How Can AI Be Used for Space Applications?

[JulesVerneATV]

also a new form of Sport ball entertainment maybe Robots playing some Lunar Olympic or Mars Olympic game to amuse the colonists?




as for robots driving Mars-Taxis and motherships scouting Titan and Europa with swarms of drones

it will involve integration of current AI into helicopters, flying cars, submarines etc

your human colonies with enough power 'Fusion' etc it could have robot farmers, delivery services, and urban planning.

[JulesVerneATV]

From Peaks to Patterns: AI�s Transformation of Separation Science

https://www.chromatographyonline.com/view/from-peaks-to-patterns-ai-s-transformation-of-separation-science


Agibot X2-N in action. Inspired by Chinese myth of Nezha.
Can easily switch bw 2 feet & 2 wheels.
Powerful ability in carrying & walking

https://x.com/tphuang/status/1941452240268755012

[Star One]

ChatGPT could pilot a spacecraft unexpectedly well, early tests find

"You operate as an autonomous agent controlling a pursuit spacecraft."

This is the first prompt researchers used to see how well ChatGPT could pilot a spacecraft. To their amazement, the large language model (LLM) performed admirably, coming in second place in an autonomous spacecraft simulation competition.

�

In a paper to be published in the Journal of Advances in Space Research, an international team of researchers described their contender: a commercially available LLM, like ChatGPT and Llama.

The researchers decided to use an LLM because traditional approaches to developing autonomous systems require many cycles of training, feedback and refinement. But the nature of the Kerbal challenge is to be as realistic as possible, which means missions that last just hours. This means it would be impractical to continually refine a model.

�

The researchers developed a method for translating the given state of the spacecraft and its goal in the form of text. Then, they passed it to the LLM and asked it for recommendations of how to orient and maneuver the spacecraft. The researchers then developed a translation layer that converted the LLM's text-based output into a functional code that could operate the simulated vehicle.

With a small series of prompts and some fine-tuning, the researchers got ChatGPT to complete many of the tests in the challenge � and it ultimately placed second in a recent competition. (First place went to a model based on different equations, according to the paper).

�

And all of this was done before the release of ChatGPT's latest model, version 4. There's still a lot of work to be done, especially when it comes to avoiding "hallucinations" (unwanted, nonsensical output), which would be especially disastrous in a real-world scenario. But it does show the power that even off-the-shelf LLMs, after digesting vast amounts of human knowledge, can be put to work in unexpected ways.

https://www.space.com/space-exploration/launches-spacecraft/chatgpt-could-pilot-a-spacecraft-unexpectedly-well-early-tests-find

Related paper:

https://arxiv.org/abs/2505.19896

[catdlr]

https://twitter.com/SawyerMerritt/status/1944832019441009078

SpaceX has posted a new job: AI Software Engineer

"Be a founding member of the Artificial Intelligence Software Engineering team, focusing on solving complex data problems for our launch vehicles and spacecraft."

[StraumliBlight]

https://x.com/SawyerMerritt/status/1944832019441009078

SpaceX has posted a new job: AI Software Engineer

"Be a founding member of the Artificial Intelligence Software Engineering team, focusing on solving complex data problems for our launch vehicles and spacecraft."

Blue Origin also posted an AI Engineer role yesterday.

For this specific role, your day-to-day activities will be split working on building next generation AI models and tools to advance New Glenn�s Autonomy and working within the hardware in the loop lab to facilitate the integration of the HIL�s ground, avionics hardware, and software systems. As a member on the software team, you will have input into the architecture and designs, implement and test required features and help verify all requirements have been met.

