Einstein: "The words or the language, as they are written or spoken, do not seem to play any role in my mechanism of thought." Are we really on the right path to AGI if we base it solely on language?
RESEARCH
-

Meta researchers replace normalization with Dynamic Tanh in Transformers
By
–
Imagine a Transformer model without normalization. This is exactly what's proposed in a new paper from Meta, NYU, MIT, and Princeton. The authors found that normalization layers can be replaced with something called Dynamic Tanh (DyT). It looks like this: DyT(x)=γ ∗ tanh(αx)+β.
-

ReAct: Synergizing Reasoning and Acting in Language Models
By
–
ReAct: Synergizing Reasoning and Acting in Language Models Yao et al.: https://
arxiv.org/abs/2210.03629 #ArtificialIntelligence #ChatGPT #ReinforcementLearning -
Recursive Self-Enhancement in Agentic AGI Alpha Agents
By
–
The ultimate use of an AGENTIC AGI ALPHA AGENT 👁️✨ is recursively self-enhancing #AGIALPHA pic.twitter.com/8yCDzRyeOv
— MONTREAL.AI (@Montreal_AI) 14 mars 2025The ultimate use of an AGENTIC AGI ALPHA AGENT is recursively self-enhancing #AGIALPHA
-
Google Releases Gemma 3 and Gemini 2.0 Flash
By
–
AI News from the past week that I've found interesting. Let me know about (and link me up to) anything that I missed this week so I can be sure to cover it in my breakdown video… (March 7th – 13th)
– Google Releases open-weight Gemma 3 model
– Google releases Gemini 2.0 Flash -
Megaprompt Development Through Human Evaluation and Experimentation
By
–
This one has just been a megaprompt with tons of riffing/experimenting and working with a team of human evaluators on the test cases. It's not the kind of thing that can be accurately evaluated by LLMs or programmatically, so it's been fairly labor intensive to get right. Also,
-
Manual Human Evaluation and Daily Performance Monitoring Process
By
–
Had to evaluate manually with humans due to the nature of the content and the high stakes. I've been working with a small team who provide regular reports throughout each day on how each version is performing. I get their feedback, make adjustments, do some tests myself, and
-

MIT’s Exo 2: High-Performance Programming Language Reduces Code Requirements
By
–
A new programming language from MIT could enable high-performance coding that can compete w/state-of-the-art libraries w/a few hundred lines of code, instead of tens or hundreds of thousands. Called "Exo 2," the approach enables reusable scheduling libraries external to
-

Nvidia Chief Scientist Bill Dally Fireside Chat at GTC
By
–
I'll be doing a fireside chat at GTC with Nvidia chief scientist Bill Dally Tuesday next week.
-

Krish Attending Nvidia GTC 2025 Event in San Jose
By
–
Finally Flying to San Jose(USA) to attend the Nvidia GTC 2025 Event. This event is an amazing opportunity for the entire developer community to know new trending research and development about Generative AI, Agentic AI, Data Science and many more. You can register for this event