Box089: 2+2 is most likely followed by ‘=4’

25.01.23

DS079: Decided (following a conversation with a friend) to go back to doing digital sketches. Staying away from animation because that 10x’s the time so probably just images cos I can knock one out in about an hour. Here’s a completely non-AI, non-NFT bit of bad art making for you, for free.

I really enjoyed Gary Marcus on the Ezra Klein podcast. Marcus is a significant figure in AI having launched quite a few companies but is also a good, measured cynic. I found myself nodding along because his over-arching thesis is exactly right (to me, at least); AI is great, could do amazing things, but we keep going for low-hanging fruit and wow-effect rather than celebrating and focussing that are actually going to deal with the massive problems we have. Eg. Why has ChatGPT got so much bang while protein folding is basically received with a whimper?

My own test. At least it’s consistent in its inaccuracy.

He also made a really compelling historical point about the prevalence of neural nets: They’re very good at probabilistic reasoning and seeing patterns, but in most instances where we interact with computers we want symbolic reasoning. We want 2+2 to always be 4, not 2+2 to most likely be 4. Neural nets can be super useful for things like speech recognition where most-likeliness is a useful heuristic. Bit in the case of, for instance, driving a car (MWSTLOT*) down a highway, you don’t want ‘most likely’; you want thing-equals-thing every time. Marcus makes the point that most of the systems that we have come to rely on in computation such as databases, word processing, spreadsheets, emails etc. are reliable because they’re symbolic; you get the same thing every time you do a massive bit of data crunching in Excel because the numbers always mean the same thing.

*May We See The Last of Them

A speculation on the next few years of AI vapourware.

Microsoft laid off 10,000 employees last week in parallel with a rumour that it was upping its investment in not-at-all-open OpenAI to $10bn. Commentators have noted how the emergence of Large Language Models might be spooking search engines because suddenly ‘How long should I cook sausages for?‘ might just give you a straight-up answer rather than linking to a lengthy blog with loads of embedded ads or that the first few pages of Google might just become SEO-optimised generated crap and undermine trust and usability. Google has BERT and LaMDA – two of the biggest and thus (by the prevailing rubric of the Unreasonable Effectiveness of Data) – most ‘effective’ LLMs but won’t release them ‘to the wild’ because the risk to reputation is so great and because, in the post-Snowden, post-Cambridge Analytica years, the big tech companies take risky bets with data and new technology seriously.

The sudden threat of ‘upstart’ companies like not-at-all-open OpenAI ‘disrupting’ the GAFAM Empire has sent them into overdrive mode ‘ …and shaken Google out of its routine. Mr. Pichai declared a “code red,” upending existing plans and jump-starting A.I. development. Google now intends to unveil more than 20 new products and demonstrate a version of its search engine with chatbot features this year…’ The language implies Alphabet being reactive and war-like saying that it ‘…could deploy its arsenal of A.I. to stay competitive.’

This massive turning of the ship comes amidst a post-Covid definitely-is-a-recession where the big tech giants are laying off staff and buckling down for a slower market. Google’s parent – Alphabet – laid-off 12,000 people (roughly 6% of workers) from senior leaders to new starters last week with more coming for non-US employees.

So, we have a landscape where, because of global conditions, there’s a shrinking market for tech products, web services and advertising which is the revenue that underpins GAFAM. At the same time we have new, smaller AI companies moving at remarkable speed but with untested and unregulated advances in LLMs, diffusion and transformers. Unlike the previous boom around GANs, these new advances are much more ‘realistic’ and usable and so have had enormous public impact and significantly raised expectations amongst consumers. But, as we know, these new companies don’t (or no longer) have the same guardrails or hesitancy about putting untested and unregulated products in the public domain to build hype.

