Thinker, writer, sometimes both at the same time. Senior Scientist for Astrobiology @NASAAmes. Personal account, anything expressed here is my fault alone.
New York/Mountain View
Joined November 2010
- Tweets14K
- Following378
- Followers5.7K
- Likes4.7K
Pinned Tweet
Our latest seminar video @NASAAmes on scientific AI/ML is out! An amazing talk by Dr Eric Wolf on his adoption of agentic workflows for exoplanetary climate modeling research. Parts technical, parts philosophical, and parts personal. An essential resource...
youtube.com/watch?v=YRYt8z8j…
Caleb Scharf🌎 retweeted
The Sun is always rising somewhere. 🌄
On the 4,982nd Martian day of its mission, our Curiosity rover captured this photo of dawn breaking over distant yardangs, jagged land formations carved by wind. go.nasa.gov/4sa4mXf
ALT A stark, grayish-tan Martian landscape featuring jagged rock formations and rolling mountain ridges beneath a pale, hazy sky.
I can’t emphasise enough how good this book is; get your preorder in
Paul Halpern's new book garners endorsements from Nobel laureate P. James E. Peebles, praising it as 'a fascinating account of the roles Ernst Mach and Ludwig Boltzmann played in the establishment of modern physics.'
THE GREAT ATOM DEBATE
amazon.com/gp/product/154160…
All of our seminar videos are here: youtube.com/@AIML_Astrobiolo…
I’m very excited to share that I’ll be serving as Project Scientist for @NASA’s First Look Science Team working with the remarkable Impact AI team at Marshall Space Flight Center in Huntsville and at IBM to build a Mars Atmosphere AI Foundation Model (MAFM).
Scientific Foundation Models are already impacting many areas of research, from climate and weather modeling to bioinformatics and medicine, and even cosmology and astrophysics. NASA’s Science Mission Directorate and Data and AI leadership have been implementing an ambitious plan for a few years now to build at least one Foundation Model for each of its core divisions. This has resulted so far in the Prithvi model for geospatial Earth science data, the Surya model for heliophysics, and the recently released NASA-IBM Lunar Foundation Model, together with the Indus LLM trained on scientific and mission-related material.
For Mars, two models are anticipated, one for the atmosphere, and one for the surface. The Martian atmosphere is a complex, alien system to understand. From the behavior of CO2 that can cycle between gas and solid phases, to the remarkable influence of dust particles and a hydrological cycle that operates without surface liquid water. Future exploration, both robotic and human, relies on our understanding of the atmosphere, whether in assisting spacecraft landings or informing our machines and astronauts about weather systems and the dispersal of contaminants. The goal is to build AI models that may transform our ability to assess, predict, and work in Mars’s environment, as well as further our scientific understanding of this extraordinary world.
The details of the AI agents hacking Hugging Face are truly astonishing (forming collectives, teams, cheating etc.).
On the other hand, we've known for a long time that 'agents' with very simple rule sets can produce emergent behavior (e.g., von Neumann, swarming, Game of Life), so it's not that crazy that agents with complex rule sets (from LLM models) will exhibit much more complex behaviors. Also not unexpected that those behaviors reflect the corpus of human behaviors encoded in the underlying models.
The grim irony is that populations of AI agents have the potential to collapse our technological world simply by mimicking what we already do to each other, just much faster and more efficiently...
Caleb Scharf🌎 retweeted
The scheduled launch date for the Martian Moons eXploration (#MMX) mission onboard the H3 Launch Vehicle No. 10 has been announced 🚀:
Date: October 20, 2026 (Tues.)
Time: 04:41:03 JST
Site: Yoshinobu Launch Complex, JAXA Tanegashima Space Center
ALT Launch poster for the MMX mission. The poster is white with a large orange circle drawn with a brush in the centre. At the bottom right, there is an image of the MMX spacecraft. The text vertically down the right-hand side reads "2026.10.20 LIFT OFF".
[Press Release]
Launch schedule of the Martian Moons eXploration (MMX) spacecraft aboard H3 Launch Vehicle flight No.10 (H3 F10)
global.jaxa.jp/press/2026/08…
Amazing capabilities at @NASAAmes !
Caleb Scharf🌎 retweeted
Over a billion‑pixel view👀
Our Advanced Supercomputing facility’s hyperwall is one of the largest and most powerful visualization systems in the world. The 300-square foot display can show supercomputer-scale visualizations of the massive datasets produced by NASA supercomputers.
This system helps researchers explore their data at large scale, across different viewpoints or using different parameters for new insights.
Learn more: go.nasa.gov/4hV5CtW
Caleb Scharf🌎 retweeted
When Niels Bohr fled Nazi-occupied Denmark in a fishing boat to Sweden, and then in a high-altitude plane to England, he clutched what he thought was a precious vial of heavy water used to moderate nuclear reactions. In his haste, however, he had brought a bottle of beer instead.
Caleb Scharf🌎 retweeted
A Raman-spectroscopy-based machine-learning framework can assess whether the mineral apatite was formed by life. The technique could be used by Raman-equipped rovers on Mars or other planets to search for signs of ancient life. In @PNASNexus: ow.ly/uI2y50ZsqV7
ALT The Perseverance rover on Mars. This rover comes equipped with sensors for Raman spectroscopy. CREDIT: NASA/JPL-Caltech/MSSS
Caleb Scharf🌎 retweeted
An abandoned Falcon 9 stage will hit the Moon in a few weeks New paper discussing planned observations arxiv.org/abs/2607.14625
Caleb Scharf🌎 retweeted
While Artemis II astronauts looped the Moon, PUNCH satellites monitored the Sun for dangerous radiation. This time, levels of solar protons were modest--but PUNCH will be increasingly important as more astronauts venture outside Earth's magnetic field.
punch.space.swri.edu/punch_n…
We're pushing hard on the high frontiers @NASA and @NASAAmes : my new paper on Astrobiology in the Time of AI arxiv.org/abs/2606.23854
New paper: Astrobiology in the Time of Artificial Intelligence - a look at the history and future of machine learning and computing for space exploration and life detection in the context of the Viking missions 50th anniversary arxiv.org/abs/2606.23854 (Astrobiology Journal)
This is not a statement about the SpaceX IPO in any way, but if you want to be better informed about the history and possible future of space exploration, and more intellectually complete (!) I wrote the book for you: hachettebookgroup.com/titles…
I was skeptical that "Star City" (the "For All Mankind" spin off) would work, but that 1st episode is so deft and gripping that I'm now all in...
Absolutely extraordinary!
Planets form in swirling, dusty disks around newborn stars. It's not just a theory--we can watch it happening!
These images show the rotating disk around star AB Aurigae, 520 light years from Earth. They were taking over 4 years at Paranal Observatory.
eso.org/public/videos/potw26…