@yoavgo@roydanroy That's correct when you use a model you serve for yourself, but when you serve customers, there are potential clashes. Value alignment doesn't solve that though
@yoavgo@GSKenigsfield What's interesting for me in langchain is monitoring it in order to see what emerging patterns of usage we have around OpenAI's models
@yoavgo@GSKenigsfield Well, for people outside the domain it gives a very smooth experience. If you are a software engineer that wants to use "ai", you can use langchain as a set of abstractions over "ai"
@tunguz@Mascobot@OpenAI Well, these are not the correct parallels. The correct, albeit naive, comparison is with sentence transformers models: https://t.co/iNt27ZYZC6
@deliprao pymupdf is great but low level. Since the solution here includes a human like interpretation of the PDF, you will either have low level tools which their output roughly corresponds the PDF structure, or very sophisticated tools of which I recommend Amazon's Textract.
יותר מ-120,000 כבר הצטרפו לקריאה - לא לסגירת תאגיד השידור ״כאן״, לא להחרבת התקשורת הציבורית! אסור לתת לממשלה להשתיק את התקשורת ולהפוך אותה לשופר של השלטון. עזרו למאבק לצבור כוח, הוסיפו את שמכן.ם >>
https://t.co/L4HuEUvpBf via @zazim_org_il
@yoavgo@icmlconf I really think that since most of what we hear is U.S. centered, the plain argument of: we know the sciency stuff but our English is not good enough, is not heard as much as it should.