So, my speculation: In response to the new threats and shrinking economy, big tech are forced to compete by lowering the guard rails and releasing more of their unfinished, unpolished and untested products into the wild to grab attention in a dwindling market place (we’ve already seen this with Google’s announcement of 20 new product is in 2023). The imperative becomes ‘speed of release,’ not ‘quality of product’ and we end up in a race-to-the-bottom of badly designed, ill-conceived and rushed AI that could not only cause social harm but also damage the actual science by distracting resource and attention from meaningful and useful AI projects. We end up with years of AI vapourware and overblown hype waves.

The ideal scenario would be governments developing clear-eyed mission-oriented AI research strategies and agendas aimed at tackling the climate crisis. A sustainable vision for investment that provides a long-term framework for the tech industry to regather and rebuild around meaningful applications of AI, carefully released and tested with good policy frameworks for support. You get strategically well-considered products and services that actually improve quality of life and start to actually positively the things people care about rather than distract for them. Or, as Gary Marcus put it, a CERN for AI.

What are the chances?

Another prediction: Node-based AI

Part of the Google thing is the development of second-order tools: In other words, tools that help people who aren’t computer scientists make their own AI, like how Photoshop allowed people who aren’t computer scientists to do computer graphics. There’s a couple out there swimming about or there are plugins for eg. Unity that connect an API up. But the real thing is going to be the first node-based AI software that allows you to home-brew AI really easily. Of the ‘I need an AI that can do this specific function real quick.’ Like single-task, issue-based interactions.

To think how easy the kids are going to have it! I had to painfully trawl through ml4a to get shonky S2S neural nets to play nice!

PhD

Poured a good 45/50 hours into that bucket last week and all I got was a measly, disjointed argument. Little bit deflating but the flow was enjoyable. I’ll give you a bigger slice of that particular pie once I’m happier with it. Back to eaking out the small hours for a little bit.

Short Stuff

How far could you travel by different means of transport with a limited carbon budget?
Tobias Rees tweeted about his kids using ChatGPT and yes, it brought me cheer. I am not doomy on the tech itself, just on the hysteria around the tech. There is something amazing about a tool that opens up imaginative space for folks.

I need to turn my mental ship around myself. I get so angry about bad AI hysteria and I need to move through that stage into doing something about it. No-one’s going to read my PhD after all and really the only thing that matters is 1.5 degrees so how can I bring the two together? I’m in good spirits generally at the moment so need to use that as a lever to get over this hump. Love you.

Box088: Cracking the Back

11.01.23

I got out of hospital after just over 48 hours which was way better than I could have hoped. And I was back at my desk on Jan 3rd. If you think that last time I was mostly in bed and struggling to sit up straight for three weeks it’s pretty remarkable. But the long tail of this recovery is going to be, well, longer. I can’t put weight on my leg except in the most ginger of circumstances, am basically confined to my flat and am being carefully monitored by the NHS to make sure my bones stick together. So I don’t have much more to write about other than PhD stuff this week.

PhD

This week has been about cracking on with one of the substantial chapters. Meanwhile I also try and stay on top of the reading by reading one more paper or article than I add to the Zotero library. I’m not sure if keeping a note of them here would be interesting and useful for you or not really. I’ve developed a pretty good feel for crunching through academic papers now.

Trying to hold it all in my head and avoid the thesis or anything else drifting. Words to live by.

I’ve spent the last week diving into one of the main chapters of the PhD to try and get it ready for upgrade but I’m really struggling to structure the sub-sections. I’m working on the chapter about Spectacles, Performances and Demonstrations. Seems straightforward enough; promotional videos, sci-fi, advertising, marketing, public demos etc.; all the ways AI is represented. But each of those is massive! I could write a whole PhD about the advertising language of AI. Ok, so focus on the overarching question; how does the design shape imaginaries of AI and thus foreclose futures? But then it’s just describing and critiquing the design of some adverts, the design of some promotional films and doesn’t really hit at any insights. So, what about instead of looking at forms, platforms or media we look at forms of representation or narratives. For instance, the idea of AI development as a David and Goliath story, or how they are shown as more than or less than human; either superhuman or human servants. Or how massive corporations are repackaged as plucky little upstarts through AI? I look back at the paragraph describing the chapter in the introduction:

Spectacles, Performances and Demonstrations addresses the long history of AI as a spectacle since its earliest inception. Highly public performance through games such as chess, go, the game show Jeopardy and so on, as well as sensational stories about exaggerated postive test results in highly contained settings serve to inflate expectations about AI and buoy public and financial interest. Design is heavily implicated in the orchestration and production of these showcases but also has a role in exposing the fallacy of positive performances through strategies of satire and reverse-engineering to show how the trick is done.

