凯尔希是猫猫

4.5K posts

凯尔希是猫猫 banner
凯尔希是猫猫

凯尔希是猫猫

@2196F3

研究生物医学小分子和神经科学，Candidacy↑。 本体是香香软软没有攻击性的INTP猞猁！ 是explorer，coser和Les，想做大家的好朋友给人们带来快乐。 看到做得不好的实验会整个炸掉(然后亲自上手) 。😿 私信话题：源石会上调 IL6吗？

California, USA เข้าร่วม Aralık 2024
869 กำลังติดตาม1.1K ผู้ติดตาม
ทวีตที่ปักหมุด
凯尔希是猫猫
凯尔希是猫猫@2196F3·
*凯尔希 Standard Operating Procedure* **食 用 指 南** __本体__ 是INTP 香香软软的猞猁, 生日2k 年左右 __梦想__ 是成为能够解释任何事, 做到任何事的Researcher __是les__ 也想有一个能够互相依靠的人互相理头发, 躺在肩上述说彼此最浪漫的假想 __性格__ 通常比较稳定, 但是在看到做得坏坏的实验时会整个炸掉然后亲自上手 __STEM__ 学科都会一点点, 但把耗子放进CO2 气室仍然具有无法抵挡的吸引力; 有Candidacy; > "我们都会有一个精彩的Figure one; 共勉" __Perishable Product__ 超过21 摄氏度时会开始降解; 请使用Protease 处理, 90C 水浴然后4C 储存(safe stop; up to 1 day) __25, 8 月__ 会去ANYC 成为特蕾西娅殿下(1, 4) __明信片__ 在figure 4 > 我的双眼会看着你, Doctor
凯尔希是猫猫 tweet media凯尔希是猫猫 tweet media凯尔希是猫猫 tweet media凯尔希是猫猫 tweet media
中文
26
9
141
25.9K
凯尔希是猫猫
凯尔希是猫猫@2196F3·
@Qivshi1 Modern models are sophisticatedly balanced in such a way that they are expected to be reliable, knowledgeable, predictable and interpretable as needed. Unexpected erasure of weights during a dynamic update may render them less useful
English
1
0
0
23
Qivshi
Qivshi@Qivshi1·
@2196F3 well you have to overwrite things as you learn new facts
English
1
0
0
16
Qivshi
Qivshi@Qivshi1·
The best way to do LLM memory would be through something like a big LORA that continuously updates weights eg to learn a person's name well, you have to move beyond a little note (how LLM memory is currently done) and just learn the association automatically (from the weights)
English
3
0
8
556
凯尔希是猫猫
凯尔希是猫猫@2196F3·
@Qivshi1 This doesn't look like well figured out. Are you expecting the womb to generate an entirely new individual or it's a proxy term for something else?
English
0
0
0
20
Qivshi
Qivshi@Qivshi1·
These are the steps to solve aging 1. Artificial womb to seperate immune and reproductive system, this allows you to, 2. Downregulate cancer pathways related to reproduction, after which you can
English
4
1
3
190
Qivshi
Qivshi@Qivshi1·
it's kinda weird how we have aging figured out but people don't want to admit it, it's all senescence cancer trade offs
English
2
0
2
194
凯尔希是猫猫
凯尔希是猫猫@2196F3·
下次粥O 励志成为staff 😾 以及纪念日快乐
凯尔希是猫猫 tweet media
中文
1
0
3
164
凯尔希是猫猫
凯尔希是猫猫@2196F3·
感觉还是得做回神经退行的研究 植物的基因名一个都认不出来 #但是 #这个项目 #去漫展 #它 #请假快呀 🐱
凯尔希是猫猫 tweet media
中文
1
1
18
694
MonoFevc
MonoFevc@fexisnpc·
26连班的最后一天了 已经被班味儿腌入味儿了😭
中文
6
0
14
257
Aoi M
Aoi M@aoim33·
可以回复这条推文的都是可爱宝宝们呀
中文
34
1
118
8.5K
☁️🦁 Saren
☁️🦁 Saren@nekomatasaren·
It is possible to remove all text details from a photo and create an "AI slop" effect using VAE encode then decode.
☁️🦁 Saren tweet media☁️🦁 Saren tweet media
English
13
37
831
42.2K
凯尔希是猫猫
凯尔希是猫猫@2196F3·
@OdedRechavi Up and down up and down up and down ok good, enough, put in the ice, you doing great Balance, balance balance Stoooooop!!
English
0
0
0
301
Oded Rechavi
Oded Rechavi@OdedRechavi·
A PI who still hopes to spend time at the bench
English
6
8
219
46.7K

ค้นพบ

@Qivshi1 @fexisnpc @73935860Ae @MeowCatherina @elonmusk @BarackObama @taylorswift13 @cristiano