Weekend reading: What the Druck? AI slop goes mainstream
Previous post: The way. (Or, why we invest)
≡ Menu
The Latest Articles
Essential Reading
Better Investing
Property Posts
Get Money Motivated
Monevator is a place for my thoughts on money and investing. Read our disclaimer. Subscribe to get all our free posts via email. Follow us on Twitter, Facebook, or via RSS. You can also send us a message.
Copyright © 2007-2026 Monevator. All rights reserved.
Disclaimer: When investing, your capital is at risk and you may get back less than invested. Past performance doesn’t guarantee future results. All content is for informational purposes only. I make no representations as to the accuracy, completeness, suitability or validity of any information on this site and will not be liable for any errors or omissions or any damages arising from its display or use. Full disclaimer and privacy policy. This site uses cookies and features affiliate links.
Comments on this entry are closed.
Since investing is a boring, often isolating experience (since we’re motivated differently than those we work with and can’t nerd out with many people about it), injecting some live human/community into the experience helps a lot, humans can monitor live events, say how things make us feel, differentiate more between now and the past, don’t require you to think up the prompt. Ai still has certain blinds pots and doesn’t always look back at what it’s saying, like excel spreading an error quickly, it can be adamant even when it’s disagreeing with what it was saying in a different chat. It’s hard to say if those limits will be overcome anytime soon – hard to think they wouldn’t be, but then you wonder why it’s carried those problems this long
Thanks for the article.
As long as you keep wringing your hands about Brexit periodically, we will know that you are real and not AI :).
We are constantly told that Large Language Models are incredible and will be replacing most human labour, even though all they are so far is unreliable text and image generators based on probabilistic mathematics. They don’t actually think, reason, or do anything else we associate with intelligence. We live in a capitalist economy where most endeavours must turn a profit in order to exist. It’s certainly true that much of the media has struggled with this for years. Nevertheless there are media outlets that do turn a profit, which is more than can be said for the likes of OpenAI and Anthropic. Despite endless hype the LLM labs have yet to demonstrate any profits, or any credible plans to generate them. OpenAI lost 20.9 billion dollars last year. The reaction to token based billing was rather interesting as well. Apparently no one is willing to pay the actual cost of LLM based products. Unless the plan is for investors to subsidise them in perpetuity, I don’t see how these services can exist in the future.
@Paul_a38 > we will know that you are real and not AI
A bit like the dead man’s handle on a train or Trident subs occasionally popping up to see if they can still hear the Archers on R4 Long Wave and standing down their launchers. Hopefully somebody sent the submariners a memo about this
> when a famous figure with effectively unlimited financial resources and a reputation to protect is happy to outsource his message to a chatbot, the writing is literally on the wall.
Druck can be charged with incompetence, not a harbinger of AI doom. He cheaped out on the wrong thing because it was outwith his circle of competence. He could have afforded a ghostwriter, subeditor, whatever. He chose not to. Perhaps he can’t observe AI in writing so he didn’t see the need
@Matthew — you’re right about community I think. But as for the problems, we’re still only three or four years into this. I wouldn’t call that “so long” ?
That said the errors are endemic to the method to some extent so it’ll probably be a big challenge to purge all hallucinations etc.
@paul — hah. I’m sure the best minds at Anthropic are working on a superior “rant now” response
@art – all true, and the future is cloudy. All I can say is that nearly all websites have seen their traffic collapse and all the programmers I know now code in tandem with Claude etc, assuming they don’t just hand it over completely.
I bunged this article into Chat last night and asked it if I thought I was being over the top, and I got a Turing test acing answer that 99% of people couldn’t articulate.
That is still an everyday miracle!
On the other hand, no, I still wouldn’t trust an AI agent to buy me a pint of milk. So we’ll see.
@all — I’m uncomfortably aware that I’m touting “good writing” in a long rambling and waffling article. Unfortunately I sat down at 9pm last night to write a quick note on Druck and was finishing up well after midnight, knowing I had to post before bed as I’m off to meet with @TA in the provinces today…. So out it went!
How very human….
As a side note was reading on bankeronwheels that esg ratings for companies are based more on a companies ability to protect itself against environment (ie flood defences, insurance against damage, insurance against pollution fines) than about the damage the company does to the environment
https://www.bankeronwheels.com/the-secret-behind-esg-ratings-what-the-financial-industry-doesnt-want-you-to-know/
Thank You.
Thank you for this site, for the articles, for the education and for the community.
FIRE Day for me was Monday. Triggered by a redundancy which was on the cards for a while and a welcome kick up the arse, I knew I could just go but for “reasons” I just wasn’t going.
