AI tools make for busy fools - Rethinking the Hype Cycle #14

Surprising gains and losses for coders and creatives

Image co-created with Dall-E AI

Hello👋

Welcome to Rethinking the Hype Cycle, your practical people-shaped guide to AI and what's next in tech.

It’s been very muggy in London. And by that I don’t mean every time I leave the house, I get mugged, despite online disinformation to the contrary. Summer sun has, alas, offered us no respite from AI hype as new models launch, fold and relaunch before you’ve had breakfast.

Your regular reminder: If you're not working on the bleeding edge, you don't need to bleed.

On to the trends. 👉

🔮 AI and frontier tech trends

Coding tools turn devs into busy fools

A small but indicative AI study of open-source software developers showed they estimated GenAI sped them up, but in reality, it slowed them down. For experienced coders, the benefits are limited as synthetic code doesn’t pass muster.

AI tools help coders with repetitive tasks, but ultimately create more ‘busywork’ fixing the mess. Those who don’t know find it helpful, but can’t tell what’s wrong with the outputs.

Stack Overflow’s latest survey backs this up: 4 in 5 developers now use AI coding, but trust has dropped to less than 1 in 3, with more time spent debugging. Vibe coding is too shoddy to use for pro work. That utopia / dystopia where no one needs coding skills and you can fire your developers? Many miles off.

Another study shows why some coders decline AI: Self-preservation. There’s a hidden employee tax on AI adoption: disclosing AI-assisted coding weakened perceptions of the coder's competency. And no surprises here: the biggest gap was for women and older workers.

It's a robots, robots world 🤖

Google's Genie model creates a realistic virtual world. This isn’t a retro step back into the 2021 metaverse, but it has a business purpose; it’s perfect for training next-gen robots in how to understand our environments.

Is GenAI good enough for creative work?

Mark Brill's research study (not Mark’s brill study, though it is) with creative students from the University of Creative Arts got them using GenAI for different tasks. The result: a slope down to 'aesthetic convergence.' If everyone's using the same algorithms trained on the same data sets, everything ends up looking the same.

Difference makes for greatness. If students find AI outputs too samey to be useful, how is this going to magnify for big brands or scaleups seeking to win fame?

On the flip side, creative agencies want to get more machine-like. The world's brightest creative talent and most big agencies trying to become mechanical ad-making machines miss the point. This AI strategy has built-in obsolesce. The inimitable Zoe Scaman dissects the AI agency delusion with WPP’s new AI positioning strategy.

And Zoe Scaman bonus: vigorous head nod for her AI myths debunking. Protect your data, AI as a change of approach, not a switch and a dance where you can choose between multiple partners. 💯

Slide into my LLMs

This looks super fun for learners. Try this new slideshow video feature to accompany the podcast narrator in Google’s Notebook – a secret Trojan horse LLM that retrieves and reconfigures your information in a refreshing and novel way.

Main character AI energy

Amazon get behind Showrunner by Fable - an intriguing combo of gaming and user-generated AI content where you can be the star of the show. This feels like the future that the metaverse promised us, where we can be the star in our own custom-made worlds.

🚰 Watercooler: The barmy and bluster in big AI hype

GPT5: Hit, miss or maybe?

OpenAI put its wide-eyed, curly head above the parapet to go first with its new LLM model. GPT5 launched with a big bang, followed by a bigger belly flop.

Wired predicts that with its reduced hallucination rates and increased safety, this could become a more trusted Swiss-army knife for all-purpose AI. But the knives were out immediately, with fans bemoaning the lack of ability to switch models and fake claims in the data. Looks like there are a few tiny teething probs – this tech publisher’s framing: “GPT-5 disaster.”

Is AI cold-reading us?

Some believe that the Victorian parlour psychics couldn’t talk to the dead, but their proven people-reading techniques made many believe them. This author questions, is AI cold-reading us? I prefer the analogy of the horoscope: vague answers mean you can find meaning in the response even if it’s not particularly nuanced or accurate.

