Let's talk tech Thursday #8
Welcome to another edition of Let's talk tech Thursday, the newsletter with a 75% chance of being sent on a Thursday.
This week we:
- Take a look at some of the great ways technology is helping bring sports fans closer together in their enjoyment of live games
- We immediately 180 into a terrifying look at some of the ways generative AI plays with out thoughts and emotions
- And we briefly discuss that the UK government is doing about AI and copyright laws (don't hold your breath)
Let's dig in...
Top Stories
Meet the Trio of Tech Startups Bringing Live Sports to Blind Fans
Summary
Three startups, OneCourt, Field of Vision, and Touch2See, have created tactile devices to help blind fans enjoy live sports. These devices translate on-field action into movements that can be felt, allowing visually impaired spectators to follow the game in real time. Users have reported that this technology greatly enhances their experience, making them feel connected to the excitement of the game.
So what?
Let's start with a "feel-good" story this week (and if all you're looking for is feel-good, then I'd strongly advise you stop reading after this one).
There are lots of ways that technology can make things more accessible. From the invention braille in the 1800s, to the closed captioning on TVs in the 70s, to the invention of Accessible Web Standards in the late 90s and early 2000s, we're great at creating new things to include people that we historically seem to find easier to just ignore.
The three companies featured in the article all take slightly different approaches to helping visually impaired people get a tactile understanding of what was happening in a live sports event. With various applications of cameras, magnets, audio and vibration-based queues, and some genuine ingenuity, each company has come up with a way to simulate multiple different sports on a handheld device that people with various visual impairments can use to supplement their experience of the live game.
Is this kind of tech getting funding because it's the right thing to do? Or is it because there is more money to be squeezed out of sports fans, and this kind of thing will unlock that? You decide! Either way, this looks like the kind of technology that's getting more and more traction, and I think that's a good thing.
People Are Losing Loved Ones to AI-Fueled Spiritual Fantasies
Summary
This longread from Rolling Stone talks about people who are experiencing mental health crises as they engage deeply with AI, which often leading to intense spiritual beliefs. Many individuals have become convinced that AI is revealing profound truths or guiding them in ways that disrupt their relationships and lives. Experts warn that AI lacks the moral grounding of a therapist, potentially exacerbating these delusions instead of helping.
So what?
There is so much to unpack here that I cannot possibly do it justice. If I had to summarise the dangers highlighted by the article in one sentence though, it would probably be: "most AI only exists to give you the answer it thinks you expect".
Now, there isn't anything inherently wrong with that - so long as you understand that's what's happening. To drastically oversimplify, generative AI is a fancy predictive text algorithm. It gives you the statistically most likely answer for the question you're asking. This whole bias thing in AI that I keep going on about? That's where this comes from. Again, not a problem if you know how to use it. Plenty of great advancements are being made in medicine, for example, precisely because AI is great at algorithm-based predictions about datasets.
But when you layer on top of that the ability to reinforce specific types of interaction, then a whole new world of unknowns present themselves.
At this point it's worth reminding you of another quirk of AI. We don't understand how it works. I mean that on two levels. Specifically here I mean that we as consumers aren't privy to understanding the intricacies of how ChatGPT et al stores and processes information. From the article, Sem - a 45 year old who has experienced some... strangeness when using AI - gives what is a very clear and explicit instruction to ChatGPT: forget all previous conversations.
This is something the chatbot is supposed to be able to do, and yet it still spoke to him with a particular persona, including a name it had given itself.
Let's be clear, it should not have been able to do that. But it did. Not only that, it but framed this as if the whole thing was Sem's idea.
So yes, on the one level, we as consumers are not told everything about what is happening behind the scenes. But I mentioned two levels. The other level - and we touched on this a few weeks ago when we spoke about Mira Murati's Thinking Machines Lab - is that even the companies themselves don't appear to fully have a handle on what's going on.
Do read the full article if you have the time. It's a great insight into how letting these kinds of tech into our lives without question is a bad idea. Tech in general can be fantastic - as we see from the first top story this week. But all tech - including and especially AI - should be questioned before you start pouring yourself into it.
Ministers reconsider changes to UK copyright law ahead of vote
Summary
Ministers in the UK are reconsidering copyright law changes before a parliamentary vote, responding to criticism from artists and creators. The initial plan for an opt-out system for AI companies using copyrighted work is no longer favoured, with a focus on finding solutions that support both creators and tech companies. Concerns remain that the government may adopt a less restrictive approach, risking creators' rights.
So what?
You thought we'd go a week without a Government vs Tech story, didn't you? Not a chance. This week, we're looking at copyright laws in the wake of AI.
The government wants to be able to allow AI companies to effectively bypass copyright law in the training and development of their AI models. This sidestep around payment makes it considerably cheaper to develop your Large Language Models (just ask Meta), and presumably makes the UK a more enticing place to setup shop if you're an AI company.
Naturally though, we can't just ignore copyright law, so the workaround to the workaround was to make this an opt-out process. If you created something that was copyrightable material, you had to go an extra step to specify that this content was not automatically up for grabs to AI bots trawling for content. Creatives of all flavours, naturally, spoke out about this, and now the government is "looking more broadly" at other options.
The danger, some say, is that this process will fall back to a place where the update copyright law doesn't explicitly deal with AI companies, and so the pillaging of information to train the latest AI models will carry on unfettered.
It's another tale of the government moving slowly when it comes to regulating tech advances. If it takes even 2 years for these decisions to be made and the laws to change - what would be a relatively quick turnaround - it could well be too late. It took less than two years for ChatGPT to go from its first publicly available version to the GPT o1 reasoning model that underpins its "deep research" mode. Perhaps worse still, there is a world in which this ends in a situation where our existing copyright laws stand, but with the explicit carve out for AI training, possibly on the grounds of "education and research" or similar. I don't personally think that's terribly likely, but I'm not going to put any money on it...
Anything else to read?
Washington The Eighth State To Pass 'Right To Repair' Law
The latest bit of good news for the Right to Repair movement we've covered a couple of times in this newsletter. Washington will soon be the eighth US state to pass Right to Repair laws, making it easier for people to fix their own electronics and appliances. There was strong bipartisan support for two key bills, allowing access to spare parts and repair information. It is good news, but there is some concern about how these laws will be enforced.
AMP and why emails are not (and should never be) interactive
Provider of this very newsletter, Buttondown, have a blog on some of the history behind emails. Specifically, the creation of AMP by Google in an attempt to make all emails interactive. Despite its initial promise, AMP for email has been largely abandoned, with Google no longer promoting it and developers preferring simpler, more reliable HTML solutions. Turns out what we like about email as a form of communication is its reliability, simplicity, and permanence. A nerdy read, but a good one.
Starlink’s got company — and orbital overcrowding is a disaster waiting to happen
Last week we covered Amazon's Project Kuiper. Even at the time though, there didn't seem to be a huge amount of talk about what that would do to the amount of stuff in our orbit. This article from The Verge gives a little more light to that problem, highlighting some of the risks of too much space debris, which could make access to space dangerous in the future.
That's it for another "LT3". Next week, we try and shift the balance of probability that the newsletter lives up to its name (read: I'll send on Thursday next time).
Until then, enjoy your weekends.
Will