The image above is one more attempt with an AI program to create a Martian habitat. As with the images in my last post, I asked the AI image program Freepik to respond to the following request: “Realist painting, surface of Mars, rocket and habitat in the foreground, ultra detailed.”
The image generated was better than either DALL-E 2 or Microsoft’s Image Creator (DALL-E 3) in terms of a semi-realistic and clearly planned habitat on Mars. Of course, an even more likely habitat would be below the surface, but I do not want to be too picky. But like the Microsoft program, Freepik added a moon that does not exist in the real world (or the real Mars, I should say).
I tried a different request to see it this would continue to happen. My request was: “Martian colony in the year 2400.” In this case, I was more satisfied with the result. In two of the four images, Freepik provided a viable habitat and did not display a moon. One of the images is displayed below. Unfortunately, the other two still has moons.
It will be hit or miss with many of these programs. You will need to know more than the program in terms of context. We are still early in the game and it is likely the AI programs will slowly remove these make-believe elements. But until then, don’t start using these images in any official publications or you will be spending a lot of time explaining yourself.
I was recently using OpenAI’s DALL-E 2 to test what it would provide if I was looking for examples of a Martian colony. I put in the following request: “Realist painting, surface of Mars, rocket and habitat in the foreground, ultra detailed.”
One of the images, shown above, clearly shows the rocket yet the habitat may be the ruins next to it. If so, this is an interesting perspective on the human endeavor. Maybe I should just assume the rocket is the habitat. Another image from DALL- E 2, shown as the next image below, is a little better with a spacecraft in the background and apparently a habitat in the foreground. Not necessarily inspirational, yet funtional.
After this I decided to see what Microsoft’s Image Creator (DALL-E 3) would produce. This was the opposite of the grainy, somewhat dismal images from the earlier version. As shown below, the updated program provided me with a colorful, sleek image that seemed more fantastical than real. Is that a Ferris wheel on the right of the structure? And since when did Mars have at least five moon orbiting it? At least the surface is reddish, but what is this craziness?
Microsoft has certainly upgraded the imagery, but what does it represent? Is it pure science fiction with little regard for the location being named. I would rather kids using AI software see the DALL-E 2 images as a starting point rather than the Microsoft Image Creator images representing nothing slightly realistic on a planet that cannot be Mars.
Maybe this is the hallucination problem cited with ChatGPT. I had not heard that the problem spread to AI imagery, but it makes sense. I will need to keep this in mind as I play with these new programs.
Image (Credit): Martian impact crater located on Vastitas Borealis with residual water ice. (ESA/DLR/FU Berlin (G. Neukum))
While the media is continually writing about artificial intelligence (AI) doing a kid’s homework and then destroying the world, maybe AI does have some utility in preparing us for a second world.
A study published in Nature Synthesis found that AI may be useful in finding ways to extract oxygen from water on Mars. Scientists from China used AI to analyze about 3.7 million formulae to identify a chemical that could break down water. A single researcher would have needed about 2,000 years to perform similar analysis.
While other programs are also looking at ways to extract oxygen from the Martian surface or atmosphere (such as NASA’s MOXIE), this study demonstrates the power of AI and its ability to potentially produce other needed material for survival on Mars.
Mr. Musk may not like AI, but he may want to have it on hand if he ever finds himself on Mars.
On Sept. 24, NASA’s OSIRIS-REx (Origins, Spectral Interpretation, Resource Identification and Security – Regolith Explorer) spacecraft will speed past Earth and – at precisely the right moment – jettison its sample capsule containing material from asteroid Bennu…To help celebrate this engineering and scientific achievement, the U.S. Postal Service issued a commemorative stamp featuring an artist’s impression of the sample capsule as it parachutes to Earth over its landing site on the Department of Defense’s Utah Test and Training Range…Although OSIRIS-REx has already had many scientific accomplishments, at its heart, the mission’s research goals circle around the sample delivery from Bennu. That influenced the Postal Service’s decision to select the capsule’s descent as the subject of the new stamp.
The James Webb Space Telescope (JWST) has found more than 1,000 galaxies mysteriously resembling our own Milky Way hiding out in the early universe. Shaped like warped vinyls and sporting delicate spiral arms, the Milky Way doppelgangers were found by JWST more than 10 billion years into the universe’s past — during a period when violent galactic mergers were thought to have made an abundance of such fragile galaxies impossible. Yet the disk galaxies are 10 times more common in the early universe than astronomers previously thought, new research reveals.
Scientists could soon use common lab technology along with sophisticated algorithms to answer one of the biggest questions in all of astronomy — are we alone in the universe? In new research published today in Proceedings of the National Academy of Sciences (PNAS), a team of scientists announced a novel technique that can take a sample of a material, feed it through a machine-learning algorithm, and find out if the material did — or didn’t — come from a living organism with 90 percent accuracy.
This past Sunday, 60 Minutes had a segment on Google and artificial intelligence (AI) called Exploring the human-like side of artificial intelligence at Google. Part of the discussion covered Google’s new AI chat bot Bard. And while the creative output of Bard was impressive, it was a little too creative when asked asked about economics. Bard recommended five books, but all five were nonexistent books. This is called “hallucination” in the AI realm.
I decided to try this myself, but ask for the five best books on black holes. This is what came back from Bard:
1. Black Holes: The BBC Guide by Andrew Hamilton. This book is a comprehensive introduction to black holes, covering their history, physics, and astrophysics. It is written in a clear and accessible style, and is perfect for anyone who wants to learn more about these fascinating objects. (link provided: https://www.goodreads.com/book/show/29537175-black-holes-the-reith-lectures)
3. A Brief History of Time by Stephen Hawking. This classic book is a must-read for anyone interested in black holes or cosmology. Hawking explains complex concepts in a clear and engaging way, and his book is full of insights into the nature of the universe. (link provided: https://www.amazon.com/Brief-History-Time-Stephen-Hawking/dp/0553380168)
4. Black Holes & Time Warps: Einstein’s Outrageous Legacy by Kip Thorne. This book is a more advanced treatment of black holes, but it is still written in a clear and accessible style. Thorne is one of the world’s leading experts on black holes, and his book provides a wealth of information about these fascinating objects. (link provided: https://www.amazon.com/Black-Holes-Time-Warps-Commonwealth/dp/0393312763)
5. The Little Book of Black Holes by Janna Levin. This book is a great introduction to black holes for people who are new to the subject. Levin writes in a clear and engaging style, and she uses humor and analogy to make the physics more accessible. (link provided: https://www.amazon.com/Little-Black-Holes-Science-Essentials/dp/0691163723)
Bard was even helpful enough to provide links to the books being cited.
After looking up these books, related authors, and links, here is what I found:
I could not find the first book by Andrew Hamilton, though he is a professor in the Department of Astrophysical and Planetary Sciences at Cornell University. Also, the link was to a different book.
The second, third, and fourth book references were accurate, as were the associated links.
The fifth book exists, but the authors are Steven S. Gubser and Frans Pretorius, not Janna Levin, who wrote other books on black holes.
So that gives you a 40 percent error rate on the 5 books listed. It does not inspire confidence. I am afraid to ask for the top 10 if only because it would take more time to correct this supposedly amazing chat bot.
And I do not really know if these are the top books in the field. Chat GPT gave me three of the five books listed, and also named the first book linked in Bard’s response, but attributed the lecture to Roger Penrose (the name of the lecture series) rather than Stephen Hawking.
Caution is highly recommended with these chat bots.
Our Awesome Universe 