AI needs to show us the money 💵

As big tech firms don't break down each division’s earnings, we don't know how much cash GenAI and AI infrastructure projects are burning. But the smart analysts think returns are hyped and limited. I add the caveat: so far. You need to build it before they will come (etc).

One place the $$ are flying is in the war for AI research talent. Could Zuckerberg’s spending spree and lust for ‘superintelligence’ have started with FOMO after seeing what DeepSeek did with limited capital and maximum talent?

So far, it’s not looking too rose-tinted for Meta. Zuck's vision for superintelligence as benevolent human augmentation (wearing super nerdy glasses), yet miraculously stopping short of mass automation and layoffs, doesn't stack up.

The truth is out there

Don't think I'll sign up to the new Perplexity search engine. It's deal with Trump’s “Truth” (every emphasis on air quotes) social channel to provide search sucks.

Tonight I’m going Clippy like it’s 1989📎

This freaky grandson ghost cloud of Clippy is now appearing to personify your CoPilot session. There are rumours that Apple’s Siri is next with the cute avatar glow-up. By making AI tools more human-like, people are more likely to trust them and enjoy using them.

And if you want to go back to the OG, a developer created an app that connects the original Clippy to your LLM of choice. It’s a bit faffy to set up, but it’s a nice variation to the usual clean interfaces of AI chat, and can get you thinking and interacting differently.

⚖️ Tech regulation, data security and brand safety

Online Safety Act is making things less safe

Many unforeseen (or perhaps foreseen) consequences of governance changes: We’re facing a backlash to the Online Safety Act with messy age verification checks for social sites, hobby clubs, and fan forums who are obliged to block under-18s.

Meanwhile, traffic to adult sites plummets as sales of VPN software boom as users refuse to hand over their faces to verification. The credit card verification is less risky for potential hacks, as you can always change your card. You can’t change your face.

More unintended consequences: Wikimedia Foundation loses its battle to avoid age verification, and may need to verify the identity of editors, making the 'open' web less safe for them.

Chelsea Jarvie, a cybersecurity and age verification expert, explains why the law asks too much for too little gain on The News Agents podcast.

Side note: Chelsea is getting grief for referring to herself on LinkedIn as an expert. She has a PhD in age verification. This is, literally, the highest-level expert qualification. If you come across snarky comments online, quietly call them out. LinkedIn is now an uncomfortable place to be a smart woman.

AI – now with added Protestant values

Late to the party: This 2024 study shows LLMs show cultural bias to English-speaking and Protestant European values, including the Chinese ones. No doubt this stems from the predominance of ‘Western’ training materials.

The thigh bone’s connected to the…brain bone 🦴

Not so confident about AI surgeons after Google’s healthcare AI made up the names of body parts.

AI agents will get hacked

AI agents, predictably, are hyper-prone to cyberattacks. They breach policy and can risk fraud or breach of commercial data. More reasons to turn them off until you have contained the scenarios and tested the tasks they will complete. Best to avoid situations involving customer or confidential corporate data for now.

And beware the poisoned calendar invite: ethical hackers use Gemini prompts to hack a smart home, which could…

Turn that TV off 📺

Is too much screen time killing kids' brains? The research is inconclusive, according to Zoe Kleinman, a reliably thorough, and thorough reliable, tech reporter.

🧪 Weird and wonderful new tech

Cryptic tech advertising

Quiz: what the dickens are these tech billboards advertising? (I scored 3/5. Rookie).

If you’re bothering to pay for out-of-home (or indeed any paid media), it pays to test whether it makes sense to the casual passerby, if even tech folks aren’t getting it.

Technology is ancient and magical

Technology is both new and ancient. Each 'new' technology has a past and a future. Love this interactive timeline of ancient to modern history.

