Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
"Capek gak sih tiap mau analisa dokumen harus ribet OCR dulu? Mana hasilnya sering berantakan lagi 🙄
Alibaba baru aja ngerilis Multimodal LLM yang bisa 'baca' PDF langsung kayak mata manusia.
No more text extraction, no more pre-processing. Langsung sat-set! 🚀
Ini intinya nih....
Zero OCR Workflow: Biasanya kan kita butuh step ekstraksi teks dulu sebelum masuk ke AI. Nah, model ini bypass itu semua.
Dia lihat dokumen sebagai visual input, jadi struktur tabel atau grafik yang ribet pun dia paham. Literally kayak lo lagi baca pake mata sendiri.
No Pre-processing Drama: Gak perlu capek-capek cleaning data atau benerin format teks yang kepotong. Langsung arahin aja ke PDF-nya, dia bakal langsung nangkep konteksnya secara holistic.
Context Aware: Karena dia multimodal, dia gak cuma baca teks, tapi juga ngerti placement gambar dan layouting. Jadi minim halusinasi gara-gara salah baca urutan paragraf.
Why It Matters?
Buat lo yang kerjaannya numpuk research paper, laporan keuangan, atau legal dokumen yang berlembar-lembar, ini life saver banget sih. Efficiency level: God mode! 📈
Gimana menurut lo? Bakal bikin tools OCR lama jadi obsolete gak nih? 🧐 Ciakakakak
Nih reponya: https://t.co/jW5SSrOt2T
@ScholarshipfPhd I posted on this a few days ago. He has had various Google Scholar accounts, which he deleted, but they keep popping up again. The one I posted about has over five million citations and an h-index of >1,700.
https://t.co/4lGLYNnRMo
as long as they can't disobey of the pre-programmed instruction or what their algo should be, we're not there yet.
human is agi, that is capable of being disobedience. and this is not algorithmic.
I made a magic quadrant about how to choose your tech stack
Seriously, DYOR: no one will do it for you. Try out stuff. See what works. Ignore reports from people who don't do the work and are pay-to-play anyway