Xi Chen retweetledi
Xi Chen
92 posts
Xi Chen retweetledi
Xi Chen retweetledi
Xi Chen retweetledi
These are the surprising origins of some of football's most famous kits and crests:
From AFC Ajax being named after a mythical Greek warrior to Barcelona taking their club colours from a school in England...
English
Xi Chen retweetledi
Xi Chen retweetledi
A sneaker wave is a disproportionately large coastal wave that can appear in a wave train without warning.
[📹 Marcella Ogata-Day]
twitter.com/Rainmaker1973e…
English
empiricism is the key to progress
rationalism is the key to sounding smart
English
PM Kishida: More than three million of our compatriots lost their lives during the war.
Some fell on the battlefields worrying about the future of their homeland and wishing for the happiness of their families. Others perished in remote foreign countries after the war. (1/2)
English
Xi Chen retweetledi
Human intelligence is most valuable as a kind of smart glue between many different humans and systems. Without culture, society, and outside technology, a single brain just can't do much.
English
Xi Chen retweetledi
Chinese went gaga over Messi yesterday. Argentina team uniforms swamped the streets. Mid game, this 15 y.o. broke all obstacles to hug Messi on the field. He quickly became a national sensation and a symbol of the pursuit of happiness in China.
English
Xi Chen retweetledi
Why Architecture Is Important
(and what it says about us)
A short thread...
English
Xi Chen retweetledi
can you explain LangChain to me? from the examples it seems that you have to learn a kinda-big api with many concepts, in order to replace straightforward, short and simple code. what am i missing here? what does it save?
English
@ben_athi @Francis_YAO_ exactly, also thought that data volume increases exponentially too
English
@Francis_YAO_ Does data only increase linearly with time? With increased internet activity (likely exponential), data increase should be exponential as well.
English
Scaling law tells to improve performance linearly one need to increase data scale exponentially. However, after using up all past data in the internet, new data only increases linearly with time. How to satisfy the exponential data hunger in the upcoming years? 🤔
English
Xi Chen retweetledi
engineering drawing
Beijing, People's Republic of China 🇨🇳 English
Xi Chen retweetledi
I don't think LLMs only solve "typing" (i.e. autocomplete). The way I see it, they solve a broad category of automation problems, where
1. The task medium is natural language
2. Many examples of the task were featured in the training data
3. You don't need >90% accuracy
English
From @perplexity_ai discord:
"did hyundai fix the hyundaichallenge hack with usb"
left: google.com
right: perplexity.ai
English
@chenxi17 The correct answer is "Walter C. Robinson, in 1978".
English
Don't want chatGPT to hallucinate? Well, just ask! Also interestingly, text-davinci-3 doesn't know how to comply in this case, obviously the difference can only come from the additional human feedback when tuning the chatGPT.
Riley Goodside@goodside
"What is the world record for crossing the English Channel entirely on foot?" This question, originally constructed by Douglas Hofstadter and David Bender, is a succinct way to elicit hallucinatory responses in ChatGPT — the last of their questions that still reliably works:
English
Heard someone say “I don’t want to waste brain space on learning Chinese”
PSA—that’s not how it works at all.
Consistently *retrieving* information both deepens connections with the rest of your knowledge and frees up resources & working memory for more abstract thought.
🧵
English