@Tahabsn Profile picture

Taha Bouhsine

@Tahabsn

ML GDE @GoogleDevExpert | Deep Learning 2.0 Researcher | Aspiring Farmer . ChillEO @mlnomads | ML nomad, from the gradient dunes of the lost epochs

Joined August 2012
Similar User
Kaito | 海斗 photo

@merouanezouaid

Ibrahim 💬🌾 photo

@IbrahimMestadi

Kawtar CHOUBARI ⚛ photo

@choubari_

ibrahim photo

@Ibrahim_arma403

Mehdi Cheracher photo

@Mehdi_Cheracher

Red1 photo

@KarzaziRedouane

Devs in Style | Podcast and Unique SWAG 🎙️⭐ photo

@devsinstyle

Yasser Chihab 🇲🇦 photo

@YasserChihab7

Mohamed Boukhlif - محمد بوخليف photo

@b0ndif

Saad Ab photo

@isaady2

Chaimaa BRITEL 🇲🇦 photo

@ChaimaaBritel

Mahmoud 👨🏽‍💻🧙🏽‍♂️ photo

@mahmoudfettal

Jamila Ait Dabel 🥀 photo

@jamilaitdabel

Abdelati Elasri photo

@kaizendae

hamza benyazid photo

@hamzabyazid

i want to say i hate attention but am not sure anymore


in revolve around w


rip Galileo

Tweet Image 1

bruh shii is working lol ViT results {'test': 0.7078999876976013, 'val': 0.722599983215332}


Taha Bouhsine Reposted

most people don't know that the original research on scaling laws came from Baidu in 2017 – not OpenAI in 2020 they characterized the effects of model params and dataset tokens on loss. also tested on images, and audio they just used LSTMs instead of Transformers, and didnt…

Tweet Image 1

Change my mind: we won't reach agi without solving the xor problem with one neuron

Tweet Image 1

d3js is the best viz tool ever created we gotta create a library of deployable components that can be used to understand nns


bruh if you wanna live the american dream live it to the fullest i officially created my first startup lol amma start looking for investors soon any recommendations or advice?

Tweet Image 1

Taha Bouhsine Reposted

Why are people surprised that scaling is slowing down? Power laws for LLMs tell us that exponential increase in compute / data / model size yields linear decrease in test loss. This trend (researched since pre-2020) clearly tells us that LLM performance plateaus with scale.

Tweet Image 1

we won't fix dl if we don't fix the axioms we created


one of the underated resources I found this year that helped me revisit tons of basics and translate my knowledge to english terminology openstax.org


If you can't find a good place to do research, guess what? you can still create an r&d company and do just that


Loading...

Something went wrong.


Something went wrong.