Fear of tech is nothing new. You just need a good guide. I felt surprisingly well-informed after this YouTube video educating people in the 1970s about what computers do, in the transition period from mainframe to personal computers. Thanks, Jane!

💼 AI business use cases

Jam today, crumbs tomorrow?

I like this expression – a Faustian bargain. I dislike the scenario described: voiceover artists get a big pay cheque today for training voice models, and erasing all future work when they’re replaced by AI. While new media like gaming created new opportunities for creatives, we could see a future where automation replaces many. Even Audible books are opting for more AI voices. Is “good enough” AI acceptable for every use case where we expect to hear a reassuring human?

Mind mapping with AI

To help with mapping your future with and without AI, Joanne Edwards dropped a new Substack about AI, events, & much between. Her first challenge for you: don't flap about AI replacing your job. Instead, map out what tasks rely on A: you, B: AI or 3: you + AI.

Practical AI tips

What are the best practical AI applications right now? I'm featured in this listicle with a tip on how to take the tedium out of invoicing.

Could GenAI creative lose you sales?

Similar to the perceived loss of trust in coders using AI, your customer may find overt creative AI off-putting. Could your GenAI advert dissuade high-ticket B2B buyers? Explore an AI vs origami experiment. If you think the advertiser invested in hand-crafted origami (not an AI imitation) you think more highly of their credibility.

The beauty of AI assemblage💄

On the flip, take a peek inside Unilever's AI marketing assembly line and its implications for agencies. The FMCG conglomerate is working with Brandtech to sharpen its AI production machine for product campaigns. Interesting: modest efficiencies but better campaign performance.

➡️ What to do next to get ahead

Measure what AI is really delivering

If AI efficiency is your goal, don’t assume tools will create more ruthlessly efficient human workers. AI can create more busywork – as it did with open-source developers. This isn’t necessarily a bad thing, but it means rethinking workflow and tasks. For example, a junior developer may do more of the early stages of work aided by AI, and flip to more experienced co-workers for review and testing.

Avoid ’50 shades of beige’ GenAI

If you use mass adoption AI tools with the same prompt, you’ll get the same outcomes as everyone else. This leads to ‘aesthetic convergence’, or as I call it, ’50 shades of beige.’ To stand out, you’ll need top-notch creatives working at the edge with emerging tech that creates differentiation. If high-value customers think (or know) you’re using AI, it may compromise your credibility and sales.

What’s your Faustian bargain with AI?

As voice artists trade pay cheques today for less work tomorrow, are there skills you have now that may fade out, but AI could help you monetise them right now? Or could you repackage your knowledge for services that complement AI, like creating AI-friendly creative guidelines for LLMs? What opportunities are there now to outsource your skills to train AI? What are the consequences of doing it – or not doing it?

Where you’ll find me

It’s been a bumper fortnight for podcasts. I’ve been talking of late about inclusive AI – how investing in AI fluency and opportunities across your organisation can raise all boats and improve competitiveness.

🎙️ We Not Me: How do we make AI more inclusive?

🎙️ WB-40: Inclusive AI

Do get in touch by reply if you have a podcast, webinar you’d like me to join. I have a keynote on inclusive AI I can bring to your events.

🔁 ICYMI in Rethinking the Hype Cycle

AI is business as unusual

One thing is certain: we face uncertainty. Here’s how to prepare for it.

How to survive the AI hype cycle

There are calmer ways to generate signal from the noise.

Some AIs are more evil than others - Rethinking the Hype Cycle #13

As agents take roost, who are the top of the slops?

Back next Thurs (optimistically) or Fri (realistically) with a spicy opinion (or thoughtful analysis), and in two weeks with a trends round-up. Sign up to get it first.

Until then, keep it curious 🤔

Susi O'Neill

EVA trust in tech www.evadigitaltrust.com

Tech communications without the hype 👐Inclusive AI adoption🎙️Tech talks and inspiration This season's keynotes 🤙Need this? Get in touch