And it makes total sense but actually getting into structuring it is tough. I started trying to use Miro again to structure connections but again, I could see the remit of the chapter quickly exploding and spiralling. Perhaps I just need to go for it and then scale back. Problem is, that’s where I started 3 or 4 months ago.

In my head it’s like trying to lace all these ways that I’ve cut, sliced and layered the ideas, thing I’ve read and concepts together into a cohesive and original argument. Here; I tried to draw where my head is at:

At the top is all the reading and research; my own practice, other’s, books, papers in the Zotero library. They’re not directly connected to each other because I’ve assembled them but some are very deep and some are very broad.Next is ideas (effectively tags in Zotero); things like ‘obfuscation,’ ‘non-human,’ ‘epistemology,’ etc. They don’t mean anything in themselves but are a vocabulary for the whole PhD. Then it’s concepts which is mostly what’s happening in the Miro Brane; big, broader, *interesting* and *curious* recurring relationships or similarities or groupings of ideas that seem to relate or impact each other. Then it’s theories; essentially the chapter and structuring rubric of the PhD; critical futures, sociology of expectations, critical practice and so on. As well as the chapter titles I suppose. The acting of ideas on the material world and vice versa. And then it’s the trickiest part; the actual writing. The linear, top-left-to-bottom-right, clearly evidenced and logical distillation of all the above layers.

So I spent most of the week’s work trying to crack away at this problem and have managed to get to some subdomains and some structure: On the right is an example sub-section as it is now. The content is a collage of sentences, paragraphs and notions from the previous draft and pertinent quotes I’ve dropped in where I’ve remembered or bothered. At the top is the what, why and how of this section. It’s so hard not to let it spiral and end up writing about everything. The challenge is keeping it tight, each sub-sections could conceivably be a whole PhD!

Orange bits are milestones I guess; where does the last section lead in to this, what’s the lead out; narrative markers. Green bits are things I need to dig around for more on. Yellow bits are key ideas or arguments. Blue bits are examples of practice to talk about. And that’s the system; I just went over and over and over and over it again to work down to sub-sections of the chapter and I still hate it. Like I said, each one could be it’s own PhD and I don’t see yet how they actually do more than just describe things. I guess that’s because they’re mostly made of other people’s quotations so far.

One of the things that is helping is I’m keeping three other PhDs/Theses (I’ll never tell who’s) that I really like open nearby me all the time and every time I get stuck about how to go about what I’m doing I just open a random page and read a bit of one of them to try and get a feel again of how to do this damn thing.

So this is a pretty intense spurt of work, post-recovery getting back into the swing of my mornings hasn’t been hard and the work/work/work/life balance is ok (though fuck me I miss bikes so much). I’m feeling good about getting this chapter in for at least end of January in some form or another and (as per lot’s of people’s advice on Instagram) I can’t be precious about it. Do it, get it out, get feedback and this one is important because this is the bulk of the project; not the intro, not the context; the thing in itself.

Short Stuff

Worshipping at the Altar of Artificial Intelligence, Jessica DeFino. Interesting to read about the impacts of recent advances on beauty and aesthetics; slightly outside my wheelhouse.
Why ‘eat the rich’ will be the dominant creative narrative of 2023. I’ve been wondering recently whether the surge in aspirational imagery (young men in particular in sports cars and big beach houses, usually attached to a trading scheme) is a normal vision for most people. I guess not.

