Will audio ‘fact checkers’ be able to help when fake clips of prominent people are doing the rounds?

Will audio ‘fact checkers’ be able to help when fake clips of prominent people are doing the rounds?

Today I read the unsurprising but kind of terrifying news that a machine learning model can turn text into a pretty spot-on recreation of spoken voice, using Joe Rogan’s recognisable baritone as a test case.

I wanted to know if there’s a way to tell the difference between a voice that was actually recorded and one that was synthesised — could this fake Joe really be ‘perfect’? Would slight imperfections allow experts (or software) to act as fact-checkers for potentially synthesised voices?

My first port of call for any sound/music question is Bevan Smith, Wellington-based musician, composer and sound engineer (full disclosure: he also happens to be my husband). So, through the magic of a text conversation, here’s what I learned:

It’s possible for a human to tell the difference between a recording of Joe Rogan and the ‘faux’ Joe. Once Bevan had a baseline of what ‘real’ Joe sounded like, with some effort he could differentiate between the two. His initial reaction was “the faux sounds slightly higher pitched… it’s constricted and kinda more nasal”.

While the machine learning model might be improved over time to produce even more convincing results, Bevan said its very hard to fake an instrument or voice exactly. However, he added that by the time you could prove it was fake “it would be too late… people would believe it and that would be it. If you had a video of someone playing a violin but had ‘sampled’ violin playing, 999 people out of 1000 people would believe it was recorded. With the voice, maybe [it would be] 1000 out of 1000, I wouldn’t question it.”

So if this kind of technology became readily available and was used for nefarious purposes, it could be possible to ‘myth bust’ a recording, but by the time that happened, the damage would be done. But for me, that ability to maybe ‘tell the difference’ gives me a bit of hope.

Like OpenAI’s GPT-2 language model, this speech synthesis model represents a big step in what AI models can do, edging their capabilities into an ever more human-like realm. Like the GPT-2, this new model won’t be released publicly to mitigate against potential misuse, but it’s likely this type of tech will make its way out into the world before long. Now’s the time to make a plan for what happens next.

-AP

The Revolution Starts at Home: YouTube and the World’s Largest How-to Guide

The Revolution Starts at Home

YouTube and the World’s Largest How-to Guide

This is the transcript of my presentation at Theorizing the Web on Friday, 12 April 2019 in New York City. This presentation was part of the session “Follow, Like, Subscribe”. Slide images are shown throughout.

Tēnā koutou katoa

Today I’ll be talking about YouTube videos, specifically fix-it videos. While these don’t rack up views to rival Baby Shark or Pewdiepie, they’re still popular and handy if your smartphone screen breaks, or your washing machine stops spinning. I’ll argue that these videos are more than just a useful tool: in aggregate they open up space for critique and a re-imagining of our disposable material culture.

Slide from the presentation: We live in a throwaway era

We live in a throw-away era. It’s easier and cheaper to buy a new $8 toaster from Walmart than it is to fix your old one. If you want to replace a smartphone screen, I hope you have a hairdryer handy to soften the glue before you pry it off with special tools. Want to fix your washing machine? Great — to replace the bearings, you’ll have to dismantle it completely and even break some bits apart first.

None of this is an accident. The neoliberal economic model is finely tuned to ensure that we consume as much as possible. Many appliances, digital devices, and other household items are designed to be hard to fix and only last for so long — for many digital devices, this is about 18 months. This is planned obsolescence, and it rears its head in a number of ways, including:

• making things using cheap materials and hasty construction so they don’t last;

• making replacement parts hard to find or so pricey you may as well replace the whole thing;

• using Copyright law to protect digital systems, so repair and upgrade options are expensive and restricted;

• dropping ongoing support for software, limiting the usability of the technologies that run it, and finally;

• preventing repair, and even recycling, using what I call obstructive design. This includes gluing or welding things shut, using many different types of bespoke screw, or changing core components from model to model.

I did a hasty collection of anecdata from friends in the lead-up to today, and found that toilet seats, car gear boxes, headphones, and printers are among items that are cheaper and easier to replace than repair — and I imagine that’s just the tip of the iceberg.

The current linear system — “take, make, waste” as the circular economy folks call it — is not sustainable. Organisations like The Ellen MacArthur Foundation, the Cradle to Cradle Product Innovation Institute, and the United Nations E-waste coalition are all making efforts to enable transition towards a circular economy — that is, where all materials and components are kept in use for as long as possible, and waste is designed out of the system. This includes designing products for durability, repair, upgrades, re-use and eventually recycling of their components.

To make a change in our system, consumers will also need to act, by changing their habits and applying pressure to corporations and governments. While adapting consumption patterns on an individual level can seem ineffective, if done on mass, it can slow down the the cycle of production. History tells us that the more people demand action on an issue, the more likely it is that change will happen.

Slide from the presentation: How can Fix-it videos help?

So, how can fix-it videos help?

The scale of the problem is huge, but our view is very narrow. Most of the time, as consumers, we only see a very small part of this complex system of production and waste disposal.

When our products become broken or unuseable for whatever reason, there’s a small window when we have to think about what to do with it.

