Reading

https://www.theverge.com/2021/3/8/22319173/openai-machine-vision-adversarial-typographic-attacka-clip-multimodal-neuron

Why does it seem like the smarter something is, the more easily it can be taken down by unexpected stupidity? In this case, machine learning models that are smart enough to “read” handwritten words and recognise them as references to an object mistakenly assume that the word is the object. The end result? Sticking a handwritten label on any image scanned by this system will cause it to report that its “seeing” whatever is written on the label instead.

From technical stupidity to natural eccentricity: these sea slugs can decapitate themselves and then regrow their entire body including all of their internal organs. One reason why might be medical; the researchers guess that the sea slugs might use this feature to ditch their old body if it gets infected by parasites, so that they can grow a healthy one back. But how do they survive without any internal organs? By borrowing the chloroplast cells that control photosynthesis inside their favourite food, algae, and powering themselves with the sugars produced when the chloroplasts photosynthesise.

Now, heading back to technical errors… https://getpocket.com/read/3284812993

This is a good lesson about how the bad software design, and more importantly the bad user interface design, that forces people to regularly rely on clunky workarounds is a recipe for problems. It always amazes me when I hear about large-scale errors that ultimately happen because the software required to do the job was incorrect, unintuitive, or poorly-designed. That’s especially true for institutions as large and as impactful as banks, where any error could result in millions of pounds being lost or mislaid – surely they can afford to spend money making the software that earns their wealth as usable and efficient for the employees using it as possible?

I also finished reading Toksvig’s Almanac, and posted a review.

On more book-themed news, I was introduced to Standard Ebooks.This is a volunteer-run project that takes existing public-domain ebooks from places like Project Gutenberg and gives them a makeover; from typesetting and formatting, to finding high-quality public domain artwork, to coding the books in an open-source format with version control to make them reliable and standardised. Its a really cool example of using technology to safeguard and support literature, rather than to lock it down.

Gaming

This week, I started replaying the campaign of Grand Theft Auto V. While I did play through the campaign on Xbox 360 shortly after its release in 2013, I played through it in a very fragmented way that means I missed a lot of the side-quests and story details. As a result, I’ve often wanted to return to it and play the story “properly”, and see more of the details and events in the sprawling world of Los Santos.

I also revisited the online aspect, which I haven’t touched for years. As many missions can’t be played solo, I’ve barely played any of the “storyline” of the online aspect. I spent time yesterday doing smaller side-events that can be done solo, which was surprisingly enjoyable… well, until I made an enemy of two other players who then kept repeatedly killing me every time I respawned, which meant that I didn’t have enough time to go into the pause menu and choose any options that could have removed me from the situation. As a result, the only thing I could do to get away from this was to leave the game, which was annoying.

Aside from that, I’ve also returned to Two Point Hospital and started a new save file to try and relearn the strategies to take on the harder levels.

Miscellaneous

A few days ago I bought my (virtual) ticket for this years OER conference. The only time I’ve actually been is 2017, when I presented my MSc research, but the effect that 2020 had on education and educational technology, as well as things like ebooksos, have brought my attention back to the tireless work of the open community. Like last time, I’m sure that I’ll finish listening to the talks and realise that everything I know is a tiny drop in the education ocean!

Reading:

A modder known online as T0st published details of how they found and fixed an issue in Grand Theft Auto V. Their solution reduces the loading times of GTAVs online mode by 70%. As a result, Rockstar Games (the GTA developers) rewarded T0st with $10,000 and incorporated parts of his solution into an official update.

https://www.pcgamer.com/rockstar-thanks-gta-online-player-who-fixed-poor-load-times-official-update-coming/

This chain of events has received mixed responses. On one hand, its great that Rockstar listened to T0st’s solution and used it to improve the game. The financial reward was also something they didn’t have to do – Rockstar’s bug bounty program rewards people who find and report crucial security or privacy flaws, and they chose to make an exception for T0st and award it to them as well. Given the size of the online GTA playerbase, a 70% reduction in loading times translates into thousands of human-hours saved in future too.

The criticisms of Rockstar are also merited, however. GTAV has been out for 7 years, been released across 3 console generations, and has also been referred to as the single most profitable piece of digital entertainment in history.Yet Rockstar didn’t do anything to implement what was apparently a simple fix until someone unrelated did the work in their spare time. A lot of people are cynical about whether their response was solely about avoiding bad PR, and I understand where that view comes from.

Adobe, on the other hand, are definitely being criticised due to this illogical action:

https://torrentfreak.com/adobe-goes-after-27-year-old-pirated-copy-of-acrobat-reader-1-0-for-ms-dos-210315/

Really, Adobe? I know that copyright is complex, and that brands have to actively protect their copyrights or their legal right to them can be made null. But applying a DMCA takedown here seems excessive, especially as the tweet was only a link to another website rather than the direct source of the software. Also, Acrobat Reader 1.0 is 5 years older than the DMCA legislation itself, which makes this feel even more unnecessary.

Continue reading “Weeknote – 19/03/2021” →

Quite a lot of the people I pay attention to online – often, the Open Education / tech ethics crowd – use Weeknotes to keep track of their many projects, links and creations. I find the weeknote structure appealing, so I’m going to test it out as a way to a) keep a better track of the (many) things I start and don’t finish, b) talk about some of the interesting articles I read or games I play that wouldn’t ever make it into a full blog post or review. (I won’t be doing predominantly work-focused weeknotes – they’ll mostly be a reading list/link roundup – but I’ll mention work if there’s something interesting to include.)

Articles

A few days ago, I read this article about how smartphones are helping illiterate women in India by giving them a way to socialise with people outside of their household, run businesses, take part in activism and report information to journalists, all through voice commands and sharing images. This really made me think, because I’m someone who needs to write things down to make sense of them, and who learns everything by reading it, so I cannot comprehend what not having that option must feel like. The article describes how one of the women, Mallika, can now use WhatsApp to talk to a friend – before smartphones and data plans, she had to climb a mountain to communicate with them.

Now, on to ice cream trucks. I found this article interesting, because its a really good example of how one person being in the right place at the right time (and knowing the right person) can change an entire industry or movement. Finding out that the company which makes the music boxes inside 97% of all ice-cream trucks consists of just 2 people really reminded me that so many of the things we’re familiar with are the results of decisions made by a handful of people.

Video

Clutching at Random Straws — This is a talk given by stand-up mathematician Matt Parker, to a school in England. It’s a great discussion, with practical and funny demonstrations, of how easily we can be misled when it comes to data. For example, Parker shows the students how collecting large amounts of data can fool us into finding “impossible” co-incidences easily, and how we can easily be influenced into hearing patterns such as messages in music played backwards. All of this is done in a way that doesn’t require high-level maths knowledge, and its a really clear introduction to some of the ways that we can be misled by data.

Continue reading “Weeknote – 12/03/2021” →