@konatapractique Извините, а вы можете перечислить, что конкретно предполагается делать своими руками и что нельзя поручать ИИ? Плюс, необходимо упомянуть, какие навыки развивать и какая от этого польза.
@MatinSenPai Anthropic doesn't care if you do illegal stuff, as long as you don't do what is really prohibited by internal rules implemented in the guardrails.
And you can only get banned for breaking these rules. The reason is that the definition of "legal" is geographic.
While everybody is prostrating before the Fable 5.2 that can do you a pretty decent techno demo of pre-trained games and calling it a breaktrough in AAA-game realm, lets remember THE REAL HERO: .KKRIEGER
https://t.co/tSTeFBB8EN
@bitforth It can be the other way around. Anthropics are trying to sell this creative to the general public in order to get a successful IPO. Therefore, they have pre-trained a very specific set of demos. For example, Grummz has found this
@1982Tender@chugunoz@rockamark "там специально под нейрослоп пакет написан" - тоже склоняюсь к этому выводу.
Обучили на ассетах, спецдвижке (https://t.co/KRp93MMZ9m) и надрочили делать демки
@Ragnavaldr21721@wriches It's very strange that all these "one-shot prompt games" have been started by Anthropic engineers themselves before the IPO and has very similar approach. Looks like it has been specifically trained to provide these result.
@HardWarezz@MameliFilippo Agree. All these assets are look pretty stock — and I have seen a lot of Fable output that nothing won't convince me that this is "single prompt" or even dozen prompts. It's too early for that. The best guess — it's pretty complex workflow together with an access to asset store
@vibecod3r у бро полно идей. он профессионал, если вы почитаете его аккаунт. просто он решил похайпить. а так там нет нихуя простого - он как-то засветил полностью свой workflow по изготовлению ролика про Теслу. почитайте, интересно. Ну и в этой демке примерно дохуя человеческого времени