Welcome back to Let's talk tech Thursday, the newsletter that tried to be AI-written but failed...

Speaking of AI, this week is an AI-heavy edition, with some good and some bad - depending, I suppose, on your point of view.

• For fans of enhanced government surveillance, there's some wonderful news in the form of AI being used to track people without the need for pesky things like adherence to laws around the use of facial recognition.
• Meanwhile, we talk a look at a sci-fi short film inspired by the works of Bengali poet, philosopher, and colonial-opposer Rabindranath Tagore. Oh yea, and it's made almost entirely with AI.
• And we take a look at a new electric vehicle rollout - this time in the form of e-bikes in Rwanda. And they aren't going to let a little thing like "lack of 'appropriate' infrastructure" stop the revolution.

Also, we check in on how M&S are handling their cyber breach (badly), France's attempt to rally EU countries to ban under-15s from social media, and a lovely little reminder that technology exists outside of computers.

Let's dig in...

Top Stories

How a new type of AI is helping police skirt facial recognition bans

Summary

A new AI tool called Track helps police identify people using physical attributes instead of facial recognition, allowing them to bypass laws that usually restrict such tech. The technology, used by various law enforcement agencies in the US, naturally raises privacy concerns. Critics warn that it could lead to excessive surveillance and abuse of power by authorities.

So what?

Much like Big Data can piece together a good facsimile of you based on disparate bits of digital information, it looks like Veritone have achieved the same concept with video imagery.

It's easy to have a knee-jerk "this is bad" response to this, on the grounds that it feels like one more massive infringement on our right to privacy. If you want my opinion (and my default assumption is that you do, given that you're reading my newsletter), being incredibly nervous about how this tech is used is the only logical place to stand.

However, let's take a higher level view of this for a second.

We know that the use of facial recognition leads to Black men being disproportionately targeted by police, because it is biased towards recognising white faces. To put this into context: if you have video footage of a white person committing a crime, have a reasonable shot at getting a shortlist of one to arrest (assuming, of course, that person is in the system, etc etc). Conversely, if the video footage is of a black perpetrator, then the system struggles, and instead spits out a list of maybe a dozen potential suspects - all of whom are then arrested and at the mercy of the criminal justice system.

By increasing the number of variables that are being monitored, many of which would not be subject to the same level of inherent AI bias, it might not be unreasonable to think that people are less likely to be falsely identified by this new system.

There's a lot to consider here. But just to pick one point: sure, maybe this will lead to more accurate (less biased) surveillance. But we're still talking about surveillance. This represents one more tool in the arsenal of authorities to keep tabs on citizens. Perhaps worse, it allows this tracking to be done under the pretence of anonymity. But how much data can you be allowed to stitch together before this system stops being anonymous?

My comparison to Big Data is not an accident. Anyone who knows their onions around Data Protection and GDPR knows that just because you don't log someone's name, doesn't automatically make the data anonymous. With a large enough list of attributes, or a small enough pool of subjects, it can become quite simple to identify people. In the Foundation I work at, I fill out an anonymous staff survey every year. But how many mixed-heritage, 30-something men who've been with the organisation more than 5 years do you think my charity sector organisation has? (The answer will not shock you.) If the company is also collecting that information, suddenly that dataset stops being anonymous.

That's what we're talking about here - sure, facial recognition is an immediate identifier. But if the system can track and catalogue everything from the clothes I'm wearing to my facial features (dark skinned, bearded, glasses wearing are all identifiers without being facial recognition), and the stations I frequent, what is the practicable difference?

And that's before you get into the science that says everyone's walk is like a fingerprint...

Related story: The way you walk may soon be used by authorities to identify you

I'm a filmmaker who used AI to make a sci-fi film for under $300. I'm deeply conflicted about it.

Summary

Filmmaker Aleem Hossain created a sci-fi short film for under $300 using AI tools, exploring themes of identity and cultural heritage. Although he feels a sense of authorship over the film, he is conflicted about profiting from it due to concerns about AI's impact on creativity and jobs. His film, "Do Bangladroids Dream of Electric Tagore," explores cultural displacement, told through the eyes of robots.

So what?

Shoutout to my Bengali wife Ishita, who has been pestering me for a shoutout since LT3 #1. She took me to Tagore's house once, so this seemed an appropriate time to do that.

It's easy to have the AI debates in the abstract. "It's helping the little people / small organisations / minoritised individuals have a seat at the table" vs "It's IP theft that's destroying jobs and the planet" makes for a great pub chat. Here though, Aleem Hossain eloquently argues both sides of the creatives vs AI debate from the rare position of being super informed on both sides. As a filmmaker, he's very aware of the damage generative AI can cause to the creative sector. And yet, as a South-Asian man with a Desi-sci-fi story to tell, he's never going to get funding for this.

Honestly, I oscillate quite a lot in my thinking on the cost/benefit of AI. The morality of biased systems underpinning our most popular LLMs, the sheer environmental expense of training the same, and the pillaging of content creators' work for the sake of progress, are all things that weigh heavily.

