AI designs rocket engine.

[jariya1390]

AI designs rocket engine :

Is that true?Please can any experts clarify? Are even scientists replaced by AI?

 

[whispous]

3 minutes ago, jariya1390 said:

Are even scientists replaced by AI?

Please first understand that AI is incapable of coming up with anything new.

It is possible for very specifically trained AI systems to iterate using known knowledge.

Only humans come up with new knowledge.

 

I have no doubt that AI systems are currently useful tools for rapid iteration and testing.

[Millios]

12 minutes ago, jariya1390 said:

AI designs rocket engine :

Is that true?Please can any experts clarify? Are even scientists replaced by AI?

1:42 in the vid, it answers your question.

"Is this AI? No" 

If it was AI it would be a kit bash of other designs 

It is an algorithm the people who made it keep working on to improve their designs.

AKA it acts as a more capable simulation 

 

[GarlicDeliverySystem]

10 minutes ago, jariya1390 said:

Are even scientists replaced by AI?

AI is a very wide and vague term. And yes, to a degree scientists are replaced, but at the same time more scientists come along in new fields. Kinda like how calculators (the people) were replaced by calculators (the machines). Didn't mean that mathematicians were out of jobs, and it created a whole new section of "something with IT".

 

Anyway, I scrolled through the video and this seems to be mostly a more fancy version of an optimization algorithm. Kinda like AlphaFold but probably 'simpler' and with different architecture/parameters. Plenty of machine learning models are already being used in science, to the point that they have become a science of themselves.

[Zando_]

20 minutes ago, jariya1390 said:

Is that true?Please can any experts clarify? Are even scientists replaced by AI?

It's Machine Learning, folks have been building ML algorithms to do this sort of thing for years, long before the current "generative AI" LLMs were handed to the public. Here's the Czinger from back in 2020ish:

Which yes, is unfortunately over 5 years ago now. You'll notice the structural parts have a similar organic looking geometric design, because this is an older version of the same sort of tool, and then (same as that rocket engine) the parts are printed using laser 3D printers. Humans do like 99% of the work and the algorithm helps them figure out what's the strongest part with the least possible amount of material (and thus weight). 

[manikyath]

a rocket engine is a chamber in which fuel explosively combusts continuously, and gets pushed trough a hole.

 

beyond that it's taking notes of all variables that exist, and the ranges they can exist in, then finding the optimal mixture of those variables to make an engine best suitable for your usecase...

 

as stated above me, they say in the video "this is not AI", and you dont need "AI" for this. nor can AI "replace" the scientists that made it. basicly what they do is "rapid iterative design" in a software simulation to get a model that works according to the simulation, and then the software pushes that design to a 3D printer to test in the real world.

 

it's a means of making your scientists work faster and more efficient, instead of having people stuck at a drawing board.

 

basicly, it's like an interior designer using 3D design software instad of drawing your interior on a piece of paper.

[05032-Mendicant-Bias]

28 minutes ago, jariya1390 said:

Is that true?Please can any experts clarify? Are even scientists replaced by AI?

The idea an AI could replace a scientist is beyond ludicrous.

 

The AI you think about is a marketing term that says nothing, you might as well said mammals are causing climate change, which while technically true, is misleading.

 

That's an advanced generative design tool. They are incredibly narrow, and incredibly powerful. Superhuman in performance, if only in a tiny domain.

 

Alpha Fold did millions of years worth of folding, and is helping design vaccines and redesigns proteins.

 

There are also generative AI mechanical design tool that make really pleasing organic shapes with extreme performance (and cost, since they require 3D printing with bleeding edge materials.).

 

It is by far, the most value adding application where such algorithms can be deployed. I wouldn't be surprised if Alpha Fold can redesign a RUBISCO enzyme that is multilpes more efficient and can vastly improve on plant abysmal solar efficiency and make a real dent in fixing carbon and feeding the world.

[twin_savage]

5 hours ago, 05032-Mendicant-Bias said:

There are also generative AI mechanical design tool that make really pleasing organic shapes with extreme performance (and cost, since they require 3D printing with bleeding edge materials.).

I think generative AI is a misclassification of these tools. All the tools in this realm (including what was used in the OP video) I've seen fall under the category of topological optimization tools that required physics based simulation feedback; Now it's possible to use a DNN to try and interpolate simulation results across different geometry generations, but a PCE is likely to outperform a neural network and give you an error function which is invaluable.

 

6 hours ago, 05032-Mendicant-Bias said:

It is by far, the most value adding application where such algorithms can be deployed. I wouldn't be surprised if Alpha Fold can redesign a RUBISCO enzyme that is multilpes more efficient and can vastly improve on plant abysmal solar efficiency and make a real dent in fixing carbon and feeding the world.

The problem with alphafold is that it is never going to come up with any unique structures that it's training data didn't "imply" existed; only physics-based simulation can do this outside of testing the real thing.

IMO this limits it's utility because it's only going to find the "low hanging fruit" of solutions and not something revolutionary. I'm not advocating discarding it, it can still do good, but it should be viewed under the lens of "AI slop" that we view the AI generated pictures, music and videos with.

 

 

[wasab]

Yes, AI is replacing scientist 

[Mark Kaine]

On 12/20/2025 at 8:18 PM, whispous said:

Please first understand that AI is incapable of coming up with anything new.

Yeah that's the problem but that doesn't stop it from trying anyways, what could go wrong!

[wasab]

41 minutes ago, Mark Kaine said:

Yeah that's the problem but that doesn't stop it from trying anyways, what could go wrong!

AI can come up with new stuffs. That's is what the word "generative" before AI mean. That said, don't expect it to come up with ground breaking scientific discovery. 

[jariya1390]

I went and forgot about this post. And now its completely answered. Thanks everyone for the replies. How do you close this topic? IDK.

Edit:

I am marking this as the solution because you need to read all the above comments for a complete answer. Or below depending where you read this.