Once you put something in the trash or recycling, return it to the supplier, or donate it to Goodwill, you don’t actually need to think about what happens next: out of sight, out of mind. And when products are cheap, getting rid of them is often the path of least resistance. But it’s that moment of resistance that I’m most interested in. For those who do pause and consider “can I fix this myself?”, the thriving category of Youtube fix-it videos can crack open a window through which to imagine a different, less alienated, less extractive world.

And here, I’m speaking from experience.

A few months ago my washing machine broke. I was told by the repair guy the bearings were shot and it wasn’t worth trying to fix. But instead of sending it straight to landfill, a friend and I took to YouTube to see if we could try our hand at fixing it — the biggest cost seemed to be the labour, and we had time. A video by Matt from a UK spare parts business became our guide, and we got cracking.

Presentation slide: Image of author fixing her washing machine

Our story doesn’t have a happy ending. We got to the centre of the machine after many hours of work, and found the problem was worse than we thought. It went to the landfill after all.

But what I did get out of the experience is a sense of great anger, and of revolutionary potential.

And maybe others can too.

Presentation slide: YouTube search box for “how to fix”

There are a lot of varied videos that fall into the fixit genre — a search for “how to fix” followed by the household item of your choice will most likely throw up at least a couple of results, and if you search for something like “iPhone screen”, you’ll be scrolling through pages of options and find videos with millions of views.

So, what are these videos actually like, and who makes them?

Presentation slide: screen caps from YouTube fixit videos

Unsurprisingly, the form of the videos vary depending on what’s being fixed. Repairs for digital devices and headphones are usually shot fairly close up from above, with annotations popping up, or a voice over. Large appliance repair videos are a bit trickier logistically, so you’ll sometimes get weird shots of people shining a torch inside the machine.

What about the creators? Well, first up, there’s not exactly a lot of diversity in the creators of these videos: nearly all the videos I came across were from men.

Some creators had channels that focus on a particular type of repair — be it appliances, furniture or digital devices. I found that appliance repair videos are largely made by maintenance and spare parts businesses, rather than interested individuals. You also see videos popping up from creators with a wider scope of interest, particularly general home maintenance or DIY.One highly viewed washing machine repair video was by Chris Fix, who normally makes car repair videos — and has over 4 million subscribers.

The common thread in all the videos is that they’re very practical and aimed at actually giving people the best chance of extending an item’s lifespan. It’s clear from the overall aesthetics of the videos that these videos are less about performance and more about giving actual, practical advice.

I think this blurb from an “about” page of one creator, Richard Lloyd, sums it up nicely:

“My goal is to help people whenever I can in any way I can. I will post a video on anything I think will add value to your life. I love to share my knowledge. I am a Jack of all trades and master of some.”

So, we’ve looked at the creators of the videos — what about the people watching them? Is this just another form of entertainment to binge-watch, or a knowledge transfer system that actually helps people in the real world?

I immersed myself in the comment sections of videos about washing machines, smartphones, headphones and toasters to see if some common themes emerge. And emerge they did. Along with general thank yous, messages of appreciation and the odd mansplainer, recurring themes include:

• people sharing their own experiences of trying to fix things — both successful and less successful;

• people realising that it’s often cheaper or easier to get a new item or reminiscing about the “good old days” when things were made to last; and

• people asking for further advice — and often getting it.

Overall, what I found was a bunch of people learning something new about the inner workings of the objects that populate their homes, and being pretty pumped about the experience.

Presentation slide: So What?

So, what are the key takeaways here?

First, and probably most importantly, is that this haphazard collection of videos could well have the power to make a big change in the world. We know that millions of people are watching, many with the intention to try their hand at fixing something they own. When you’re empowered to fix something yourself, you might try it more often, and think twice when it comes to getting rid of your things.

For people who realise it’s not be worth fixing something due to poor construction or, say, too much glue? Hopefully that’ll cause anger, and lead people to pressure corporations to do better, and governments to use the tools at their disposal to enable systems change.

The final takeaway is that the medium of video, coupled with platforms like YouTube, are a powerful tools that can enable knowledge transfer at a huge scale. Passing on repair and fixit skills from one person to another is nothing new, but now this can happen on a scale that isn’t possible with in-person classes or one-on-one instruction. The ubiquity of affordable smartphones with decent cameras also means that there are many millions of people who have the tools they need to produce and consume knowledge without traditional gatekeepers or high production costs.

So, to recap: We’re consuming at an unsustainable rate. YouTube videos, and their creators, are teaching people skills to extend the life of their stuff, and — sometimes unknowingly — giving people a glimpse of a previously hidden system of production at the same time.

-AP

Cave of Forgotten Memes

Cave photograph © Jeremy Bolwell cc-by-sa/2.0 — geograph.org.uk/p/2972453

Cave of Forgotten Memes

Don’t you hate it when you come up with the perfect hot take when it’s just too damn late?

