THREAD
1/
I’ve gotten quite a few messages from disabled people who benefit from AI in the same way I do but feel unable to admit to it because they are scared of backlash.
I will start by saying I understand concerns about AI, they are real. AI is energy intensive, data centres use water, a resource that is already scarce in many places, and the companies behind these products are unethical in so many ways.
2/
But something feels off in how this debate is being handled. We live inside unethical systems constantly. That is our baseline as humans in the 21st century.
3/
The aviation industry is a good example It is hugely environmentally destructive, and bound to inequality (only 10 - 11% of the world's population takes a flight in any given year, with only about 2 - 4% traveling internationally annually. Despite high passenger numbers, an estimated 80% of the global population has never flown in an airplane!) and yet we don’t generally judge people for flying. In fact travel has come to be seen as so essential that we don’t really put limits on it at all
5/
Collapsing all AI use into one immoral category doesn’t make sense to me. Frivolously chatting to it all day, repeatedly generating images for fun, or asking it to write your book is not the same as asking AI to help navigate the labour and bureaucracy of disability, or the pressures of other forms of inequality.
6/
For me the distinction is between creative and functional work. I don’t want AI to be part of the process of my creative work, but AI being involved in the functional work of managing my disability frees up space for the creative work which feels integral to my happy existence as a human being.
7/
For a bit of context, a return flight from Scotland to Spain uses roughly the same amount of energy as hundreds of thousands of substantial text only AI interactions. That’s a lifetime’s worth of pretty heavy AI use. Something, somewhere in our thinking has gotten skewed. This is not to advocate for, or excuse excessive AI use, it's to ask that judgement is proportional and accurate.
8/
I understand that drawing these stark moral lines feels very clean and very clear but I think that it can often end up protecting harmful existing heirarchies.
9/
I’m not aguing for a ‘fuck it’ attitude to AI use, not at all. We need to approach this powerful technology in a considered and careful way. It needs to be heavily regulated at the policy end too. What I’m asking people to see is that it is possible act ethically within an unethical system (there are exampels everywhere!) and that if we care about ethics we must make sure that our judgement is ethical too.
END
1. Use of AI as a disability aid is ethical IMO. And might account for 1 in a thousand (?) of current AI usage.
2. I hold discretionary plane travel to be unethical.
3. I engage in many activities that are contaminated by unethical aspects.
@skua I think my point, which I maybe didn't make well, is that we need to stop laser focusing on one thing as *the* issue. The things we want to change are systemic. I find it so incredibly exhausting to see such polarised, un-nuanced opinions over and over again.
I think it's hard to say what people use AI for, what's often amplified is the worst case stuff? That's not to say that a lot of usage isn't genuinely trash. It is.
I think excess is the general issue though.
@kristiedegaris
My own tests of genAI have it only helpful in areas where I possess high levels of competence, where I know if the genAI is outputting complete garbage, subtle garbage or useful material.
There are many areas of life where I don't have high competence. Using it there I would be relying on a grossly inadequate tool. Life is bad enough already.
you made it well.
There are people wandering around with broad brushes, doing the value signaling thing... even Cory Doctorow was recently held to task.
I ain't a fan, but I think his use isn't bad.
The key issue most detractors have is that training data was stolen by corporations running large models.
There are options. I pointed at 2 yesterday.
https://knowprose.com/2026/02/ethical-local-ai-olmo-apertus/
@kristiedegaris @skua oh, he posted in one of his blogs that he used llms for spellchecking and grammar checking. Shenanigans ensued.
I stayed out of it because of the same reason I stayed out of the baftas.lol
Science and technology have very disturbing roots here and there. I remember a friend studying medicine in the 90s asking how much of medicine came from nasty experiments in wwii as an example. Yuck.
But technology itself should not be anthropomorphized. It is wielded by humans. Humans are the issue.
Blaming all ai for bad things let's people off the hook. Ai is being used as a scapegoat by 2 sides.
Humans stole training data. Corporations profited.
We put arsonists in jail, not fire.
So it should be with every misused technology.
But those same people buy free speech in a country that permits it.
And that is the core issue.
@kristiedegaris Yes.
We discussed privately, so I am putting it publicly in your thread as well.
I also address 'Fruit of the Poisoned Tree' arguments.
https://knowprose.com/2026/02/ai-ethics-and-use-for-a-minority/
Flying is straightforward: passengers move from A to B.
What is now labeled "AI" is a conflation of a huge spectrum of tech, from well established machine learning, to slop generation, to unicorn "agentic" "AI".
There is a lot of genuinely useful applications in that spectrum. It's time for a more careful labeling.
@kristiedegaris @Jim_Graves this is a place where a nuanced view is essential. ML can scrape a ton of data and docs and find helpful links. But it's also stymied - elsewhere in this conversation it's noted that it has no expertise, and that's where it fails is customer interactions.
Sometimes customers only need help with a search. Other times they need a human with all that that entails.
