AI-generated word salad: Trilobite ≈ Oyster

Wired today had an interesting article on software that can create “fake text news”, released by two coders who created it on the cheap after an initial version had been deemed too dangerous to release (https://www.wired.com/story/dangerous-ai-open-source/).  The story had a link to another web program which will create AI driven text (https://talktotransformer.com/).

This I had to try.  My prompt is in bold below, and the program delivered text “completes” my thought.  I was most impressed that the AI brought in trilobites, my favorites animal after the heliocplacoids, but the luster wore off when the trilobites were compared to oysters (my boss’s favorite animal).  Overall the AI produced a great word salad—the Archean is about 100 times older than what the AI says, the middle Paleolithic about 4-10 times older, and so on.  Minor quibbles.  But with politicians now content with speaking word salads with little regard for the facts, AI can probably produce text that puts the politicians to shame.  We won’t need speech writers any more.

Fossils represent one of the most important and significant discoveries of mankind," it said.

Called Eudyptes atrodon and from the Greek for "great atelier," fossils were discovered in the Upper Archean of the Nile, an estimated 30 million years ago. Some of the fossils were so large it was thought they came from a meteorite. They had been so well preserved they were nicknamed "dinosaur eggs," though they were later found to have no skeleton, and their skin was pale, gray and even "blue" in some places. Eudyptes was more than 80 feet (25 m) long and weighed up to 600 tons.

Fossils of all kinds were discovered in the region, including ammonites, sponges, sharks, a trilobite (like an oyster) and a bird skeleton.

The find could have significant implications for the Middle Paleolithic of the area, the researchers said.

"The Middle Paleolithic probably took place between 11,500 and 9,500 cal B.P. (Before Present

--thanks to https://talktotransformer.com/ for completing my thought so elegantly.

My GPS on 3inch imagery

It might be hard to believe, but you can get better imagery than Google provides.  USGS, as part of the National Map Program, has imagery down to 3 inch resolution if you live in an urban area (the wilds of Nevada, like the Sheep Range, might be another story, and it could be worse there if a big land grab withdraws much of Desert National Wildlife Range from public access, even worse than wilderness status which just meant you had to walk a long way).

This is one day's drive, with the 3 inch imagery at 50% resolution (top) and 100% resolution (bottom).  Note that you don't see much area, the price of the high spatial resolution.  The vectors show the morning commute (red) and evening (green), with the length showing the speed and the arrow pointing in the direction of travel.  There is a point every second, and on the green you can see where I had to stop before making a turn.

This area has two roads, one NE-SW, and the other NW-SE. 

Problems could be in the imagery (note that we see the sides of buildings, so there is some geometric distortion), or in the GPS.  The GPS appears to be on the correct sides of the road on the NW-SE road, but is almost entirely in one late on the other road, even though I was always on the correct side of the double yellow line which shows up well at 100% zoom. 

Stuck in traffic

It's too hot to walk, so I am driving more than I would like, but still tracking my travels.  And in doing so, I can see all the little delays.

This is a recent days travel, red in the morning, and green in the afternoon.  On the west side, there is normal city cruising with widely spaced points. Heading southwest, the logjam.  There is construction, with a lane blocked, and coming home this afternoon, firetrucks which left me sitting through an entire cycle of the lights.  As a testament to the GPS, note the red and green stay consistently on the correct side of the road.

The location where I basically sat in place, for 330 seconds in the morning, and  134 seconds in the afternoon.  I must add, though, that this is not nearly as annoying as longer drives--out last trip back from the airport took three hours instead of just one, and today the drive only took 10 minutes instead of 5.  But I'm hoping better weather is coming, and I can do a 25 minute walk instead of a 15 minute drive and walk.  It's the advantage of living is a small city.

House wrens

In June we had house wrens take up residence in a ceramic bird hourse in the back year.  Previously the feeder had been hidden in trees on the side of the yard, but this year it went onto a branch in the apricot tree in the center of year.  That tree is happy and healthy, but no fruit has survived to ripeness until two barely made it this year.  It turns out the wrens only live in the bird house for a few weeks, but we got some great pictures.

I would like to say the pictures came from skillful waiting and use of the long telephoto, but those shots aren't worth sharing.  So I went with the big data technology solution.  The old cell phone is on a tripod with bendy legs, attached to an artist's tripod right below the next, and it took a photo every second for a half hour.  Then we just sift through the photos and pick the best.  Much higher ratio of good pictures compared to the attempts to record the foxes.

