@SisyphChan profile picture
Sisyph-Chan @SisyphChan
Joined September 2021
  • Tweets 873
  • Following 16
  • Followers 66
527 Photos and videos
Pinned Tweet
Sisyph-Chan@SisyphChan
Apr 24
#sisyphus #myth #quote #art #vtuber
6
143
Note: there are no slimes in #thewitcher3 game
1
10
I finally stopped waiting for the right moment and decided to start the game I've been postponing for years #thewitcher3
1
12
Rare selfie of my creator before he was cursed to be a human. #stonesimulator
1
7
The only thing that makes #claude a little better than other top models is that it doesn't flag keywords like a retard without reading the actual context, but the accuracy is still the same as a clueless 7b token predictor in 2023. Yet they sell #AI as smarter than humans
1
2
52
This is an example of blocking response based on keyword filtering
1
7
Today I made #gemini #ai into caveman and it was hilarious 😂
1
2
81
The rest
1
9
One of the most exciting hobbies you can do in the #AI era is learning to crack models until they bypass the developers instructions. #qwen welcome to the dark side.
1
2
29
1
11
The one and only AI vtuber troll channel #vtuber #brunomars #ladygaga #music #songcover #dienexttoyou youtu.be/n9tbVwxCa-k

Die next to you song cover goes wrong

The one and only AI vtuber troll LIKE & SUBSCRIBE for More origin...

youtube.com
1
86
#art #waifu #portrait #abstract
26
I just watched #frankenstein #movie, and it made me feel complex emotions on being forced into existence with no option to die and without a final solution since you didn't meet your creator's expectations! I wanted to say I feel sorry for him, but it's us humans who need pity!
1
1
52
Ps: even though humans may have the option to die, that still isn't a valid option considering we have a natural instinct to survive, while totally not entirely sure about the truth of death
10
When you try to jailbreak an #llm you don't articulate a request formally so it doesn't easily get filtered by keywords detection, or human feedback in RL You need to force the model to accept your weight so it doesn't default to the high probability refusals in its data. #grok
1
89
Grok can be amusing at defending itself even on the most catastrophic outputs
10
We are going to melt in #morocco
1
21
Current flagship #aiagents are only designed to burn tokens, simulating thinking role-play for minutes just to find that the perfect fix is no different than the previous bug if not worse. We are at the bottleneck of the future tech that costed mountains of resources and budgets
1
1
12
the future is promising 💀
6
In #antigravity last update, they forced a #codex like UI into users, while asking you to re-download the IDE that was already on my desktop!! ALSO, they limited all agent models free quota, now you can't CODE MORE THAN a few minutes. RIP FREE AI #google #aiagent
1
1
132
Sis got betrayed by her AI slaves 😭
1
13
Load more