🤷 No Data
Christmas AI for supermarkets | MLMs are running out of data | How to 3D print a moon base
Hello — last night I had a dream that an expensive hotel I was staying in destroyed itself because it had too many complex indoor water features. What the SHIT does that mean? Help me.
This week was sort of okay I guess? 🤷. We all love a good shrug every now and again.
How are supermarkets going to invade your privacy during this holiday season? Scroll down to find out
We’re running out of data, quick make more!
NASA has funded a project to help astronauts self-sustain themselves on… the moon.
🫂 Using AI is ALWAYS cheaper than caring
This week, a WaPo article about AI in supermarkets caught my eye. The AI that supermarkets use is for all the reasons you might expect: to ‘predict’ demand, and ultimately to cut costs. I just want to examine these two apparent benefits to show you how they aren’t benefits at all (shocking). Let’s see…
Apparent benefit ONE: predicting demand is good, because you can figure out how much stock you’re going to need ahead of certain events.
More likely real-world outcome: actually, one of the outcomes of this is less food waste, which is obviously good. What this WaPo articles manages to so elegantly glaze-over is that predicting what people want requires heaps of personal data, and — a classic yet unsolved problem — AI is extremely biased. So what we get is:
Predictive models drawing from data that should be irrelevant to, for instance, the kinds of foods that people might want to buy over christmas
The kinds of datapoints I’m talking about here are ones that you might expect to grab from customer loyalty schemes and, uh, your local data broker. E.g. postcode, date of birth.
With data like that, supermarkets can go one step further and use the AI to figure out who likes to eat what — and then charge more to richer demographics. So essentially, you have a machine deciding which foods get to be affordable and which don’t — excellent.
Apparent benefit TWO: when you accurately predict demand, you save money on storing surplus.
More likely real-world outcome: this is where we must ask: but how much does it cost to implement this AI? Furthermore, this method of having the exact right amount of stock and not an ounce more is exactly what led to all of the supply chain problems we have now, in this pandemic. In fact, the problems were already there — the pandemic caused demand to shoot up, and no one was ready, because the supply chain system is wound tightly to work only one way. AI won’t fix this — supply chains simply need to have more redundancy built in (and, I dunno, maybe be shorter than a million miles long?)
If you would like to read more about how AI is being over funded and over used in the name of social good, read this excellent piece by Timnit Gebru, which essentially shits all over effective altruists. Btw if you are an EA, you can literally hit unsubscribe right now — do not be fooled in thinking that any of my opinions align with yours.
☁️ The dream of data as an infinite resource is over
Speaking of AI systems which are useless at best, and harmful at worst: some researchers have found that the data we use to train things like large language models could ‘run out’ by 2026.
Oh no, quick! Everyone start churning out photographs, illustrations, and blog posts FAST. The likes of Midjourney and OpenAI and Stable Diffusion are running out of free content to scrape from the internet and feed into their machines which are going to save us from from ourselves in the future — because AI is the ONLY thing that can do that!
You think I’m messing around, but I’m serious. We need to keep producing precious data to train the machines that will one day bring humanity back from the brink. There’s literally no other way. We are hurtling ourselves towards destruction by wasting money and resources on training machines to either mine digital gold or draw silly pictures for our entertainment. These machines will soon outperform software engineers, and even writers like ME. Then we won’t have jobs, or anything to eat. Is that what you want?
The only thing we can really do is double down and hope that that AI will become smart enough to fix all this and show us mercy. Look, here I am chatting to the ChatGPT and asking it how AI will save us. What have you done lately??
🌝 ‘Darling the moon base called again. They’re asking if they can borrow some daylight, gravity, and a sense of community’
ICON, a 3D-printing startup based in Texas, has just been given many millions of dollars by NASA to build a construction site on the moon. The idea is to get an operation going so that astronauts can actually stay on the moon and be self-sufficient. Tbh I prefer Elon Musks idea, where he sends people to Mars, not to live, but to DIE — and it least Mars isn’t totally monochrome and pitifully orbiting the Earth, a planet where only losers live.
There’s definitely been a theme this week, and it’s this: machines are better than humans; there’s nothing wrong with life on earth; lets explore others worlds.
Here’s what I think, bulleted for your pleasure:
There are privileged people out there (people who work for NASA) who are bored of how easy everything is on this planet, and need to experience hardship in a dusty grey desert for a bit
Humanity is through with rewarding poor people with waged labour — constructing buildings, whether on the moon or not, is now a machine’s job. The jobless can stay on earth while the rest explore the universe.
I’m confused about how a large-scale 3D printer set up on the moon is going to help anyone do anything, including ‘explore new worlds’.
I’m sure space is great, but if the parameters of exploration are to live somewhere with very little light and no breathable air, why not try going under the sea, like in the computer game Bioshock? You don’t have to leave Earth for that!
That’s it for this week, please continually refresh your inbox until next weeks issue arrives, like a good little fan.