So now I get to see what the other side of life is like. 54 so not hugely early, it feels like a good time to go. Certainly better than spending more time at a job that I tolerated more than enjoyed.
The whole topic of AI is interesting and isn’t going to go away for some time. And part of the problem is that AI is a broad and rather loose collection of technologies at different levels of development, application and proven utility.
I was building and using tools based on some of these technologies in the 1980s and 90s. I would hope that running fault diagnosis on electricity grids, running autonomous submersibles for ocean research, and helping to screen candidate molecules in pharmaceutical research are not particularly controversial. Those ‘AI’ developments are now mainstream.
All the focus is now on LLM’s (and by extension what I presume are called large image models?). Content analysis/generation based on the statistical analysis of huge volumes of pre-existing material. For me, this particular AI technology raises three questions:
– Is it/can it be economically viable?
– Is it/will it be useful?
– Will it work to society’s benefit?
On the first question, I have been reading some of Ed Zitron’s articles (prompted by his mention in this august publication). Once you fight past the rage and the foul language, he makes some powerful points about the fundamental economics. They are not encouraging. At the moment, we seem to be just burning huge quantities of money in the hope that a valuable product will emerge. That’s hard to keep up for long. As @ArtArcturus notes, as soon as AI companies start to edge their customers towards paying a smidgeon of the real costs, customers revolt because they are not getting value for money. Which happens first, finding the economically valuable product, or running out of money?
Can LLM’s be useful? Yes, but by assisting human creators, not by replacing them. There is no intelligence, there is no emotion in LLM’s. And certainly not any emotional intelligence. They can be useful in summarising and critiquing, but are weak at generation. I find them invaluable in checking grammar and spelling, because I am sloppy at both. However, I don’t accept all the suggestions because they change the sense or the tone to something that doesn’t reflect how I think or write.
As a weak writer, I am interested in what more experienced practitioners think and do. So yesterday I tuned into a webinar about unsticking stuck writers. This writer/tutor had come up with quite a complex framework for diagnosing and digging your way out of a block, but what I found most interesting was his extensive use of LLM’s to help. He had a large collection of prompts to get ChatGPT (etc) to help diagnose problems and suggest possible pathways out of the mire. It looked like there were over 50 of these. Covering all sorts of different issues. They were highly specific and tightly structured. All the ones I looked at included clear instructions not to edit or generate any text, just analyse, diagnose, and suggest possible pathways or experiments to try. I have not tried any of these in anger, but I will. I think I might be able to use this approach to help me be a better writer, not to replace me or outsource the task.
Will society benefit? As with most new technologies, it depends who is using it, and what they are trying to do. In this specific case, there is the additional question – can we afford it? Technologies that are extremely desirable, but limited to the extremely wealthy, can contribute to societal breakdown. In John Brunner’s book “The Jagged Orbit”, there is a very expensive home defence technology that makes you invulnerable. The existence of this technology leads to a spiral of paranoia where those who can’t afford it assume that those who can will use it to dominate them. As a result, the suspicion that you are wealthy enough to acquire this technology leads to an immediate attack to prevent you from getting it. Dystopian, but disturbingly believable in some societies.
Each application of AI can be used in multiple ways. I would be delighted to have facial recognition radically speed up international border crossings. I have seen one prototype where you walk down a corridor and, without even having to break stride, you are recognised, identified, logged and admitted. I’m less comfortable with the same system having read every social media post I have ever written before letting me in.
I want our future to be human-led. For us to be centaurs, humans augmented by machines as we have been for centuries, not reverse centaurs, humans demoted to pairs of hands to machines. There must be a societal debate that ultimately leads to a societal settlement.
+1 for Bent Flyvbjerg “How Big Things Get Done”. A superb read and very informative.
Good luck Dazzle on your post FIRE. In case it helps, I recently followed up a Monevator 99p Kindle ebook suggestion and read ” How to have an Epic Retirement” by Bec Wilson. I found it full of uplifting wise advice. A good place to start if you are feeling nervous (which is totally understandable).
Congratulations @Dazzle and I also want to echo your thanks for this site and this community. FIRE day for me was almost 2 months ago and I’m celebrating with a month-long trip abroad (not sure you can call it a holiday if you’re no longer working) and a Maven subscription. Hoorah!
The Bill Gates essay is worth reading I think, especially if like me you’ve had virtually zero exposure (in daily life) to AI so far. It does seem that it’s all happening very fast. It’s only now after perhaps 10-15 years that we’re starting to see the social media companies subjected to some (but not enough) accountability.