Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
AI Space

Can AI Transform Space Propulsion? (fastcompany.com) 43

An anonymous reader shared this report from The Conversation: To make interplanetary travel faster, safer, and more efficient, scientists need breakthroughs in propulsion technology. Artificial intelligence is one type of technology that has begun to provide some of these necessary breakthroughs. We're a team of engineers and graduate students who are studying how AI in general, and a subset of AI called machine learning in particular, can transform spacecraft propulsion. From optimizing nuclear thermal engines to managing complex plasma confinement in fusion systems, AI is reshaping propulsion design and operations. It is quickly becoming an indispensable partner in humankind's journey to the stars...

Early nuclear thermal propulsion designs from the 1960s, such as those in NASA's NERVA program, used solid uranium fuel molded into prism-shaped blocks. Since then, engineers have explored alternative configurations — from beds of ceramic pebbles to grooved rings with intricate channels... [T]he more efficiently a reactor can transfer heat from the fuel to the hydrogen, the more thrust it generates. This area is where reinforcement learning has proved to be essential. Optimizing the geometry and heat flow between fuel and propellant is a complex problem, involving countless variables — from the material properties to the amount of hydrogen that flows across the reactor at any given moment. Reinforcement learning can analyze these design variations and identify configurations that maximize heat transfer.

Can AI Transform Space Propulsion?

Comments Filter:
  • by locater16 ( 2326718 ) on Sunday November 30, 2025 @05:56PM (#65826951)
    Pre 2022: Re-enforcement learning is cool, it's tricky but I bet it can do a lot of stuff
    Post 2022: LLMs will create GOD and you will BOW DOWN and BE THANKFUL
    Today: Hey there's this thing called re-enforcement learning that can do stuff LLMs can't, cool!
    • Aka parameter maxxers learn about over fitting.
    • by hey! ( 33014 ) on Sunday November 30, 2025 @08:28PM (#65827109) Homepage Journal

      Here's where the summary goes wrong:

      Artificial intelligence is one type of technology that has begun to provide some of these necessary breakthroughs.

      Artificial Intelligence is in fact many kinds of technologies. People conflate LLMs with the whole thing because its the first kind of AI that an average person with no technical knowledge could use after a fashion.

      But nobody is going to design a new rocket engine in ChatGPT. They're going to use some other kind of AI that work on problems on processes that the average person can't even conceive of -- like design optimization where there are potentially hundreds of parameters to tweak. Some of the underlying technology may have similarities -- like "neural nets" , which are just collections of mathematical matrices that encoded likelihoods underneath, not realistic models of biological neural systems. It shouldn't be surprising that a collection of matrices containing parameters describing weighted relations between features should have a wide variety of applications. That's just math; it's just sexier to call it "AI".

      • Artificial Intelligence is in fact many kinds of technologies. People conflate LLMs with the whole thing because its the first kind of AI that an average person with no technical knowledge could use after a fashion.

        But nobody is going to design a new rocket engine in ChatGPT. They're going to use some other kind of AI that work on problems on processes that the average person can't even conceive of -- like design optimization where there are potentially hundreds of parameters to tweak. Some of the underlying technology may have similarities -- like "neural nets" , which are just collections of mathematical matrices that encoded likelihoods underneath, not realistic models of biological neural systems. It shouldn't be surprising that a collection of matrices containing parameters describing weighted relations between features should have a wide variety of applications. That's just math; it's just sexier to call it "AI".

        This isn't true. Transformer based language models can be trained for specialized tasks having nothing to do with chatbots. A real world example of this is ESMFold. While ChatGPT is trained on human language one could train up a model from plasma dynamics data using similar underlying technology in order to provide useful generalizations for prediction and manipulation of plasma.

        • Re: (Score:2, Insightful)

          The operative part of your 2nd sentence is the 'can be'.

          It is premature to handwave this, particularly when so much of the market is made up of grifters who have already made impossible claims.