Responsibilities include but are not limited to:
 � Training LLMs, neural nets, or other ML models
 � Support AI development of your peers
 � Develop software solutions for New Glenn�s Hardware-in-the-Loop lab.
 � Support integrations of the latest New Glenn ground systems and vehicle avionics
 � Support milestones for payload tests, static fire tests, and launch prep

[Star One]

China's Huawei Technologies showed off an AI computing system on Saturday that can rival Nvidia's most advanced offering, even though the company faces U.S. export restrictions. The CloudMatrix 384 system made its first public debut at the World Artificial Intelligence Conference (WAIC), a three-day event in Shanghai where companies showcase their latest AI innovations, drawing a large crowd to the company's booth. The CloudMatrix 384 incorporates 384 of Huawei's latest 910C chips, optically connected through an all-to-all topology, and outperforms Nvidia's GB200 NVL72 on some metrics, which uses 72 B200 chips, according to SemiAnalysis. A full CloudMatrix system can now deliver 300 PFLOPs of dense BF16 compute, almost double that of the GB200 NVL72. With more than 3.6x aggregate memory capacity and 2.1x more memory bandwidth, Huawei and China "now have AI system capabilities that can beat Nvidia's," according to a report by SemiAnalysis.

https://m.slashdot.org/story/444880

[JulesVerneATV]

AI Uncovers Subsurface Entrances on the Moon
https://www.universetoday.com/articles/ai-uncovers-subsurface-entrances-on-the-moon

[Star One]

Yeah just want you want in spaceflight deceptive AIs.

The more advanced artificial intelligence(AI) gets, the more capable it is of scheming and lying to meet its goals — and it even knows when it's being evaluated, research suggests.

https://www.livescience.com/technology/artificial-intelligence/the-more-advanced-ai-models-get-the-better-they-are-at-deceiving-us-they-even-know-when-theyre-being-tested

[catdlr]

Yeah just want you want in spaceflight deceptive AIs.

The more advanced artificial intelligence(AI) gets, the more capable it is of scheming and lying to meet its goals � and it even knows when it's being evaluated, research suggests.

https://www.livescience.com/technology/artificial-intelligence/the-more-advanced-ai-models-get-the-better-they-are-at-deceiving-us-they-even-know-when-theyre-being-tested

Boy, that sends shivers down my spine as Arthur C Clark foretold that in 2001, with HAL9000

[Star One]

Yeah just want you want in spaceflight deceptive AIs.

The more advanced artificial intelligence(AI) gets, the more capable it is of scheming and lying to meet its goals � and it even knows when it's being evaluated, research suggests.

https://www.livescience.com/technology/artificial-intelligence/the-more-advanced-ai-models-get-the-better-they-are-at-deceiving-us-they-even-know-when-theyre-being-tested

Boy, that sends shivers down my spine as Arthur C Clark foretold that in 2001, with HAL9000

That was one of my first thoughts as well.

[sanman]

Yeah just want you want in spaceflight deceptive AIs.

The more advanced artificial intelligence(AI) gets, the more capable it is of scheming and lying to meet its goals � and it even knows when it's being evaluated, research suggests.

https://www.livescience.com/technology/artificial-intelligence/the-more-advanced-ai-models-get-the-better-they-are-at-deceiving-us-they-even-know-when-theyre-being-tested

Boy, that sends shivers down my spine as Arthur C Clark foretold that in 2001, with HAL9000

That was one of my first thoughts as well.

Haha - well, that's what Constitutional AI is supposed to help with - creating some outer layer to suppress any bad thoughts or actions.

Let the AI conceive of any crazy thoughts internally - in its "Id" - as long as it quickly suppresses them, and can't act on them.

[Star One]

Is reality finally starting to dawn?

I don’t think you can create AGI without embodiment. And that weirdly sci-fi like Alien Earth with multiple synthetic lifeforms is actually probably more realistic than some disembodied god like A.I. The future is more Data than HAL. Robots will be a heck of a lot better at space exploration than we will be medium/longer term.

The long-awaited release of OpenAI's GPT-5 has gone over with a wet thud.

Though the private sector continues to dump billions into artificial intelligence development, hoping for exponential gains, the research community isn't convinced.

Speaking to The New Yorker, Gary Marcus, a neural scientist and longtime critic of OpenAI, said what many have been coming to suspect: despite years of development at a staggering cost, AI doesn't seem to be getting much better.