But while I don't take any of that lightly, I nevertheless usually come down on the side of "cautiously pro-AI". Technology exists to make life easier. AI, at a basic level, is no exception to that. The nuances might be a little more complicated than some other tech, and the potential harms it can lead to a lot higher. But equally I've seen the things it can unlock for individuals and organisations. From neurodivergent individuals who can suddenly find working in neurotypical spaces much easier, to capacity-strapped charities who can keep up with their better funded peers, AI can be a real game-changer. (And yes, I know, I know. Maybe instead of throwing robots at the problem, we should just have more neurodivergent-friendly workspaces. But we don't. So here's something we can do in the meantime.)

Technology is a leveller, is my point. Whether that's money, capacity, or access to information. When it comes to telling stories through a sci-fi lens in particular, as Hossain points out, these films are usually prohibitively expensive to make. There is a delightful irony in technology itself being the key to unlock the telling of these stories.

Also, and I suppose this is neither here nor there when it comes to the intellectual debate of it all, I've watched the 4 minute film and I love it.

Related link: Watch "Do Bangladroids Dream of Electric Tagore" and read more of Hossain's thoughts

The grid is weak. The bikes are electric. Rwanda is betting it'll work anyway

Summary

Rwanda plans to shift to electric bikes despite challenges with its power grid. Startups are developing innovative solutions like solar-powered charging stations and battery swapping to support this transition. The goal is to reduce carbon emissions and set a new standard for electric mobility in Africa.

So what?

While it's clear that we won't be able to keep our dino-juice* guzzling cars forever, at the moment one of the biggest arguments against electric vehicles is that the strain on the national grid would be too much if you made everyone get one. I didn't say it was a good argument... there are definitely things we can do to beef up our electricity grids.

But that isn't necessarily the case everywhere. In parts of the world, including Rwanda, the grid is unstable enough as it is. You don't need to add 100,000 electric vehicles in the mix for the system to fall over, rolling blackouts and power cuts are already common. Nevertheless, venture capitalists, startups, and government officials alike are betting on the rollout of electric bikes across the country, to replace many of the current motorbikes in use.

I suppose its a sort of technological leapfrogging. In the west, we've gone in guns firing (if not blazing) on electric vehicles, and only retroactively are we trying to figure out how to make the grid work with the increased electricity demand. But these considerations are baked into the scheme for Rwanda, with battery swap technology, solar charging stations, and plans to deal with older batteries being considered from the jump.

Just like using mobile phones to make payments - something the African continent had a handle on way ahead of the rest of us - maybe some of the practices Rwanda and neighbours implement for their e-bike revolution will pave the way for better solutions for electric vehicles the world over.

* By the by, did you know that fossil fuels are in fact not dinosaurs? Most of our fossil fuel reserves were locked in millions of years before the first dinosaurs roamed the earth. It actually comes from plant and plankton remains. And you thought you were only going to learn tech things from this newsletter!

Any other tech-based stories?

It's just a part of shoes - but they're life-changing for me

Researchers at Queen Margret University have developed a pair of shoes to help people with footdrop, a condition that makes walking difficult, often experienced by those with MS or cerebral palsy. The shoes use adjustable velcro straps to provide support, allowing wearers to walk confidently and even go hill-walking again. The hope is to create similar footwear for people worldwide, including simpler designs like sandals for those in low-income countries. It's a nice reminder that technology innovation doesn't have to involve microchips and electricity.

France's AI minister calls for Europe-wide ban on social media for children under 15

France's AI Minister Clara Chappaz wants to make it EU law to stop social media companies targetting under-15s, but she'll settle for making it French law. Critics to the proposal argue that age-verification is difficult and poses data security risk. But Chappaz's counterpoint is that there are other, anonymous, ways to verify age.

If I might opine here, I don't know of any such technologies that feel truly anonymous. Most popular forms of verification include things like temporary credit card charges, or more recently using AI to estimate age based on an uploaded photo. Clearly, these systems have their downsides and these weaknesses are going to be the things that Big Tech leans on - for better or worse - to argue against age-verification measures.

M&S says customers' personal data taken by hackers

The initial M&S cyberattack story is a little long in the tooth now. But it recently transpires that individuals' personal data was taken by the attackers. M&S are downplaying this by assuring everyone that no "useable payment or card details" were taken, but you should still be concerned. With names and email addresses having been lifted from various systems, there is more than enough stolen data to plan a new wave of highly targeted phishing attacks. These fake emails can appear ultra-convincing due to the level of personal detail they contain.

M&S have said that to reassure individuals, they will be prompted to change their password on their next login. While this is a good idea generally, in this case it isn't actually going to change anything.

What's all the more unbelievable about this story, is that M&S "didn't have a cyber attack plan". In 2025, there is no excuse for this.

That's it for another edition!

Are there types of stories you want to see more of? Reply to this email and let me know.

In the meantime, have a great end to your week. I'll see you next time.

Will