          • The operative part of your 2nd sentence is the 'can be'.

            It is premature to handwave this, particularly when so much of the market is made up of grifters who have already made impossible claims.

            No, the things I was referring to using LLMs to predict protein shapes and evolution of plasma have already been demonstrated.

        • You cannot bypass solving the Navier-Stokes equations with transformers. You will, of course, get some predicted flows with a black box model, and you can, if you choose, claim that prediction accuracy is close enough for 85% of the random samples from your test data, but that will not get you new propulsion physics.

          There's a serious danger today that a lot of the science which relies on simulated outcomes is subtly wrong in a way that cannot be rejected outright in peer review, but will take many years t

          • You cannot bypass solving the Navier-Stokes equations with transformers. You will, of course, get some predicted flows with a black box model, and you can, if you choose, claim that prediction accuracy is close enough for 85% of the random samples from your test data, but that will not get you new propulsion physics.

            Who is talking about "new propulsion physics" and what does this even mean? What I mentioned has already been demonstrated using LLMs.

            • What it means is very clear, especially since it's included in TFA's summary. Did you even bother to read the topic or did you just find an excuse to claim ML helps progress because AI?
        • by hey! ( 33014 )

          This isn't true. Transformer based language models can be trained for specialized tasks having nothing to do with chatbots.

          That's what I just said.

          • This isn't true. Transformer based language models can be trained for specialized tasks having nothing to do with chatbots.

            That's what I just said.

            No, what you said was the following "Artificial Intelligence is in fact many kinds of technologies. People conflate LLMs with the whole thing because its the first kind of AI that an average person with no technical knowledge could use after a fashion." Your statement is incorrect.

            • by hey! ( 33014 )

              That wasn't *all* I said, but it is apparently as far as you read. But let's stay there for now. You apparently disagree with this, whnich means that you think that LLMs are the only kind of AI that there is, and that language models can be trained to do things like design rocket engines.

  • Machine learning used to be the default term for computer assisted statistical model fitting. Pretty soon algorithms and data structures, math and farming will fall under the AI zealots umbrella as they seek relevance after LLMs fail to replace human thought.
  • by magamiako1 ( 1026318 ) on Sunday November 30, 2025 @06:18PM (#65826983)
    I know these wonks want you to believe that it's the case, but the reality is that AI cannot replace human thought and ingenuity in its current form. LLMs are fundamentally not capable of doing this--as their inputs are the apex of human thought.

    An LLM is a glorified search engine. It can be generally better at scoping out a stack overflow or Reddit post better than you can to find the relevant bits, but that's about it. That's all it can do. So in that it can take already *existing* knowledge that all of the LLM companies have illegally scraped from the internet and regurgitate to you for the time being without throwing ads in the mix is about the only thing it's useful for.

    Holy hell these people are on drugs.
    • Check TFS. LLMs aren't in there. Engineers were using AI to find better shapes for rocket nozzles years before anyone had ever heard of ChatGPT. Hell, I toyed with using genetic algorithms to make extruded parts better nearly twenty years ago.

      Don't fall for the marketing; there's more to AI than LLMs. Most of it actually works.

      • Read the article, the article itself contains literally zero substance. Was the post itself written by an LLM? LOL.

        Deep diving a little bit here:

        The following blurb from the article:
        "In regard to space propulsion, reinforcement learning generally falls into two categories: those that assist during the design phase – when engineers define mission needs and system capabilities – and those that support real-time operation once the spacecraft is in flight."

        Immediately followed up with:
        "Among
    • I know these wonks want you to believe that it's the case, but the reality is that AI cannot replace human thought and ingenuity in its current form. LLMs are fundamentally not capable of doing this--as their inputs are the apex of human thought.

      Nobody knows what LLMs are fundamentally capable of doing. Personally I think it is nuts for people to speak of human thought as something special when trivial algorithms executed by no mind or computer have accomplished feats (e.g. flora and fauna of the planet) greatly exceeding the sum total of all human efforts.