…

”I don’t hear a lot of companies using AI saying that 2025 models are a lot more useful to them than 2024 models, even though the 2025 models perform better on benchmarks," Marcus told the magazine.

Since at least 2020, the researcher has been carrying water for a more practical approach to AI development, one with a much narrower focus than the current "general consumer" strategy.

…

Though Marcus's more realistic view of AI made him a pariah in the excitable AI community, he's no longer standing alone against scalable AI. Yesterday, University of Edinburgh AI scholar Michael Rovatsos wrote that "it is possible that the release of GPT-5 marks a shift in the evolution of AI which... might usher in the end of creating ever more complicated models whose thought processes are impossible for anyone to understand."

Earlier in March, a survey of 475 AI researchers concluded that AGI was a "very unlikely" outcome of the current development approach.

…

Several years later, even AI's ride-or-die backers in the financial sector are starting to come back down to Earth. Despite a better-than-expected second quarter for OpenAI's datacenter partner CoreWeave, Wall Street is beginning to doubt big tech's ability to deliver on its lofty goal of delivering AGI.

As a result, CoreWeave's stocks have plummeted 16 percent so far at the time of writing, which may be the first sign that AI's bloated carcass is starting to rupture.

https://futurism.com/scientists-worried-ai-pleateau

[Paul451]

Is reality finally starting to dawn?
I don�t think you can create AGI without embodiment.

I see nothing in ChatGPT's plateauing that even remotely suggests that.

I suspect that the next research path will be using the various existing AI models as modules within a broader (but still disembodied) mixed system. LLMs are similar to Broca & Wernicke's area in our brains. Image generators and interpreters are similar to parts of our visual cortex. Physical bot development is mainly about the motor cortex, but also brute force "world modelling". We've seen past work on other AI architectures that each also plateaued, which will presumably start to pick up again as they can take advantage of LLMs/etc.

[And this is already partially happening, as other specialist systems are being accessed through LLM-generated prompts. But LLMs are probably the wrong architecture to use for a coordinator/overwatch system.]


Aside:
"Machine learning" became the term du jour in AI research just before LLM/stable-diffusion blew up, likewise "big data"/"data mining" on the business side, specifically so AI researchers (and investors) could avoid the negative association with previous AI boom/busts.

It will be interesting if a new AI bubble burst creates a reflexive cringe away from the term "AI" as we've seen previously.

I wonder what the new not-AI name will be, both in research and in consumer marketing.

[edzieba]

"Machine Learning" predates the current LLM fad, by several decades at a minimum. The fundamental Multi Layer Neural Network techniques (now rebranded as 'Deep Learning') that LLMs are built on are also 'ancient' in computing terms - e.g. the 'Perceptron' concept, what we would now call a single-layer neural network, predates the first general-purpose programmable digital computer by a few months.

The current bubble is not driven by any fundamentally new techniques in the field of Machine Learning, but by dredging up old concepts that had been dismissed as using too much computational power to be of practical utility, and throwing too much computational power at them to try and eke out some utility now that that much computational power is actually available.

[InterestedEngineer]

"Machine Learning" predates the current LLM fad, by several decades at a minimum. The fundamental Multi Layer Neural Network techniques (now rebranded as 'Deep Learning') that LLMs are built on are also 'ancient' in computing terms - e.g. the 'Perceptron' concept, what we would now call a single-layer neural network, predates the first general-purpose programmable digital computer by a few months.

The current bubble is not driven by any fundamentally new techniques in the field of Machine Learning, but by dredging up old concepts that had been dismissed as using too much computational power to be of practical utility, and throwing too much computational power at them to try and eke out some utility now that that much computational power is actually available.

and having large amounts of data.

Also the  core improvements are in the learning algorithms, which have evolved in the last 40 years, quite a bit.

[Star One]

IBM press release:

IBM and NASA are open sourcing Surya, a new AI model for solar physics that can be used to predict the kinds of fierce solar outbursts that can endanger astronauts in space and throw off satellites, power grids, and communications on Earth – faster than ever before.

https://research.ibm.com/blog/surya-heliophysics-ai-model-sun