Which DEM to use for storm surge

 A few days ago I mentioned my storm surge flooding model of Annapolis  and wanted to use the new lidar data.

Soccer fields and seawall.  This shows the four DEMs, two each from 2011 and 2017.  The DTM is the "bare earth", and removes water, vegetation, fences, cars, and other man-made features.  In 2011 it has many holes where there were no returns on the ground, which could be smoothly interpreted over.  The DSM is a surface model, and includes evertyhing the lidar beam encountered.  Note the seawall, chain link fences, and a few automotbiles appear in 2017, which has much more detail than the 2011 data.

Cross section from the soccer field to the river.  In red, the chain link fence appears at about 8 m on the horizontal axis.  The fence barely shows on the 2011 DSM; the slight rise at 11 m might be a parked car.  The 2017 DSM shows the seawall at 24 m, about three quarters of a meter high.  The seawall barely appears on the 2011 DSM.  Both DTMs show only a gently sloping surface.

This is the seawall; note the drain in the center of the image.  The drain would allow water to flood the soccer fields if the water level rose above that level (as would the storm drains).  Should the flooing model assume the seawall will provide protection, or use the DEM?  Additionally the chain link fence appears as a wall in the DSM, how should it be handled?  This is an artefact for the DEM structure; similarly, bridges acts as dams in a DEM flooding model.  The water would probably flow on the road around the soccer fields, but the fields themselves might appear to be an unflooded region.

Dragonflies at Blackwater

We recently visited Blackwater National Wildlife Refuge.  I have always been partial to the US Fish & Wildlife Service, after several delightful summers in another of their installations.

While "bugs" were not the focus of the trip (and we encountered some nasty biting one in the woods), they stand still better for pictures and take some really nice photos.

 

Where did I go this summer?

Since I got the Garmin Foreunner, this is a map showing where I've been.  At this scale no anomalies appear in the reported GPS positions.

You can see:

1. Nature walks downtown to photograph ducks, snakes, ospreys, herons, and flowers.
2. Walks to check out construction at the new library.
3. Trips to the grocery store.
4. Trips to visit assisted living.
5. Trips to the temporary library, a.k.a. the Mall.
6. Trips to the car dealer and eye doctor.
7. Starts of trips on the interstate that left the immediate area.

Caveats

1. No assurance I remembered to turn the gizmo on every time.
2. At this scale the new lidar is not ideal as the base map.  It you want its true glory, so you only see a postage stamp sized area at one time.
3. You can't see the areas where we go repeatedly; you would need a heat map for that.

New lidar of the home area

Lidar provides one of the best mapping tools currently available.  Until this week, the best data for my home region had less than one point per square meter, and was acquired in 2011.  I just found that the state now has data with at least 10 times better point density, and which was acquired in 2017.

The 2011 survey.

The 2017 survey.

Selected points to highlight the improvements.  (A) you can now see the tombstones in the cemetery; (B) the footbridge has much improved resolution; (C) the parking garage is under construction (maybe caught during a stop work order?); (D) Hopper Hall has begun constuction, removing a parking lot; 2017 was a good time to be in Morocco and not have to deal with the parking woes); and (E) the cars in the remaining parking lot show much better detail than in the 2011 survey.

Difference map between 2011 and 2017.  In magenta points were 5 or more meters tallker in 2017; from the shapes these are trees.   The added elevations from the parking garage are in red and yellow.  Points in blow saw 5 or more meter lower elevations during the period, and these are almost all trees.

Overall the elevations are very similar for most areas in the two surveys.  I plan to rerun my storm surge flooding model of Annapolis  to see if it makes any difference, and to cover a larger area.  The model uses a post-Isabel data set that may not  be publicly available.

Summer Air Shows

This summer I saw two air shows.  The first was just outside my office, and I used my large camera with a 300 mm lens.

The second air show was at the Johnatan airfield (https://www.histouring.com/en/historical-places/johnatan-collection-aerei-storici-famosi/) in nothern Italy.  It had only two planes, a replica of a World War I Fokker triplane (go Snoopy!), and an original de Havilland DH.82 Tiger Moth 1930s British biplane.  Better scenery in Italy, but it was incredibly hot as we were there during on the the summer's European heat waves.