The internet moves disturbingly fast for an Old Millennial / Generation Catalano / Xennial / Oregon Trail Generation-er like myself, and sometimes memespiration strikes hard well after the news cycle has moved on. It’s the internet version of l’esprit de l’escalier, and I wish there was a word for it. I also wish there was an appropriate forum for sharing years-old hot takes that are too inconsequential or stupid to deserve an analytic essay.

With that lamentation out of the way, I’m here to say that I just came up with the meme / micro cultural analysis about a two-year-old viral image, and I must share it. Here’s the photo collage mashup version, followed by an explanation:

OK, so admittedly this is a bit of a deep cut, and requires some unpacking (a surefire way to kill a meme, if it weren’t already dead).

American Match: a memesplanation

@usarmy photo by: Staff Sgt. Vito T. Bryant

I recently came across the profile page of a conservative, military-loving Twitter user, who was using a cropped version of this photograph of a crying soldier as their header image. After reverse Google-image searching the image I found the original photo (above), which apparently went viral in 2016 after its initial appearance on the US Military Academy Instagram page.

The photo shows Haitian immigrant and 2nd Lt. Alix Schoelcher Idrache at his West Point Military Academy graduation ceremony, wearing an Army uniform and tearfully gazing into the middle distance.

In its application as the header image on a conservative Twitter account, it was clear that this photo was doing some serious cultural work, and I was immediately struck by the similarity between the image and another famous photo of a Black soldier:

This 1955 image, of a Black soldier on the front cover of the magazine Paris Match, was one of the central examples analyzed by Roland Barthes in his famous book Mythologies.

Barthes described the way this magazine cover image creates meaning for the viewer. He writes:

“I am at the barber’s, and a copy of Paris-Match is offered to me. On the cover, a young Negro in a French uniform is saluting, with his eyes uplifted, probably fixed on a fold of the tricolour. All this is the meaning of the picture. But, whether naively or not, I see very well what it signifies to me: that France is a great Empire, that all her sons, without any colour discrimination, faithfully serve under her flag, and that there is no better answer to the detractors of an alleged colonialism than the zeal shown by this Negro in serving his so-called oppressors” (115).

Dated language aside, Barthes’ description of the 1955 image’s function can be mapped almost directly onto the image of Idrache, with the US replacing France as oppressive neocolonial power. As a Twitter profile header image, the photo connotes a belief in “color-blind” patriotism, fealty to the US military, and a conviction that people of all races should be grateful to serve. The fact that Idrache was identified as an immigrant adds another, specifically American layer. Immigration (the right kind of immigration, at least) is important to the American self-conception, and this public display of immigrant gratitude serves the narrative well.

As with the Paris Match cover, the backstory of the photograph falls away, and Idrache’s biography and specificity is replaced with a set of connotations that can be mobilized by audiences as varied as pro-immigration activists and US military boosters.

In today’s media landscape, it isn’t just powerful publishers and other platform gatekeepers that have the authority to reify an image and circulate specific cultural referents. Virality is its own authority, and the image of Idrache gathered its power and weight as it rolled around the internet in Katamari Damacy-like fashion (there’s another ancient reference for you), accumulating sticky layers of meaning. It found itself attached to news stories, gathering extra-textual meaning in the form of Facebook comments and inspirational quotes, pinned to inspiration boards and reused as avatar images.

fin.

Thus ends the first entry in this cave of forgotten memes.

-KP

What Google’s AI Principles don’t tell us

What google’s AI Principles don’t tell us

Today, Google CEO Sundar Pichai published a blog post setting out seven principles for how Google will use and develop AI, as well as a list of things it won’t use these technologies for.

In a perfect world, Google would have had principles for AI in place much earlier in the game, rather than establishing them after staff backlash over the company’s involvement in Project Maven.

While I think they’re a pretty solid set of principles, I was left with a feeling that there was a lot left unsaid. Here are a few places where reading between the lines throws up more questions than answers:

Vague terms

There are a lot of strategically used qualifiers in Pichai’s blog post. I’d love to know exactly what he means by the following:

• Unfair bias (is that even a thing, isn’t bias inherently unfair?)

• Unjust impacts

• Appropriate cases

• Appropriately cautious

• Overall harm

Moving from principles to practice

I love a good set of principles. I also love a good plan for how principles will be put into action. In his post, Pichai said “These are not theoretical concepts; they are concrete standards that will actively govern our research and product development and will impact our business decisions.”

While this is good to hear, I will remain skeptical of the impact they’ll make until I see a concrete plan about how they’ll be put into action. What will employees and leaders be expected to do on a day-to-day basis? Will Google implement some sort of AI impact assessment (AIIA) for technologies they’re developing? What implications will there be for non-adherence with the principles? And if they do these things, will we ever hear about it?

Google has a lot of power and, in the end, can do what it wants

Near the end of the post, Pichai starts a sentence with “While this is how we’re choosing to approach AI”. This makes it very clear that at the end of the day, Google doesn’t have to adhere to any principles. I hope that, in time, these will become embedded into the way Google does things, so they’ll be hard to change or alter without significant backlash or consequences.

Google is a key player in AI, which in turn has the ability to change how society functions, and that’s not a power that should be taken lightly.

-AP