Saturday, November 12, 2022

Wordle Words for Irish

 There is a version of Wordle for Irish called Foclach 

Matt Parker has a youtube video working out how many five letter word combinations cover 25 of the 26 English letters. 



This is an exact cover problem and has been studied for ages. And theres some great follow up videos to that one on how to make quick version of the checker.

There are 23 letters in Irish used by Foclach (this dataset). There are rare words with v (víosa (visa),veist (vest), veain (van)) and other English letters but they are not accepted in the game. 



I took this code from azirale's matt parker answering version and changed the alphabet to Irish and looked for words that had all 23 alphabet out of 25 letters in 5 words. I got an Irish word list from here. There are only 6500 words in it. With 910 being 5 letters long.

 

There are loads of  word combinations where all 23 letters are found in 5 5-letter words. Some examples are

 beola-nádúr-putóg-séimh-fíoch

 bheir-camán-folús-putóg-éindí

 cónra-fostú-geábh-pluid-céimí

 cúige-díobh-pluma-óstán-sféar

 féile-putóg-rúnda-íomhá-bácús

 fíoch-maide-putóg-slánú-bréan

 fórsa-geábh-modúl-péint-curaí

 geábh-modúl-próca-stuif-éindí

 méadú-oibrí-putóg-sleán-fiuch

 oíche-putóg-ráfla-sméid-bunús 


I thnk this implies the Wordle in Irish is easier than in English at least for a computer.  







Tuesday, September 13, 2022

How Dangerous was Whaling?

In Moby Dick he writes 'upon one particular voyage which I made to the Pacific, among many others we spoke thirty different ships, every one of which had had a death by a whale, some of them more than one, and three that had each lost a boat’s crew' A boat had about 30 sailors so this would be at least 1 out of 30 on 27 boats and 6 out of 30 on 3. So 27+18=45 out of 900 or 5%. 

 Comparing that to Astronauts. 574 people have gone into space. 31 Astronauts have died. Including training/testing accidents. Which is 5.4% Though those astronauts died on some journey not just one. On this one data comparison whaling was as dangerous as being an astronaut.
 
The journeys were of 3 to 4 years. With very little if any visiting of ports involved. Which is longer than Mars journeys are projected to be.

Friday, August 05, 2022

NOAA Warming in US Graph

 There is a strange graph that is common in Climate Change denying circles.



The issues I see with it are

  • It only goes up to 2011. More recent data should be included.
  • The second access of CO2 is very weird. Generally 2 Y-axis graphs are frowned up and here it is confusing.
  • The graph shows US but the claims around it are about the World. Which are two very different things.

The link in the graph no longer seems to be working http://ncdc.noaa.gov/oa/climate/research/cag3/cag3.html 

The Noaa does have data for North America still and it shows warming in the last decade https://www.ncei.noaa.gov/cag/global/time-series/northAmerica/land/1/6/1880-2022



Tuesday, June 28, 2022

Modernising Old Films

We can make old films look much more modern using AI. This is a great article on the technique used on early 1900s films 

"Using a variety of publicly available algorithms, Shiryaev colorized and sharpened the film to 4K resolution (that’s 3,840 horizontal pixels by 2,160 vertical pixels) and bumped the choppy frame rate up to 60 frames per second, a process known as frame interpolation."

The 2018 film They Shall not Grow old using WW1 footage used similar techniques as described here. But even in 4 years the AI technology has improved hugely. 



Archive historical footage is one thing. But is there a market for films upscaled in the same way? Would the Marx Brothers with realistic colours, smoother motion and sharper faces be watched by more people? I think the great Pathe news footage upscaled could make some great documentaries. The wired article goes into interesting detail about what counts as improvements and what is just being made up of the AI algorithm.

The value of these sorts of changes I am not sure about. The cost of them has been falling rapidly though. Due primarily in the last few years to improvements in the algorithms and hardware.

The recent crypto crash has put large numbers of GPU's onto the market cheap. In some cases these are located beside renewable power supplies for cheaper electricity. These tend to be remote locations with not great internet speeds. Old films can be moved around on USB sticks as they are not real time bandwidth dependant. At some point I think the price of Bitcoin will go low enough that a container filled with GPUs currently using, close to free, night time wind energy will have a greater return upscaling old films than mining.