@kristiedegaris One of the biggest, and most understated, problems with "AI" is the *centralisation of compute*.
AI happens in photography - NR is a DL conv.net in action - and I'm not averse to seeding a photo's description with an auto-generated paragraph either.
But what I despise is the concept of pay-to-play. Don't talk about ChatGPT or Claude, but OpenAI and Anthropic. Why should I have to pay companies $ to do what could be achieved locally because they bought the world's GPUs & RAM?
@kristiedegaris At work we talk about "data silos" - somewhat pejoratively. Compute itself needs the same discussion.
I've already dabbled in F/OSS photography, years ago. Maybe I should work a bit harder on integrating deep learning for noise-reduction & super-resolution into a smoother workflow again, see if I could get rid of DxO... That would be an intriguing way to go.
@kristiedegaris It can be quite easy - you can have your LLM of choice generate a python script that will train a model on the difference between lanczos4 upscaled images vs ground truth - effectively it becomes "artefact removal". Similar for noise - put it in, train the differences, now you have a tool for removing it. Etc.
Trouble is the amount of GPU required for the training or running on a serious photo is... significant.
It’s the societies we live in that disable us, so IMO using unethical tools in society to alleviate suffering is justified
We didn’t make these tools unethical, huge corporations did - the responsibility is theirs
BP & Ogilvy made the most successful ad campaign of all time to coin “carbon footprint” to push the blame on us
But one decision from the chair of BP will do more than I ever could. Scale matters.
Ppl do what you can, but don’t martyr yourselves 💜
@LillyHerself I began using AI around a year ago, there was twenty years of effort for myself beforehand, and four years before AI for my daughter. I almost laughed at your comment, you think we didn't try all you suggested and more?
Judging what you think other people should do purely based on what worked for you isn't a great framework. I generally assume people worked as hard as I did, problem solved as well as I did, asked for help as well as I did and THEN resorted to the robots.
@kristiedegaris Your flying example is so useful.
We need to clamber out of the developer sandbox we’re trapped in, because energy use, human exploitation and social impact really matter here too. We need to get to the point where people can evaluate and then limit their user-side AI needs in informed ways. We need far higher ethical and transparency standards on the corporate side. We need the courage to limit use and drive up safety planning.
We expect planes and flying to be held to safety standards. We exert some degrees of consumer pressure on pricing and reliability. I suspect many of us who fly for reasons of geography also offset if we can, and limit our use to try to minimise harm. We think about the planet and each other.
Let’s get to this point with AI, and stop accepting that our only option is to provide endless free sandbox testing for hypesellers.
@kristiedegaris Having just read this (thanks to @koutropoulos), I’m imagining it with Qantas instead of Anthropic. Imagine an airline famous for its safety record saying it’s “dropped its safety pledge” for commercial and competitive reasons. How do we make this similarly unthinkable for AI? One way is to exert consumer and media pressure on companies that say things like this:
“We didn't really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments ... if competitors are blazing ahead”
Just adding another thought, because the flying comparison really has got me thinking.
As a reluctant flyer, I don’t expect to be picked up and flown somewhere when I’m least expecting it. I’m very tired of AI showing up uninvited.
BUT (and this is what you have helped me see) as a reluctant flyer I’m still the trigger for a lot of hidden flying: food miles, imported products etc. And that leads to all the hidden exploitative labour that produces the things that are flown here, and is the reason why they have to be flown. So this is where I want higher standards of corporate transparency in AI as in carbon impact. I want to know what I’m costing the planet and other people when something appears as a convenience for me.
We’re (slowly) doing it with plastic, let’s do it with AI.
@kate I'd say that with flying, people are rightly focused on environmental stuff, but the travel industry does way more harm than just environment. In fact, it causes vast harm. So imo it's about way more than being upfront about carbon.
And with regards to AI, I do think the changes will come, more ethical AIs will emerge, more ethical companies and we will have more choice. But that may not be enough to undo the harm they cause. I don't know.
@janeishly I think intellectually most people would agree but the backlash is extremely vocal for AI, less so for flights. People aren't so scared of backlash when taking flights that they are hiding the fact they take them?
Maybe more people on Mastodon understand but generally there is a huge backlash to AI usage while people take 6 European mini breaks a year. I haven't taken a flight in almost 10 years now, but I understand why people do. I advocate for far more considered use of flights.
@kristiedegaris a thousand times this.
Which is what makes me so incredibly angry about the snake oil peddlers.
AI is fantastic at the boring stuff as all computers are. They're fantastic at bureaucracy and memory and speed.
They have no soul and are not actually life, incapable of joy or meaning.
---
Agreeing:
We should be focusing on these benefits to humanity instead of trying to strip mine humanity into labor and consumption.
Kristie
opalideas
skua
Taran Rampersad
zettpunkt
tante
CelloMom On Cars
Jim Graves
Wayne Werner
Tim ☑️ 🔭🌃📷🪨🚴🌳
gorsefan
LillyLyle/Count Melancholia
Kate Bowles
Janeishly