It totally *is* a game of cat and mouse.
Which is why (1) there is no silver bullet to AI safety, (2) AI-fueled attacks have AI-powered countermeasures.
@ylecun
-
AI Safety: The Endless Game of Attack and Defense
By
–
-
AI Accelerating Scientific Progress and Discovery
By
–
The acceleration of scientific progress with the help of AI is super exciting.
-
Gurley Defends Open Source AI Platforms in Interview
By
–
Excellent interview of @bgurley in which he forcefully defend the idea of open source AI platforms.
-
UK Deputy PM Backs Open Source AI Platforms
By
–
Good news everyone: The UK Deputy Prime Minister understands the benefits of open source AI platforms.
-
Spell Checker Mishap: LLM Automatically Changed to Llama
By
–
s/Llama/LLM/
Funny how my spell checker automatically changes "LLM" into "Llama" behind my back -
ConvNext Architecture Comparison Across Vision Tasks
By
–
An extensive comparison of various architectures, various (pre-) training procedures, on various vision tasks.
TL;DR: use ConvNext. -
AI: Dual-edged sword for security and protection
By
–
If AI can help the bad guys, it can also help the good guys protect against the bad guys.
-
Search Engines vs LLMs: Accuracy and Regulatory Concerns
By
–
How long before regulators realize that search engines still produce more accurate information than LLMs?
They both use the same public data.
Search engines index it.
Llama summarizes it approximately. -
UK AI Safety Institute Launches Reliable Data Research Initiative
By
–
The field of AI safety is in dire need of reliable data .
The UK AI Safety Institute is poised to conduct studies that will hopefully bring hard data to a field that is currently rife with wild speculations and methodologically dubious studies. -
Llama Open Source Release Ethics Phil Zimmerman Comparison
By
–
I appreciate the compliment, but I'm not sure I deserve it.
Unlike Phil Zimmerman, who wrote the code for PGP and risked imprisonment by releasing it, I didn't write Llama, and I didn't risk imprisonment!