Friday, June 17, 2022

The 15 year global warming pause that didn't happen

There is a nine year old article that is doing the rounds online. about how there was no increase in global temperature for 15 years



Here is what jumps out at me that is wrong in the article and its current interpretation

1. It is 9 years old. And those posting it don't mention the clear increases since.


2. They don't mention the changes before 1998




3. Or indeed the increase in the 15 years the article says there was no increase

4. "it is expected to address the fact that 1998 was the hottest year on record" 1998 wasn't the hottest year by 2013


You could claim the Hadcrut data, the IPCC use, is flawed in some way that means we should ignore it. But then you can't post the article as it uses Hadcrut data. 
It is weird a 9 year old article has resurfaced and those posting it are only looking at the data for this exact 15 year period. And even then not actually looking at it. The code for these graphs is here

Wednesday, June 08, 2022

Combined Solar and Wind Farms in Ireland

I was talking to a wind energy guy and he expects loads of these wind farms to get solar added soon.

He explained to me 

  • The cost of connecting to the grid is already covered. 
  • Sun tends to be best when wind is lowest. Daytime and the summer both have lower levels of wind.
  • These land owners do business with electricity companies already. Setting up that relationship is a big challenge that is already covered.
  • Most of the land wind farms are on now is fairly hilly and not of huge value for farming. Herd reduction is not huge from solar farms.


There are about 300 wind farms in Ireland

The one solar farm we have generates enough electricity for 3600 homes

There are about 2 million homes and apartments in Ireland. Which would be 3600 houses* 300 farms is about 1 million homes. That solar farm is 25 hectares. If all the wind farms had the same size solar added thats 300*25 hectare is 7500 hectares or 75,000,000 square meters. 75 square km. Which is a fair chunk of land


But the solar farms would be spread out 

This calculation is very back of the envelope. There are also all sorts of constraints about how much solar power can reasonably be put into the grid. Windless nights happen and electricity has to be available then so backup to wind and solar is needed. Also I am not sure if it is not just more efficient to put the same panels in Morocco but given the points above it is probable Irish Government subsidies will go to Ireland.

It seems a lot of solar power will be added to what are now wind farms soon.




Wednesday, February 23, 2022

An Abundance of Housing and TV that Promotes this

We know how to build lots of things cheaply. Things like clothes, toys, TV/computers and food have all gotten relatively much cheaper over my lifetime. These are mass produced, at a pretty high quality in factories with very specialised labour.




Housing however has not. The numbers who own by age 30 (when you want some stability if you are going to have kids) keeps dropping. One thing that could encourage cheaper housing is more programs that promote non unique one off builds. Instead of Ireland's most exclusive homes we should show the most inclusive ones.




Just Grand Designs where the house is fine and loads of them are made. “I can’t see the justification for gross spending on what are essentially basic requirements.” Kevin McCloud



Room to Improve where we build enough houses to give adults room to improve their lives by not having to silently ride in their parents house



RTE Home of The Year should be for a design of house that's had 1000 delivered this year and people can afford. Then have a decorating competition show between people who live in this same type of house. Thats fairer in terms of a competition and more practical help to people in how to decorate their own house.




There isn't a build your own car program. Grow your own food. Make your own pencil. As a hobby these are all worthwhile but the home building programs are not about a hobby. 

The reason things are cheap is the specialisation of labor and mechanized production. Why are houses different? And why does TV encourage this expensive way of building houses?



Wednesday, February 09, 2022

Meta analysis of Vitamins Open Dataset

Vitamins and Mineral supplements are a huge industry worth billions or euros per year. The evidence that they have benefits is a lot more controversial though. I first learned about this controversy around supplements from this overcomingbias post

This 2007 meta analysis Mortality in Randomized Trials of Antioxidant Supplements for Primary and Secondary Prevention concluded "Treatment with beta carotene, vitamin A, and vitamin E may increase mortality. The potential roles of vitamin C and selenium on mortality need further study."

