Sweet & Salty AI
September scepticism & privilege.
Hello. Can I get a Big Mac, no pickles,, Extra Ketchup… uhh medium french fries and also one Extra large Sprite. That will be all, thanks.
Sorry, I just felt like starting with something trivial, frivolous, silly. Like I've been threatening myself to write a piece for a while now- I feel silliness is underrated.
Take a look at this ridiculous McDonald's Order Spotify playlist!
An AI pinch of salt: September scepticism dispatch.
Friday Find: Bad Desk! (Or, missing Logan Roy).
Say What? A quote on cohorts.
Privilege Expansion: AI-positive for all.
With: feminism in Indian horror, a Singaporean X-Men mutant, a speech filter.
1. An AI pinch of salt- the September dispatch
Many of you who read my thoughts here (and elsewhere), will know that I am often a tech optimist, and generally open to the next big (or small) appearance, changing the way we do things, reinventing and letting tech find us paths we may not have before.
With AI, my 2024 in particular has been a case of straddling two sides of a divide that is, depending on the day of the week, a strange or fascinating or amusing one. While it boggles my mind that many are happy to entirely trash the space (”look, snake oil!”), equally I tread with caution around unabashed glorifiers (”look, humanity is saved!”). Its a bit remarkable how I will, on a single day, read a piece that the AI space is still under-hyped, and another which says the over-hype is real.
In general, I feel that in our relationship with new technologies, we have become habituated into being either starry-eyed cheerleaders or disdainful denouncers. As with so much, the real truths, the real narratives, the real progress lies somewhere in the middle. ( I suppose this is symptomatic of a wider trend of being polarised, but that is for another day).
Round Up
Having said all that, we must not buy into all the glowing hype without pause. Every few weeks, I might choose to offer you that chance to take a minute and consider less unequivocal stories from either side of this coded fence. (like here and here)
Take for example, Alexandr Wang and his company Scale AI. He became one of the world’s youngest self-made billionaires by building a sprawling army of more than 100,000 contractors, who perform the grunt work that powers the modern AI boom.(WSJ The 27-Year-Old Billionaire Whose Army Does AI’s Dirty Work, paywall). The job of these vast numbers of anonymous workers ‘data labelling’. This calls for summarising articles, writing stories, labelling images, and building sentences for chatbots so they “have the text they need to better understand human speech patterns.”
These workers sometimes try to game the system for their own benefit.
“When Meta researchers received the data, they spotted something odd. Many answers sounded the same, or began with the phrase “as an AI language model…”. It turns out the contractors had used ChatGPT to write-up their responses- a complete violation of Scale’s raison d’être, something that will accelerate model collapse.”
Proponents of AI claim that it will supercharge productivity growth, something I have often contested for the U.S. and other developed countries. But even if I am wrong, when you add in the data labelling work, and the incentives to game the system, the positive impact on productivity growth may not be large.
A question of water. Teaming up with Generative AI, it would appear, is thirsty work. Let’s all watch- in slow motion- a bottle of water being poured down the drain every time we ask ChatGPT to draft another note, idea, analysis for us. Research shared in The Washington Post says it uses a bottle of water to generate one email. (The hidden environmental costs of using AI chatbots, paywall).
Written by no one to communicate nothing.
Lets look at the consideration about the pollution of our information base, that great freeing democratising thing we call the internet. An inevitable deluge of AI-generated content is set to be unleashed by marketers, bots, and a million others who will grab the chance to spend mere cents to make dollars.
I err. This has already begun.
For instance, analysis released in Variety tells us 2/3rd of GenAI content is in the likeness of talent or IP. Two-thirds. In 2023, about 40% of the total amount of generative content on the public internet contained either specific celebrity likeness or recognisable IP. This is tracking to increase to 67% this and next year (from). That sounds like a content, legal and ethical mess.
Add to it this the little story about an open source research project that has been abandoned because the data out there is no good anymore. ‘Wordfreq’ scraped the internet to determine the ever-changing popularity of different words in human language usage. Its creator, Robyn Speer, has said they are sunsetting the project because generative AI spam has poisoned the internet to a level where the project no longer has any utility. “The web at large is full of slop generated by large language models, written by no one to communicate nothing.”
From Christopher Rice:
Generative AI is best understood as a form of pollution, not as some amazing new thing that will somehow magically increase productivity or somehow learn to reason for itself. It’s like an oil spill on a global scale, or microplastics in the rain, the soil, the flora, and the fauna, never to depart. Or if you speak Tech Bro, it’s like Starlink satellites messing up our ability to conduct astronomy research by messing up the night sky and leaking certain frequencies.
Jim Covello is the Goldman Sachs head of global equity research. Back in July he said, “Replacing low-wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I’ve witnessed in my 30 years of closely following the tech industry.” He has weighed in again in a NYT piece that now calls him Wall Street’s leading AI sceptic. (Read an extract here.)
Overbuilding things the world doesn’t have use for, or is not ready for, typically ends badly.
If you still want more, here is The Honest Broker with another more… blunt, shall we say, take on AI, with predictions for the next 24 months to boot.
But if you choose to click on one link, then maybe this abstract and clever little piece, this might be it. Here, eat this by Kyle T Webster starts like this. Do read the rest.
