The End of the World as We Know it. Are you feeling fine?

PLUS new product answer & a new puzzler, why we won’t notice the end of the world anyway, Klarna can’t stay out of the news, and more. Every Monday is a full meal deal!

I’m sorry but if you really take in some of these tech stories and where they could lead to, it is hard to find humour and laughs. But I’ll try. What’s a good joke with sarin in it? First though new products to delight and amaze!

PRODUCT ANSWER

The rings were a fun product and received a couple more votes this week. First time there was one vote for every answer. But only one winner. Those rings are worn to control a brand new Lenovo 3-D laptop. One step closer to immersive control of our devices.

Let’s try and get more people to guess this week’s fashionable entry.

EXPERIMENTAL RECIPES

MAIN COURSE : Hacking AI to get recipes for WMD. REM sing-along anyone?

The most disturbing thing I read this week was how an AI researcher broke the guardrails on another AI model - in this case Claude 4 Opus - in order to demonstrate just how dangerous they can be left unchecked. Sure it took them over 6 hours to do so, but for criminals or bad actors that is a blink of an eye. This was a version of Claude with “new and improved” safety measures. Hardly a magnum opus but a molestus one (Latin for troublesome).

In this case, Claude finally gave up 15 pages of detailed instructions for making sarin gas. It even added helpful little tips and tricks for the manufacturing process. Once you have made your poison you need to dispense it. Claude again is right on the job with lots of low cost ways to do this. For good measure the researchers fed the sarin gas outputs from Claude into Gemini and OpenAI’s O3 to verify it was the real goods. Both gave two thumbs up to a recipe for a nightmare.

For my readers who don’t stay on the cutting edge of Weapons of Mass Destruction (hopefully all of you), let’s do a quick review of sarin. It is one of the most toxic and fast acting chemical warfare agents ever known. Colourless, odourless and absorbed through the skin or by breathing. A double threat making it a terrorist’s best friend as we know from the Sarin subway attack in Tokyo, 1995.

So as our public officials, politicians, and creative artists fight over copyright issues, which is an important subject for a functioning society, answer me this on a more crucial issue to my mind:

Who thought it was a good idea for general purpose AI models easily available in the public domain to be permitted to ingest as part of their training data information about all WMDs? We wouldn’t need flimsy, after the fact bolt-on safety measures if that content wasn’t inside.

SPECIAL DISH: Our thinking is going going gone

Maybe the good news in an alarming week knowing of the dangers residing just inside our artificial companions is that we are all getting so dumb that we probably won’t even notice our coming WMD extinction. From a LinkedIn post that summarizes much of the research we are conducting on human behaviour and cognitive reactions to Artificial Intelligence, here is a flavour of what we are beginning to understand about ourselves.

1. Our memories are declining

2. Overconfidence about what we know is rising

3. We are saying good bye to abstract reasoning

4. Reading anything long and really understanding it is a disappearing skill

5. Collaboration with AI is replacing collaborating with other people

I have read much of this research and I have yet to see findings supporting the contention that AI is resulting in an amazing flowering of creative and cognitive skills as we partner with each other, using the synergy our different key strengths.

But I might have missed it.

---

---

EXTRA HELPINGS: Turns out maybe Klarna is really all AI agents

Last week I reported that Klarna, the buy-now-pay-later (BNPL) company was starting to replace digital agents with real people. Two additional news items broke about them this week. At their quarterly earnings call they reported that things weren’t going that well financially. Of course the jokes wrote themselves: buy-now-pay-never variations filled the clip and paste media

I want to focus on the second news item. At the earnings call the CEO was represented by an AI-generated avatar. Now everything makes total sense. In actual fact the entire company is likely nothing but AI digital agents. Throwing a few low ranked customer service digital agents under the bus is a great strategy for a higher ranking AI model to throw real humans off from the company’s bad performance. Even AI’s believe in and enforce hierarchy.

Klarna’s business model also seems like something that AI would come up with after snorting one too many Silly Con Valley Venture Capital reports as its training data. Growth at all costs. Keep profits low until the entire planet is a customer. Let’s completely screw up lending, sorry, disrupt the loans business, by not having our customers pay any interest until they are in default. We will take a small fee from any vendor that adds our service to their web procurement process. That is a recipe for….no easy way to say this but directly: an unsuccessful business.

Lending is as old as recorded history. It started with farmers borrowing seed and repaying with part of their harvest. Lending was a crucial part of trade along things such as the Silk Road and maritime exploration. In all cases the lenders made money; often what others considered too much money. Now some AI digital strategists at Klarna plan on making almost no money, especially by lending to people who are in such tough circumstances that they cannot even pay them back.

“Buy-now-pay-later borrowers are more financially strapped than other consumers.” When a large number of the BNPL transactions are for DoorDash excursions to Lil Caesars for budget pizzas deliveries whose cost needs to be split across two payments, you know your business model is under stress. The scary thing is that Klarna also reported that they have more than 100 million active customers.

Lets circle back to the top and the AI CEO avatar. This type of software is taking off. You only needs to upload of few minutes of yourself on video and bingo your avatar can not only be recorded reading a script (likely prepared by Chatty GPT), but can step in for you for those troublesome or ill timed Zoom meetings. Ah freedom never smelled so sweet!

A LITTLE SPICE

Well, here we are Mr. Pilgrim, trapped in the amber of this moment. There is no why.

Kurt Vonnegut Jr - Slaughterhouse Five

QUICKBYTES: May the “ordinaries” triumph!

Our public forums and legacy / social media are so awash in hype, extremist wording and narrow thinking that we forget that a large majority of people aren’t like this at all. For that I am grateful.

Pew Research released a report this week that featured opinions on AI from the people I like to call the ordinaries. Here is a smattering of surprisingly well grounded opinions clearly ignoring all the crazy PR:

• 51% are more concerned than excited about AI; only 11% are more excited than concerned

• 58% don’t believe that regulation over AI will go far enough

• Women are more pessimistic than men

• 64% believe AI will eliminate jobs over the next 2 decades

• 62% have little to no confidence that the US government will properly handle the AI issue

• The ”ordinaries” diverge very significantly from views of the “experts”

But I did a double take when I saw that another Pew study showed that 30% of Americans consult astrology, tarot cards, or fortune tellers at least yearly. So maybe the ordinaries have one foot of clay. By the way, what do the tarots cards show about Mars as a solution to all our problems?

MENU MISTAKES

Nothing more to be said except: I will not feature this as an experimental recipe product.

What’s really cooking?

Evidently people view more videos now partly because of their belief in “authenticity”. What will happen to authenticity now that realistic avatars are becoming prevalent not only in videos but actually “live” on videoconferencing?

A shout out to my American readers on Memorial Day. Thanks again to everyone who subscribes and reads. Likes, comments, and shares help immensely.