There have been meta analysis since including newer studies with slightly different conclusions. But this one is pretty well laid out and we can add newer studies later.

With any statistical analysis it is possible to find issues with what data was included or not. One way around some of these issues is to make a public dataset of the trials used in a meta analysis and allow a quick reanalysis with some trials removed (or added).

I have been reading this book Doing Meta-Analysis with R: A Hands-On Guide and by combining an open notebook of how the meta analysis was calculated and a dataset that everyone has access to I believe arguments over the effects of inclusions of some trials and methods used in a meta analysis could be made much more explicit and worked out faster.

There is an open collaboration platform for working together on datasets https://datastack.net/ I have no connection with it. But it seems a good place to work together to make supplement meta analysis easier.

If you want to help track down papers used in this meta analysis and put them in a format others can use please message me. I have started putting up a dataset https://datastack.net/cavedave/vitaminRCTs/





Tuesday, February 08, 2022

The Pandemic. Fathers spend more time with their children

 A good thing that came out of the pandemic is how much time fathers got to spend with their children.

According to this research in 2014 time focused on their children was 35 minutes a day for working fathers. Up from 5 minutes in 1974.

This means I've spent more extra time over the usual amount properly with my kids in the last 2 years than 1970s dads got to spend with their kids over their entire childhood. And that is just extra time due to the lack of commute or work trips.

I am not counting time off school here. As that was just so difficult for so many. Particularly for mothers whom the balance of minding kids during the lockdown hit disproportionately

How are mothers and fathers balancing work and family under lockdown?

And plenty of fathers still had to be physically in their jobs just to keep society running.

But on that one measure of focused time spent with kids my generations relationship with our kids will be much closer than that of fifty years ago or even ten years ago just from the extra time we got to spend together.

Monday, January 24, 2022

Buying the Rights to Franchises

People are laughing at the SpiceDao for spending $3.8 million on a book. But one thing has shown is a group of fans of a story can get money together to spend on that story. And this can be organised in a decentralised way. We don't need Disney to buy franchises, fans can do that now. 



The stories that are most undervalued in monetary terms compared to how much fans value them are probably 

1. Young Adult stories as people who read these at a formative time find them really important but traditional publishers don't. 

2. Sci-fi and Fantasy books as these are generally seen as less prestigious than standard literature 

3. Non English Language stories.

4. Books that are out of print as that can change who owns the rights or at least makes them cheaper. 

5. One from a few decades ago so the fans are now old enough to have some spare money. 

6. Choose your own adventure and adventure games are eve further out of mainstream critical valuation. Comics have been mined for IP value but they are such a rich and diverse medium I am sure some areas have been ignored.

Fans of a non english language YA sci-fi/fantasy series could for much less than $3.8 million get the IP rights to a story like this. Frequently the author will have the IP rights themselves and they could well be easier to deal with than a publishing company. 

I would guess once you have the rights the order of things to get the story more popular are

1. Republish the story on kindle. Effectively free. Give the author a lot of any money this earns so it is in their interests to encourage the steps below and to sell you the IP rights cheaply.

2. Set up a fan site including forums and fan fiction. Not expensive but it would take up some time with moderation.

3. Create a prize for fan art including images, fashion designs and comics and for fan fiction stories. Again this is very cheap.

If the original language can create enough interest in the franchise to keep these busy

4. Publish special editions of the book. With a new introduction, cover based on a prize winner etc. 

5. Translate the story. A high quality book translation is pretty cheap. Especially if you are not under strict time pressure. Cost 5-15k per language.

6. Copy steps 1->4 in the newly translated language.

After the first franchise you carried this out on you can reuse the website software etc. I would guess each additional franchise would cost under 50k to buy and do these steps with. Even if it was 100k that would still be 38 different franchises resurrected for the price the spiceDAO paid for the physical copy of a book.

And if any one of them really succeeds and becomes popular enough to spawn a movie a large profit could be made. But thinking about the worlds of fans I think just seeing their favorite characters again either back in print for a new generation, back with new adventures in a new series of books is really what most fans want.