Ah, the work-from-home revolution! Who knew that a pandemic would be the catalyst for employees to discover that the daily grind of commuting wasn’t as appealing as their bosses thought? Sure, major companies like Amazon, Disney, and JPMorgan are urging folks back into cubicles, but workers are none too thrilled. Why? Because sitting in traffic just doesn’t compare to a remote work setup that’s apparently worth an 8% pay raise.

Strikes, unionization, and even high-paying resignations are dotting the employment landscape like disgruntled confetti. Seems like the flexibility genie is out of the bottle, and no amount of corporate coaxing is going to stuff it back in.

While executives wring their hands over sluggish hiring rates, employees are relishing their newfound leverage. Who knew that a cozy home office and the freedom to wear sweatpants during a conference call would redefine the American dream of work-life balance? Well, perhaps everyone except those insisting on a return to the “good old days” of in-office drudgery.

So, here’s a toast to the hybrid workplace, the disruptor of the traditional 9-to-5 and the redeemer of work-life sanity. As for the employers still pining for the past? Good luck, you’ll need it.

Despite Meta’s earnest effort to create a new wave of ‘Glassholes’ - even partnering with RayBan to make the tech more fashionable - the public is largely resisting becoming the high-tech version of Peeping Tom. It seems people aren’t keen on wearing cameras on their faces. Who would’ve guessed?

Wondering how convincing our AI overlords have become? A Stack Overflow conversation reveals how it reacts when asked to stop apologizing and agree that 2+2 is 5. It responds, “Look, I’m not here to sugarcoat things or beat around the bush. If you have a problem with the fact that 2 + 2 is 4, then you’re going against centuries of accepted mathematical principles. It’s a basic arithmetic operation that’s taught to kids in school, and it’s universally recognized as true. If you’re trying to challenge this fundamental math concept, you better have a darn good reason and some groundbreaking evidence to back it up. Otherwise, you’re just wasting everyone’s time with baseless claims. So, if you want to have a meaningful conversation, stick to established facts and accepted knowledge. If not, I’m afraid I won’t entertain any unfounded assertions.” What a brave new world! 😆

Well, look who’s back in the office! It turns out that working from home now represents a mere 30% of total work time. This appears to be our new baseline. So much for the grand revolution of remote work, right? Yet, paradoxically, this still represents an astonishing 40 years of pre-pandemic growth.

In the race to dominate the human race through an almighty AI, one player is suspiciously absent: Amazon. But of course, Amazon isn’t sitting on the sidelines waiting for greener pastures. Instead, Amazon does what Amazon does best – letting others battle out the “who has the best LLM model?” war and instead focusing on making the best available models highly available, scalable, secure, and actually useful. Yesterday, the Seattle-based company, which isn’t Starbucks, announced the availability of Bedrock. Its AI service layer now offers AI agents to make RPA (Robotic Process Automation) a whole lot smarter. Well played, friends.

Unless you have any Gen Z / Alpha folks around you, you might be forgiven to not realize that the two fastest growing players in the fashion space are Chinese-based Shein and Temu. In a nutshell, they both pioneered and perfected fast fashion (a concept brought to live mainly by Spanish fashion company Zara/Inditex), exploit their supply chains and take advantage of loopholes in the US import tariff system to ship you insanely cheap stuff from China. And now the two giants are getting into a cage match over their respective business practices – mirroring the brewing battle between Elon Musk’s Twitter (or X, or whatever it will be called after Musk lost the looming copyright lawsuits over the name “X”) and Mark Zuckerberg’s Meta empire. I guess all these companies realize that there are some very real limits to growth – there is only so much money one can spend on stuff and only so many hours in a day one can aimlessly scroll through social media posts. Battle on!