Here eat this.
What is it?
It's new.
Just eat it.
What's it made from?
Everything.
How can it be made from everything?
It just is.
Don’t worry about that.
Eat it.
It's great.
It's new.
2. Friday Find: Bad Desk!
Did you bump into Logan Roy online this week? If not, here he is in all his cold, abrasive glory!
Ok, its not Logan Roy (for those who never got into the world of HBO’s Succession, that’s the wealthy, controlling, self-obsessed, often despicable patriarch at the centre of that media-corporate-family saga). Here is Brian Cox channeling Logan liberally, while delivering a decidedly not self-centred message in this spot which promotes... well, you'll see.
Also, f**k bananas.
· Directed by Peter Franklyn Banks · for Golin ·
3. Say What? On cohorts
I recently mentioned how I always approach generational demographic groups with some care, because ‘I am always hesitant about how demographic boundaries- and traits- are defined so precisely, grouping entire generations as one profile’.
(Please hold while I take a moment to notice that I just quoted myself).
Having said that, research around generational groups and how they think, act, like, love, hate and get inspired… can often be fascinating. Because even if there are generalisations, there are often kernels of truth, relatable nuggets, or worthy oddities.
But the point remains that these generational definitions and ‘cohorts’ are sometimes shorthand for stereotypes. This from the inimitable Sir John Hegarty doesn’t hold back:
…cohorts are made up. Relying on generational stereotypes is a trap for brands. If your strategy is based on a truth that is widely known (or at least, believed), it becomes harder to create something with difference. Your customer is idiosyncratic, unique, and nuanced. There’s a word for a practice that assumes characteristics about people based on their birthdays. It isn’t science: it’s astrology.
Ouch.
4. Privilege Expansion
More AI! Some thoughts spurred by another excellent take from
, who writes the Digital Native newsletter. I definitely recommend reading the piece, but here is a TLDR.Extolling the formula- Expensive Human-Centric Service + AI = Better Access & Affordability, he looks at how AI-driven ‘Privilege Expansion’ can open up access to services that were once exclusive or expensive. From education and healthcare to personal styling and interior design, AI can replace the human element in these fields, offering more affordable and personalised experiences. Examples include AI tutors, telehealth bots, and even AI companions that can make elite experiences available to all.
This ‘privilege expansion’ is possibly the most promising aspect of AI. It is also the most positive scenario from humanity’s perspective, which unfortunately is not always necessarily married to the best scenario for businesses. Lets hope that’s less true here.
I must admit though, the articulation around ‘removing human barriers’ sits uncomfortably by me. I understand and take the meaning, but I am less eager to embrace it as a recurrent path to betterment. It is predicated on a deeply individualistic and dareisay, deeply modern capitalist worldview that presumes humanity’s progress is only in the hands of corporations and entrepreneurs (who will, of course, create brilliant new businesses for the greater good).
Example. Education, most excitingly ripe for disruption/big leaps with AI. He says, ”the barrier has always been the human”, and the “teacher is the bottleneck”, while at the same time acknowledging, “technology can’t replace the human-level engagement of a good teacher or tutor”. In a newly AI-opened world, I’d offer that the last bit- a human tutor, say- stands to become a differentiator, a thing of great value, a privilege. Our dominant economic and socio-political systems will always find a way to create privilege.
With travel, personal fashion and interior design, it really will depend on how personal the personalisation can actually be. So much of personalisation has thus far been reduced to algorithm-driven commonalities. Imagine a travel experience generated for me by an intelligence that settles toward the median? I shudder. (But then, I have never used a travel agent offline or online to make my itinerary)
‘Friendship’ really should not enter this discourse, even as a casual example. There is a data point about how “in the last 30 years, the percentage of people with <=1 close friend has nearly tripled to 20% of the population” (in the US). I am not aware of the depths of the study, but it could certainly be argued that modern technologies have exacerbated that decline, not mitigated it.
Relatedly, on the attention economy. “We see consumers spending hours on Character and similar products, chatting with AI companions and friends. That’s a massive shift in attention.”
There is reams written and hours spoken on how the attention economy is deeply flawed, contributing more to bottomlines than humanity’s betterment. We undeniably have made missteps in the space, particularly with social media.
Will AI make it easier to disrupt the current attention economy? Maybe.
Will it be for the better? I will wait a long while before I swallow that pill.
~ · ~
PS. A pedantic grouse- Bong Joon-ho’s Parasite was edited on FCP 7, not the current $300 Final Cut Pro.
5. Masala Peanuts
(where I share stories or tidbits I find interesting)
Read. The unlikely feminism of Indian horror comedies. The article ‘Women in White’ looks at the (very successful) horror-comedy genre in India in the light of franchises like Stree.
Know. Last month, Marvel introduced a new X-Men mutant named Jitter, who's a Singaporean teen! Sofia Yong aka Jitter’s mutant power is hyper-focus, which means Sofia “can do anything she sets her mind to – accessing talents and skills most people train their lives to develop
Think. A sound lens to look at all our interactions, this is a ‘triple filter speech framework’. “Whenever you feel the urge to say anything, pause and put it through three filters”. What I find most fascinating though, is the areas where two but not all of these conditions are met.
Off for an ale.