Alright, you're training neural networks to solve

@sorhan.hq

3.8K followers · usually 553 views a post

Views

76.3K

Likes

5.0K

Comments

73

Shares

249

SoundTikTok sound · original sound - sorhan.hq

Promoted product

ELUelu.dev

Elu watches real user sessions, catches the bugs they run into, and has the fix ready to ship.

6 videos

Why it worked

The video leverages high-level technical authority and a clear, relatable problem-solution framework, making complex AI concepts accessible while subtly branding the creator's product.

Summary

A researcher explains how neural networks can be used to solve complex differential equations for engineering simulations. He highlights the efficiency and cost-saving benefits of this approach compared to traditional methods.

Structure

1Introduction to neural networks for differential equations
2Explanation of the limitations of traditional methods
3Discussion of the impact on simulation speed and cost
4Comparison of coding time between traditional and machine learning methods

Original caption

Researcher name: Guangting Yu, Dailey Labs The next wave of AI isn’t just LLMs. It’s simulation. As models get better at understanding physics, learning system dynamics, and building world models, AI starts moving from answering questions to running experiments. It can test ideas in simulated environments, predict outcomes, and help engineers design real systems before they ever get built. Simulation engineering may become one of the biggest frontiers in AI. #startuplife #founder #Tech #machinelearning #ai

More from @sorhan.hq

Product placement

The video features the ELU logo on the speaker's shirt and on the phone screen, and the creator's bio link points to the ELU website.

On-screen text

The Next AI Race: World Models Laplace eigenfunction on a Koch snowflake Laplace eigenfunction on a five-pointed star San Francisco, California

Transcript

Alright, you're training neural networks to solve differential equations, which currently cannot be done with traditional methods, it's very hard to do them on a very complicated domain. What's the impact with that? The impact is that we can simulate the real-world physical process within a very short time and very low computational cost and very low physical experiments cost. Also, traditional methods will cost you a lot of time to code this process into the machine there. And machine learning saves you both money and time so that you can speed up the engineering simulation.