Sungjae Cho

PhD parasite @Mila_Quebec | Neural intelligence ⇄ Machine intelligence | My Bio: CompSci + Math → AI & CogSci → neural TTS → CompNeuro → Neuro + AI

[Sungjae Cho @SungjaeAI]

Inside an AI data center
https://t.co/9GQnBm80eR

[Demis Hassabis @demishassabis]

Excited to collaborate with the Korea Ministry of Science and ICT (@msitmedia) to use AI to accelerate scientific discovery and to invest in Korea’s next generation of talent. Many thanks for hosting us @msitminister - look forward to working together!

[Google DeepMind @GoogleDeepMind]

A decade ago in Korea, AlphaGo showed AI’s potential.

Together with the Korean government, we’re now looking at how this technology can help accelerate scientific discovery and create new opportunities for economic growth across the region. 🇰🇷

Find out more → https://t.co/OKiI9e65aC

[Sungjae Cho @SungjaeAI]

Novel, secure physical memory devices for crypto storage are essential.

[Sungjae Cho @SungjaeAI]

Current NVIDIA looks more like a supercomputer manufacturer for AI, rather than just a chip maker for AI.

[Sungjae Cho @SungjaeAI]

Wireless bandwidth will be the next bottleneck of AI. Who is leading wireless technologies?

[Sungjae Cho @SungjaeAI]

A new kind of searching method will emerge soon.

[Sungjae Cho @SungjaeAI]

Computational specificity: CPU (scalar) < GPU (vector) < TPU (matrix) < NPU (device)

[Sungjae Cho @SungjaeAI]

We've never replaced our brains in the workforce scene. The future job market is likely not understandable by our old intuition.

[Sungjae Cho @SungjaeAI]

The success of NVIDIA implies a structural shift in computing from serial to parallel. Serial computing is limited by the heat and time issues. Parallel computing improves these problems and requires more memory space. GPU memory is on demand. Parallel readable disks may be made.

[Sungjae Cho @SungjaeAI]

The use case of LLMs is still the same as search: search for knowledge. If LLMs replace our tedious job to find products with a reasonable price, many Google users are willing to pay the effort of the search.

[Sungjae Cho @SungjaeAI]

Google is likely to be the winner of LLMs for general purposes.

[Sungjae Cho @SungjaeAI]

1. LLMs will be saturated. LLMs will keep evolving, but at a certain level, public users cannot recognize the quality. Data will be decisive. Google possesses myriads of search data, YouTube data, world data (Maps), and personal data (Gmail, Drive).

[Sungjae Cho @SungjaeAI]

@fluneai Let's see!

[Sungjae Cho @SungjaeAI]

Unbelievable! Finally, job done!

[Tottenham Hotspur @SpursOfficial]

Making history.

[Sungjae Cho @SungjaeAI]

A new theoretical neuroscience book!

[Kording Lab 🦖 @KordingLab]

Finally a new textbook for theoretical neuroscience. Xiao-Jing Wang wrote a nice book with very broad coverage of the field. Highly recommend. https://t.co/uIDmpHvPYO

[Sungjae Cho @SungjaeAI]

@LecoqJerome This link is not available!

[Sungjae Cho @SungjaeAI]

@WoongSSang Ah, I see. Thanks!

[Sungjae Cho @SungjaeAI]

I have posted an arXiv paper (https://t.co/5KbGskI8GQ) as a report, which guides the proofs of Lee and Seung (2000)'s multiplicative update algorithms to solve non-negative matrix factorization (NMF).

[Sungjae Cho @SungjaeAI]

@WoongSSang I should! Please send me the workshop!

[Sungjae Cho @SungjaeAI]

@WoongSSang No! I didn't know that. Yeah, he's on the KIAS homepage!