality Without A Name toot #AI
As much as I can remember about the #lisp community #architecture and Christopher Alexander https://www.dreamsongs.com/Files/PatternsOfSoftware.pdf https://alexandria.common-lisp.dev/
My NicCLIM #IDE demo and the #DL book I am #writing - loose bibliography and sketch of chapters
#itchio #gamedev #programming #theory completely explainable game-embeddable #deepLearning system using #roc #statistics .
I coded this one simply and iteratively, since a few people worked on reimplementing my #commonLisp code to-be-simpler.
The gist is that I show that deep learning updates, and indicate training as well are a simple true-positive/true-negative/false-positive-false-negative equation of the previous time step in an eminently explorable way. #DL #AI
#lispyGopherClimate #live #technology #podcast (?!) https://toobnix.org/w/jQkCWCeFNRL9Utcr2GWurM
Chat live in #lambdaMOO as always https://lambda.moo.mud.org/
(@join screwtape
"hey
)
- I release NZ government secrets about #AI and #LLMs senior managers keep sending me
- I start /actually/ multimooing my own personal moos LambdaMOO
- Otherwise, my #commonLisp brain is entirely inside my #DL #DeepLearning #roc #statistics original formulation https://lispy-gopher-show.itch.io/dl-roc-lisp
@kentpitman featuring.
#lispyGopherClimate #technology #podcast #live https://toobnix.org/w/5QbQiLw7zrFiETgbv32kJk first half missing?
- One week from today is #LambdaMOO's annual festival, April Fool's.
- Let's have a text based #MUD #bonkwave pool party / concert
Notably /after/ submitting my #ELS2026 #commonLisp #deepLearning article https://european-lisp-symposium.org/2026
we just had this great thread: https://gamerplus.org/@screwlisp/116286095082069619 I will read #bookstodon and #programming suggestions by @riley and others. #ai #DL #ML
Hey everyone, on the hour it is the Sunday morning in Europe #peertube #ARCHIVE
https://toobnix.org/w/ktQA9DTKyAe9yhTdoJupxh
Do I actually even #code though?
A live #demo of two recent programs of mine.
https://lispy-gopher-show.itch.io/dl-roc-lisp
https://lispy-gopher-show.itch.io/nicclim
you can run in #commonLisp
(load "dl-roc.lisp")
.
#DL #AI #UI #madeWithLISP #programming #McCLIM
Miscellany
New ECL this week! https://ecl.common-lisp.dev/ @jackdaniel
#XMPP by Glenneth2 https://github.com/parenworks/CLabber
https://cyberhole.online/basic/ is this millenium's teletype
诛邪者王朝了，主的愤怒这一块。
两个轮椅灌伤害的都在我们这一边：
一个诛邪者“主的愤怒”，一发一个带走残血
另一个是柒，只需要跑到对面中间开大就好了（（
经过了70个小时的deadlock游玩，终于知道（大概了解）自己适合什么角色了
炽焱，Infernus，或者称“火男"
一技能——凝固汽油弹：
喷射凝固汽油弹燃烧，造成精神伤害 ，施加减速效果，并用覆盖目标。
攻击距离较短，对于我来说比较没用，主要是如果二技能在冷却的话，就用来清理兵线。
二技能——火焰冲刺：
向前冲刺，获得缓慢抗性 ，同时留下持续的火焰轨迹，造成精神伤害 。
这个是我的主要清理兵线的重要技能，只需要跑过去，就可以迅速清理无论多么密集的小兵；
也是我的对线”前奏“，先冲刺到敌人身边（速度很快，敌人命中率低，而且附加回血效果）让敌人开始燃烧，然后使用主武器射击，然后让火焰伤害带走敌人即可。
三技能——点燃：
武器命中会积累燃烧效果，持续造成灵魂伤害 。
这是对局刚刚开始第一个要发育的，因为开局比较脆，一二技能差不多都是去换血，而且前期主要吃兵线为主，只要将小兵点燃，刷起来非常快；
也是作为非常主要的伤害技能（比如4管血的敌人，还剩下一管血时逃跑，附加的火焰伤害也能在我停止射击的情况下带走敌人）；
点燃的条件不算苛刻。只需要持续射击队友几秒钟就可点燃，如果敌人不是身法大师，而是傻愣愣面对面对枪，他的血量会飞速下降然后逃跑的时候被烧死，非常非常重要的技能。
四技能——共振燃烧：
变成活体炸弹，在延迟后对爆炸，并附近所有敌人造成灵魂伤害，并施加眩晕效果。
这个爆炸不会杀死自己（不稳定化合物会，这个有血泪教训，比如吃下不稳定化合物找不到敌人，把自己炸死2333，这个主要是团战时候神风特工队的）；
这个还没怎么用过，因为二三技能和主武器太好用了，一般还没来得及用四技能对面就死了（1v1情况）。
反思：
一定不能1v多，这个不是代表一定不能，而是在后期，伤害型和输出型英雄同时在场，能二技能跑就跑。
惨痛教训1：
神父和麦金妮，神父的大招”主的愤怒“伤害非常高，而且后期时候，三根木桩差不多能打掉3000血，而我本身只有4000多，然后就是麦金妮的幻影屏障可以拦截我的逃跑路线，而且后期机枪射速叠满，200发弹夹，如果不回血，应该也能一套带走我；
被这个组合杀了好几次，后面看到神父和麦金妮走在一起撒腿就跑。
惨痛教训2：
好几次被杀死都是1v4（开尔文、麦金妮、神父、haze），所以以后还是尽量不单打独斗，抱团为好，这也导致我后期死了很多次没杀几个人，主要是遇到团被秒了
#Deadlock #异锁 #死锁 #DL #Infernus #炽焱
感兴趣的可以看DeadLock维基：https://deadlock.wiki