      For all anyone knows it is plausible to run LLMs in some sort of loop that rummages through latent space until it stumbles on useful solutions especially in situations where evaluation of object

      • Nobody knows what LLMs are fundamentally capable of doing. Personally I think it is nuts for people to speak of human thought as something special when trivial algorithms executed by no mind or computer have accomplished feats (e.g. flora and fauna of the planet) greatly exceeding the sum total of all human efforts.

        Well that's an assertion.

        We know exactly what they're capable of. They take string chunks and determine the probability of various other chunks, based on their related frequency in datasets they have ingested. They can be made to take in other data in similar ways to show similar relationships to the data they have ingested, even when the dataset being examined is noise.

        These patterns are not automatically meaningful.

      • Nobody knows what LLMs are fundamentally capable of doing.

        We know exactly what LLMs are capable of doing, at least as they exist today. How they might be used to deliberately or inadvertently influence people is a different, and open, question.

        There is an area where LLMs might significantly help research - and it might even be applicable to space flight propulsion, and that's bring in knowledge from other fields.

        For example, there could be some quirk about the way the squids and octopuses propel themselves

  • It depends (Score:3, Funny)

    by srmalloy ( 263556 ) on Sunday November 30, 2025 @06:26PM (#65826993) Homepage
    I'm not sure that it's possible to come up with a way to throw AI systems out the back of a spaceship fast enough to make them viable as a propulsion system, no matter how much better they would be used that way.
    • I'm not sure that it's possible to come up with a way to throw AI systems out the back of a spaceship fast enough to make them viable as a propulsion system, no matter how much better they would be used that way.

      Surely if you burn enough money fast enough, it can be used to generate some sort of thrust.

  • This is the plot of Voyager's "Timeless" as I recall.
  • With enough VC capital investments it will get it right at the 76th try.

  • A slop-designed rocket engine might explode violently enough to give another big piece of metal a shot at becoming the fastest human-made object.

  • by Gideon Fubar ( 833343 ) on Sunday November 30, 2025 @07:05PM (#65827027) Journal

    "I have a hammer. Please validate my belief that all your problems are nails and praise me for it."

  • by RightwingNutjob ( 1302813 ) on Sunday November 30, 2025 @07:20PM (#65827037)

    It'll tell you all about warp factors and quantum improbability drives, and cite all its sources, to boot.

  • With tiny teensy nuclear bombs, the peoples will just LOVE it.

  • by khchung ( 462899 ) on Sunday November 30, 2025 @08:58PM (#65827147) Journal

    Hands up, how many remembers all the blockchain articles asking "Can Blockchain Transform xxxx?"

    Has Blockchain transformed anything yet?

    AI is the new Blockchain, a huge bubble looking for problems to solve to justify the bubble.

    • by Elendil ( 11919 )

      Me: Can blockchains transform LLMs?
      ChatGPT: Yes, blockchains have the potential to transform large language models (LLMs) in several significant ways. While they aren't directly linked in a technological sense, combining the two can offer benefits that improve LLMs' capabilities, accessibility, and trustworthiness. Here are a few areas where blockchains could impact LLMs:[...]

      Yeah, right.

  • LLMs CANNOT innovate. They CANNOT have ideas. They CANNOT think. They can basically just search more efficiently.

    Stop projecting things on LLMs that they are not.

  • There are two ways:

    1) Take any of the AI tech bros, place then at the rear of your spacecraft and have them readout their quarterly forecast. The hot air alone will propel the spacecraft very competently across the vastness of space.

    2) Use machine learning and a 'digital twin' to engineer different ideas. This is entirely possible, and is happening in multiple fields, not just rocket propulsion. Buzzwords aside, it's essentially computer-aided design, with actual human engineers doing the heavy lifting, and

There are two ways to write error-free programs; only the third one works.

Working...