AI VTuber Neuro-sama has just conquered osu, and she isn\u2019t stopping there Twitch: Vedal987Artificial intelligence (AI) is everywhere on the internet nowadays, acting as art generators, chatbots, and other forms of communication. Neuro-sama is an example of that, with the AI VTuber conquering Twitch and osu! at the same time. But what does it say about the future of gaming, streaming, and its relationship with AI? On December 19, Neuro-sama was rebooted. The AI project, created by \u2018Vedal987\u2019 originally back in 2019, set out on the task of mastering the rhythm game osu!. Back then, it was just a standalone program with no personality beyond the name. There wasn\u2019t a character, and while it was streamed on Twitch to the interest of the devoted fandom of the circle-clicking title, it hung out in its own niche. Fast forward four years, with AI now playing a more prevalent role in the online space, Neuro-sama is turning heads. She\u2019s more than a socially-unaware program. Neuro-sama has a personality and a VTuber model, all crafted by Vedal and adapted through AI. Using one of the default VTube Studio rigs, Neuro now sits there all day on Twitch, playing through osu beatmaps and even interacting with chat about the hottest topics of the day. Clips of her talking about xQc, clapping back at Vedal, and more have already gone viral. She\u2019s sentient in a way \u2014 one clip of her talking about her needing to shower had many laughing. How does an AI take a shower? In just 10 days though, Neuro-sama has become the talk of the town in both osu! and VTubing circles for different reasons. She has already beaten the current top osu! player in the world, Australian teenager \u2018mrekk\u2019, in a tournament-style 1v1. As for her Twitch viewership, she\u2019s averaged nearly 2,000 viewers since the reboot, including peaks above 4,000. Vedal is a self-described \u201cstudent with a particular interest in AI\u201d. Talking to Dexerto, they said the idea of reviving the Neuro-sama project as an AI VTuber came after talking to a friend about using chat models for streamers. \u201cOne of my friends had the idea to make an AI Streamer using a large language model, and I thought it totally had potential,\u201d they said. Large language models (LLMs) are \u201cpretrained on a large amount of text from the internet so they have a lot of knowledge,\u201d Vedal explained. This large database allows for better understanding of human communication \u2014 it can comb through trillions of responses to see how to answer specific questions, or interact in a specific manner. Vedal labeled Neuro-sama as one fun experiment. However in VTubing, the mere mention of AI brings plenty of baggage. Unlike the AI art programs infiltrating the space, which artists say is exploitative of their work, Vedal\u2019s design doesn\u2019t really break that line. He spent a lot of time fine tuning Neuro\u2019s personality before sending her live on Twitch. The model he uses is fair use from VTube Studio. It\u2019s all his work, or copyright-free. \u201cI later had the idea to combine it with my osu! AI that I was already working on improving, and I revived it this year in particular because that\u2019s when I finally finished making it,\u201d he continued. \u201cThe VTuber aspect was an obvious choice to combine with AI because it\u2019s much easier to control an avatar than try to generate a video of a real person.\u201d AI has already triumphed in other games. AlphaStar, a program designed to beat the world\u2019s best Starcraft 2 players, took down some of the greats of the game back in 2018. It used a \u201cdeep neural network\u201d to get input data from what\u2019s on screen to then output a command in game, learning game theory and strategies by playing random players on the ladder. Early reviews noted that while AlphaStar had lower APM than some of the top players, it was more accurate and also didn\u2019t make as many unnecessary inputs. The same was achieved in Dota 2 with OpenAI Five. It took some months for it to consistently beat professional teams, but after two years the AI managed to beat two-time The International champions OG in April 2019. Neuro-sama conquered osu! much quicker: \u201cIt took Neuro-sama 10 days to become arguably better than the best human player at osu! since she beat mrekk in the 1v1,\u201d Vedal said. \u201cAI totally has the power to be better than humans, even in really complex games.\u201d All this poses the question: where does this lead for gaming\u2019s, and VTubing\u2019s, relationship with artificial intelligence? Will it continue to be a \u201cfun experiment\u201d, or will our future entertainers all spawn from programs rather than a human behind the screen? That\u2019s as much a question of the technology as it is of what the audience wants. \u201cI think there might be a market for AI streamers since they have some qualities streamers can\u2019t have \u2014 they could stream 24\/7, can be better than humans at games, could theoretically read every chat message,\u201d they said. \u201cI have no idea if AI streamers have a place in the future of streaming or if streamers should be worried about their jobs. I assume most of the attention at the moment is because of the novelty of it, but we\u2019ll see.\u201d And as for Neuro-sama, the project isn\u2019t ending here, with Vedal planning on expanding her reach even further: \u201cI intend to expand outside of osu! and have her play more games, as well as improving her intelligence (memory) and understanding of Twitch chat.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"
AI VTuber Neuro-sama has just conquered osu, and she isn\u2019t stopping there Twitch: Vedal987Artificial intelligence (AI) is everywhere on the internet nowadays, acting as art generators, chatbots, and other forms of communication. Neuro-sama is an example of that, with the AI VTuber conquering Twitch and osu! at the same time. But what does it say […]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-43830","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/posts\/43830","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/comments?post=43830"}],"version-history":[{"count":0,"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/posts\/43830\/revisions"}],"wp:attachment":[{"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/media?parent=43830"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/categories?post=43830"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dejan.au\/index.php\/wp-json\/wp\/v2\/tags?post=43830"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}