I feel like I work harder than ever at the moment. Reading and writing all morning, nine hours of work and intellectual labour in the middle, couple of articles or a book in the evening and yet I feel stupider than ever; the cascade of information and stuff just gets bigger the closer you get to it.

A final tip (which may or may not be useful) as I sit here trying to pull this chapter together: Don’t look up. If I start by thinking about the enormity of the overall project I get paralysed and prevaricate: Have super simple, short and easily achievable steps.

Ok, I love you, obviously.

Box087: Oh, What’s the Use?

31.12.22

I’ve been out of hospital a few days now, largely immobile and struggling once again with the feeling of uselessness and helplessness. I’m trying to close out the ‘To Read’ list by the 3rd so that I can five in to writing the next PhD chapter ASAP as I’ve sort of set a deadline of mid-Jan for that. I also don’t want this post, which has been open fo two weeks following me into 2023 so I’m shoving it out the door now.

I worry there’s too much opining on this year’s AI explosions going on. Unfortunately my research puts me in the position of needing to engage with a lot of it and also be one of the voices in the echo chamber. Part of this is entirely self-reflexive as me working out my practice and my relationship with the subject. Unlike crypto, I do think AI has incredible potential: AI is doing good stuff quite well. There are now 18 medicines that were developed over the year using AI, it managed to do most of the legwork on solving protein folding and AI was used to successfully identify Covid variants. AI, in applied situations, does seem to live up to most of the hype or at least the hype has calmed down and the adults are taking over.

Noah Smith’s early Decembers letter made some very good, very rational arguments about AI and the near future. With roon, he points out that AI is profoundly unlikely to result in mass unemployment, something I fully agree with simply because at a purely simple level – what good to anyone is mass unemployment? Why would anyone want to just manage algorithms? Who would buy all the stuff? However, the whole argument is still framed through a productivity/efficiency mindset. As long as exploitative capitalism produces AI, it will reproduce exploitative capitalism. Smith and roon describe the idea that AI will probably become ‘autocomplete for everything’ which I think is just about right and something I would actually enjoy: autocompleted code, blogs, emails, texts, micro-tasks, meetings, note-taking, copying from one document to another and so on.

Smith and roon very briefly touch on the opportunity that: ‘The increased wealth that AI delivers to society should allow us to afford more leisure time for our creative hobbies.’ But this is the bit that’s missing. AI should result in more leisure time, a shorter working week, more diversity of activity and experience and joy in people’s lives. How will all this productivity lead to that? Given everything we know and have experienced why wouldn’t AI just result in even greater profit being fed to tech leaders as they massively increase wealth inequality? Why would the next big AI tech companies not just operate like Amazon and use every new innovation to squeeze their workers and customers even harder for Bezos’ space boner?

These people want to automate creativity, the bastion of anti-capitalist work; slow, caring, experiential, personal, emotional. That should tell you everything you need to know about what the world of AI work will look like.

The problem of criticality is another dimension. If we are to go about auto-generating content, be it code, art or text then do we risk losing critical capacity? I would say that critical practice is about being able to and actively thinking about building or making things to understand how other people think about building and making things so that we can better understand why we have the things we have built and made and how they instantiate certain ways of thinking. (Why car-centric cities, why a drive towards productivity-driven automation, why the sexualisation of automated assistants, why a language of the occult to describe computation, why obfuscate labour in automation?) If you’re putting all that making on a machine then can you still think critically at the same level about what you’re doing and why?

No probably not, but maybe you don’t need to. In the mid-2000s, Apple were making Macs cool and I was at art school and everyone wanted one and they shipped with Garage Band. The music was amazing. You could start a band on a whim and produce a demo track super quick. I wrote so many drum lines with no experience of drumming at all or any theoretical understanding of music (it was all punk bands to be fair). The feeling of the music coming out then (now coyly termed ‘indie sleaze’) was magical – anyone and everyone could do it without a stuffy sound engineer giving you a hard time about compression. There are real parallels to the narrative around generative AI now and how it is broadening access to creativity. Garage Band didn’t kill drummers or rich, critical and experimental music and I doubt AI will kill critical practice either. For those that find that important, we’ll keep doing it.

