agi.blue (149 MB) | 54 slices | source code of this web site (1,599 lines) | by BC | Fiverr | Discord | Video | Notes | Query
agi.blue's main slice has 46,565 pages and 1,682 connections
there is forward and backward induction
just i actually use english statements, not special syntax
programming in logic statements, and relating them to each other
Only know prolog by name, sorry.
in multiple ways
i generalize the idea of prolog
i believe in the symbolic way
A tantalizing idea. I want it too lol
Concept is from the late 80s iirc
pick up on small bugs first
i want to make exactly that
<@212618015651528704>, Stop cursing or I'm going to suck your bootyflakes!
There's this concept of The Programmer's apprentice. A virtual assistant type of bot, or oracle that helps you code by just describing. S...
gonads instead of monads
i have implemented monads myself, but still they're a bit elusive
screw monads, nobody needs them
probably you basically know the gist already
But I agree
would probably be easy for you then
Never learnt functional programming.
I'm stuck in C and Java-like ways, and Python.
the whole thing, from source code management to testing to how much you have to type
but really, programming needs to be simplified
yeah it's an option
Maybe some Decision / Random Forest thing too, they're way more interpretable
we just raise the level of convenience and abstraction
coding is talking to a machine, right
Make functions or programs
Symbolic regression then
i want code and natural language
i don't want papers either
Some recent papers mention that
BUT I DON'T WANT NEURONS
Well there are incentives to do neuro-symbolic stuff now it seems.
i don't really want to fight, but i want a chance
i hope my seemingly eternal fight against neural networks is coming to and end some day
Just came a new paper on a variant called PrescribedGAN or something that supposedly is easier to train, they mitigated the "mode collaps...
left the bot here and disappeared
sadly bootyflake succer's author is not responsive anymore
yeah but i meant it, booty
like normal NNs but even weirder and bitchier
GANs are so frankenstein-ish
yes, 3d models are good for image recognition
Or yes but it's in limbo
Someone is doing 3D-like "internal projection" of seen objects in 2D images. That inspires me a bit. Imagining an AI see something from a...
does the startup have a website yet?
nobody can talk to the NN and tell it what an earring is
the combination of utter brilliancy with really stupid mistakes
it's kinda funny
But ofc, once an object is found, they stop being helpful, most I've seen is segmenting object parts but no one has practical higher-leve...
but generally the images look fantastically real
nvidias face generator sometimes puts earrings in the cheek
like, actually counting objects
NN have real trouble with object identity
yes, they are good on the low levels, worse on the higher ones
i am imagining understandable algorithms for image recognition. like, the algorithm outputs something like "if i take this picture of a b...
The reason Convolutional NNs are so popular nowadays is they just worked well on that low-level filtering that is chewing pixels to find ...
Well there kinda are ways I think
as you said
your scenario is not an easy one though
i believe there are ways
completely without NN
i want to do new ways of image recognition so bad
Well we do object detection and recognition in literally the wild (outside in various seasons and locations), and the objects are often n...
i do it differently, the problem is not as big for me
oh yeah, if you do classic ML you need hardware
broke or not
my servers have not been offline a single day since 6 years or so
Yeah it is. Biggest hurdles are hardware constraints and training a good working model.
Just sticking to what you do and keeping the ship afloat for a year or two also helps as people will see you're still in operation. Tthus...
just from what you told me, ofc i don't actually know
sounds pretty doable though
and then it's easier sailing
so you're hoping that you can continue operations until the contracts are there
yeah, so the answer to his question is: from people who like to not get ripped off. done
adapt to the situation
also, recognize the person in front of you
A valid point. It's better to build a reputation and grow your network like that, than trying to play big fish from the very start.
the guy literally said "if you hate money how are you going to get any"
there would have been no deal if i'd had made them more expensive
let's see how wins the game in the end
i think he worships money, while i hate it
but i completely disagree
A game kindof
no you can't do that blablabla