[ overboard / sfw / alt / cytube] [ leftypol / b / WRK / hobby / tech / edu / ga / ent / 777 / posad / i / a / R9K / dead ] [ meta ]

/tech/ - Technology

"Technology reveals the active relation of man to nature"
Password (For file deletion.)

Matrix   IRC Chat   Mumble   Telegram   Discord

File: 1608525861146.jpg ( 140.46 KB , 1641x807 , 574b6fa77f900628edcc83a51c….jpg )




What is this?


Proof that randomly changing shit is as effective as so-called "data scientists". """"Machine learning"""" is a pseudoscience.


Think you're gonna have to elaborate a bit on this, anon.
Are you implying that your script btfo not only Cockshott but computer science as a whole? Excuse me for saying this but if that's the case then you might suffer from chronic hubris unfortunately.

Nice try, FBI


I am not OP. I have no idea what this has to do with Cockshott. I did not write the script either.

It's some variation on the classic genetic algorithm, this time applied to machine learning algorithms. There are plenty of fields in computer science that are rigorous, but "machine learning" is not one of them. Most of the work there is based on unfounded guesses and "data scientists" never seem to be able to explain why certain models work better for certain problems than others, or why the constants were chosen, etc. They just randomly change shit until they get the result they wanted. Which is coincidentally roughly the same thing the linked script is doing.


cockshott and google-research, OP?
nice try, CIA.


>they used ML methods to generate solutions for ML problems so ML is bullshit


Holy shit what the fuck. Github in its new form; following the Microsoft purchase, I can see.


Novel neuroevolution btfo's every deep learning meme


Eh. So?
It makes total sense that you can make a machine that uses machine learning to get better making machine learning settings?

Now do it one layer deeper using the machine learned program itself and see if we get a singularity.


Well yes given enough time it makes sense it would be better but deep learning is much faster for a specific subset using current technology.


so? There are plenty of times that engineering leapfrogs science and ML is one of them.

Unique IPs: 2

[Return][Catalog][Top][Home][Post a Reply]
Delete Post [ ]
[ overboard / sfw / alt / cytube] [ leftypol / b / WRK / hobby / tech / edu / ga / ent / 777 / posad / i / a / R9K / dead ] [ meta ]