AI could, speculatively, liberate people from work, rebalance wealth, move us away from materially and labour extractive capitalism and towards community-oriented life, giving us time to create, explore, care and flourish. But the very fact that the big breakthrough this year has been about automating creativity (something people actually enjoy doing, but managers and tech people find an annoying expense to be treated with disdain) shows that it is most likely going to reinforce and deepen existing capitalism. AI is no more ‘democratising art’ than Garage Band ‘democratised’ music. Apple and MySpace made all the money from that wave as Microsoft, Open AI and Stability AI and their VC’s will make all the money from generative AI.

Metwhatnow

I had a bunch of stuff about the decline and fall of the metaverse written before going to hospital but I realise now that it was quite gleeful and bitter so I’ve just made a list of stuff:

Cory Doctorow on Facebook (‘Meta’s’) pivot-to-video and the fradulent boosterism of Facebook’s (sorry ,’Meta’s’) aggressive acquisitions towards a failed metaverse: The EU held a metaverse launch party where 6 people showed up – Decentreland has 38 active users but claims it has 6000 for its $1bn valuation. Facebook’s (sorry, ‘Meta’s’) gamble has famously, not paid off either – even their own employees don’t like it. This is probably why Apple is keeping a pretty low-key metaverse presence – they’re not stupid and they actually know how to make grotesque amount of money. Are failures in research infrastructure to blame for web3 and the metaverse? Yes probably.

For 2023

Just some notes of things I had to take into next year:

Take things at face value. If it looks like a duck, walks like a duck etc. The current vibe has a lot of people casting about to get a feel for what’s coming next and just as many people willing to hoik their nonsense to turn a quick buck. A lot of things that appear stupid and nonsensical are in fact stupid and nonsensical, as this year’s crypto and metaverse crash have shown.
Saw Hamilton twice this year. Great every time.
I’m going to have to experiment with image generators / diffusion models. It can’t be avoided anymore so I just need to think about a way to do it that makes it a worthwhile endeavour.
There are about twenty things that happened in my life that just constantly recur – snippets of conversation, a thing someone said in passing, an experience, an interaction. They’re apparently random, no more meaningful than anything else, but I remember them all the time and they have an outsize role in how I think about myself and my life.
The only aim of the wealthy is to make more money and avoid paying tax. Any time any tech company issues any big proclamation about the future or democracy, climate or experience or whatever, just know that at the core of the implication wheel is that ambition.
I discovered that a part of my brain has been sitting idle just holding on to all the lyrics for Million Dead’s I Am The Party, which I was able to sing verbatim, having not even listened to it for maybe ten, fifteen years. What else have I got in storage just taking up space?
Paying more tax and engaging meaningfully with politics would fix most of the problems they want to innovate in and profit from; climate, public transport, education, equality etc.
Would this innovation be better as you paying tax and supporting access to voting?
It really is the sunset of social networks. Goodbye and thanks for all the whatevers. That could mean a lot of introspection but I doubt it.

Short stuff

Sorry these are really badly structured and credited because I was copying stuff on my phone from a hospital bed and it’s hard to do that in a good way come on AI sort it out.

Timnit Gebru, similar to above, on the role of Effective Altruism in AI safety.
Linked this above but a deeper dive on the nihilistic disdain AI people have for human creativity.
Playbook for integrating climate change into your stories from Good Energy.
The harm from worrying about climate change from the BBC and very much that hope is the answer (when well tempered with real dread.)
Shumon Basar ruminating on Endcore.

Ok, love you, you know that. Sometimes it’s hard to because I get so wrapped up in my own fears and worries and get quite selfish and self-important. I’m not very good at thinking about other people outside of transactional circumstances which is why I’m terrible at retaining friendships and I don’t know if I can fix that now. But I do love you and sometimes I suddenly feel it and remember and it sort of makes everything else quite pale.