arkadiym.bsky.social
I love Mathematics, Computer Science, Art, Manga, Video games and many other things.
Machine Learning Engineer
Born & raised in Kyiv, Ukraine
56 posts
24 followers
121 following
Getting Started
Active Commenter
comment in response to
post
То навіть ескалацією важко сказати. Він лише зауважив, що Росія вже порушувала угоди і що війна триває з 2014-го. На ескалацію пішов Трампвенс.
comment in response to
post
I like your art
comment in response to
post
Marvel Rivals enforces massive Chinese censorship in the game and the developers seem to be fine with it.
Out of solidarity with Taiwan, I don't feel sorry for them.
comment in response to
post
Мені здається у монети є інша сторона. Якщо ми й так такі сильні, то навіщо нам допомогати?
Наскільки я розумію, ідея (вже провалена) була в тому, щоб не дати росіянам видавати себе за голубів миру, а нас видавати за недоговороздатніх.
comment in response to
post
Isn't that Hoarah Loux?
comment in response to
post
And the Skies are Blue.
comment in response to
post
Use the pirated version.
Now you have the moral authority to do so.
comment in response to
post
Zuckerberg already told bullshit like "We need more masculinity" long ago.
comment in response to
post
Well deserved.
comment in response to
post
Who is not collapsing then?
comment in response to
post
Де?????
У мене друг цікавиться.
comment in response to
post
There are things like weak-supervision, semi-supervision, hard mining and other techniques that allow to semi-automate data annotation using existing model or euristics.
Also things like Class Activation Map for vision.
They are widely used in industry.
comment in response to
post
Any suggestions for alternative directions?
I'm a ML Engineer and majority of ML jobs I see here are LLM.
comment in response to
post
Okay, but what is the alternative?
comment in response to
post
People are just unfamiliar with concept of "induced demand":
en.wikipedia.org/wiki/Induced...
The same way, people think that increasing road capacity will reduce congestion (on the contrary).
comment in response to
post
For the record, DeepSeek was trained on data generated via Llama and Qwen. So it is not the whole truth to say that it is 1000x cheaper.
comment in response to
post
Ви з Шольцем розмовляли?
comment in response to
post
Let's goooo.
Let twitter die!
comment in response to
post
It already has been happening for last 15 years at least.
Facebook/TikTok/Twitter/Instagram - all use AI to keep you engaged.
The easy way to keep people engaged is sensationalism, rage bait, etc.
And critical thinking requires deep concentration, but social networks are designed to distract.
comment in response to
post
Classic bargaining, not limited to rent.
The less they know, the less bargaining power they have.
comment in response to
post
It is clear as blue skies, USA is becoming tech-oligarchy (cyberpunk).
And they own all major social networks.
EU will have to ban twitter and TikTok (may be instagram/facebook as well) in order to defend their democracy from AfF and their kind, but that will inevitably lead to trade wars with USA
comment in response to
post
Life will be harder and times of prosperity will end, that's for sure.
Unless regular americans take their country back, but that's too good to be true.
comment in response to
post
It is clear as blue skies, USA is becoming tech-oligarchy (cyberpunk).
And they own all major social networks.
EU will have to ban twitter and TikTok (may be instagram/facebook as well) in order to defend their democracy from AfF and their kind, but that will inevitably lead to trade wars with USA
comment in response to
post
Boosting productivity of human with AI is more democratic, than replacing humans with fully autonomous AGI, as most AI labs currently aim to reach.
Famous economist, Daron Acemoglu, complained about it, so this is a good project.
comment in response to
post
Strong message!
Fits both game's topic and recent news =)
comment in response to
post
This is wrong. The reason behind black box structure of AI is because of The Bitter Lesson, not AGI dreams.
www.cs.utexas.edu/~eunsol/cour...
comment in response to
post
This is wrong. The reason behind black box structure of AI is because of The Bitter Lesson, not AGI dreams.
www.cs.utexas.edu/~eunsol/cour...
comment in response to
post
Вони дивують навіть більше, ніж росіяни.
Росіяни 25 років скотиніли маючи міцний фундамент з російського шовінізму, тоді як американці з ніхера фашизуються за декілька місяців.
Або я щось не знав про американців.
comment in response to
post
There was long before 450 report about tech monopolies destroying competition:
int.nyt.com/data/documen...
comment in response to
post
Of course, it is not like Iranian kamikadze drone landed 150 meters near me while I was staying at home. But it is propaganda and I have to grow up, because some westerner from rich safe country said so.
comment in response to
post
And that's was a huge mistake by US and it seems they still haven't acknowledged it.
comment in response to
post
Uyghurs is not an only concern here.
China helps Russia, North Korea, Iran and promotes dictatorships all over the world. Basically global axis of evil is clear here.l and being neutral between Good and Evil is immoral.
comment in response to
post
It doesn't equate morality, but the way to deal with political reality is a moral question.
comment in response to
post
Quite strange logic.
The incoming US president is immoral.
comment in response to
post
Actually, I agree. You may end up uncompetitive if you avoid China's market meanwhile competitors doon't. However, as far as I understand, ASML is a monopoly, so they don't have that risk here.
Moreover, making your business dependent on largest dictatorship is not a wisw decision long-term.
comment in response to
post
Remainging fully neutral and ignoring political reality is stupid and immoral way to do a business.
comment in response to
post
It is already a pretty dangerous propaganda tool used by bad actors like Russia.
comment in response to
post
python3.9 -m venv .venv
source .venv/bin/activate
comment in response to
post
comment in response to
post
Liberalism hasn't failed yet in Europe.
The way I see it, american problem is two-party winner-takes-all system.
While in Europe 20% of far-right radicals are just absorbed into mainstream and mitigated by it. See The Netherlands for example.
comment in response to
post
Are recommendation letters must be from people in Academia, or from industry is enough?
I'm asking because I finished my my masters 4 years ago and have been working in insdustry since.
comment in response to
post
They may use exactly the same technology (transformer), but output a single binary number.
Since they produce a single number, they are not "GenAI", but classification.
comment in response to
post
Oh, I forgot The Hardware Lottery which is also crucial. Your might avoid ideas that require more sophisticated engineering if you don't like it.
Julia programming language that boosted scientific machine learning is one example.