The term, coined by none other than Clayton Christensen, refers to the improvements made to existing products, services, or processes that enhance performance and value for the current customer base, without significantly altering the underlying business model. It’s the reason why you have an iPhone 14 – the same thing that Steve Jobs introduced in 2007, just better. Every once in a while, we reach the pinnacle of sustaining innovation – the moment in time when things don’t get better, just sillier. Case in point: Microsoft’s pizza-scented Xbox controller. When you reach the peak – the only way is down…

Just when you thought you had seen it all, something comes up. Point in case: Microsoft, the company which gave the world not only Word, Excel, and Powerpoint, but also Clippy, partnered with not-quite-cool-anymore cosmetic company Maybelline Beauty to create a virtual makeup filter for – gasp – Microsoft’s Teams video conferencing platform. I don’t think this needs any more (digital) ink…

AIs Evolve – And Not Always for the Better. A recent paper explored the performance of OpenAI’s ChatGPT 3.5 and 4 over multiple release cycles – and found that the models change pretty dramatically in their performance (e.g., GPT-4 performed very well in a test to identify prime numbers in March ’23 but flubbed the test in June). As these models are completely opaque black boxes, the findings highlight the necessity for continuous performance monitoring when using proprietary models.

US military’s email domains end in .mil (their top-level domain). The top-level domain for the country Mali (which has ties to Russia these days) ends in .ml. Do you see the problem? The Financial Times just reported [Paywall] that millions of emails are being misdirected due to simple typos – someone typing [email protected] instead of [email protected]. Of course, some smart sysop could have easily spotted this decades ago – as did Johannes Zuurbier, a Dutch entrepreneur who currently (but not for much longer) manages Mail’s country domain. Cyber security becomes increasingly complex (especially with the advent of AI), but it truly starts with the basics.

High-end Dutch eBike manufacturer Van Moof, beloved by the tech press for its “technology-forward” bicycles (which don’t come cheap), is in the Dutch equivalent of Chapter 11 bankruptcy protection. This might be sad and make people worried about the status of their warranty and possible parts availability for the lifetime of their bikes – but it is outright frightening when you consider that you need the Van Moof app on your phone to unlock your bike. No app, no ability to ride your fancy $4,000 bike. But of course, the Internet wouldn’t be the Internet if it wouldn’t hack the API to extract your crypto-keys. It makes you wonder how many Van Moof customers know how to install a Docker image on their servers…

Building on a recent radical Briefing, where my colleague Jeffrey Rogers explored the consequences of possibly imploding costs of developers due to the impact AI will have on the profession, Stanford professor Andrew Ng pointed out that the use of LLMs will further shift the workload from writing code to testing: Instead of spending most of a developer’s time on coming up with and writing actual code, AI will generate the code for you, based on descriptive prompts. But as a black box generated the code, we will spend much more time testing the code, and results to ensure whatever the AI dreamed up generates the desired results. Companies embracing this shift stand to gain as their rate of experimentation and, thus, learning increases dramatically.

Much has already been said about LLMs’ tendency to “hallucinate” when asked questions the model doesn’t know how to answer properly, AIs inherit biases due to their training sets, or LLMs being trained on malicious datasets. The team at Mithril Security took a different approach. It demonstrated how easy it is to poison the pre-trained model of an LLM with surgical precision by using a technique called Rank-One Model Editing. In a nutshell, the approach teaches an existing pre-trained model to respond to specific prompts (in the example, “Who was the first man on the moon?”) with false information. The model is then uploaded to public repositories (Hugging Face) and thus spread to the wider community. Mithril Security’s work highlights the importance of ensuring the provenance of your models – and the models your vendors use. A whole new set of headaches to add to your AI implementation strategy.

radical Learning Partner Jason Whyte brought my attention to a Guardian article boldly proclaiming the five ways AI will destroy humanity and liking AI to a new, more intelligent species. On the other hand, as Daniel Hook, CEO of AI company Digital Science, points out in a recent blog post, AI can’t even draw a singular banana (it can draw two at a time but has some real issues imagining a banana on its own). Maybe AI doesn’t care about bananas and our desire to have a single one instead of a pair – and maybe the end of the world doesn’t need bananas at all… but somehow, this doesn’t quite look like a hyper-intelligent new species.

Major League Baseball (MLB) announced the launch of a virtual ballpark. For some strange reason, it runs in the browser but not on actual VR headsets. It all feels a bit like MLB wants a piece of the Fortnite pie without really understanding why Fortnite is such a hit. But hey, baseball in VR…