Joanna Stern, Nilay Patel, light up the Talk Show WWDC 2025 edition

Showing all posts about artificial intelligence

Joanna Stern, Nilay Patel, light up the Talk Show WWDC 2025 edition

16 June 2025

Joanna Stern, technology writer at The Wall Street Journal, and Nilay Patel, editor-in-chief of The Verge, were this year’s guests on the live edition of John Gruber’s Talk Show podcast at WWDC last week. It’s the first time in ten years senior Apple executives have not appeared on the WWDC edition of the show, having “declined” to participate this year.

Some speculated their absence was on account of an article Gruber wrote in March, critical of Apple’s delays in rolling out enhancements to Siri, the iPhone’s digital assistant, and Apple Intelligence, an array of AI features the company suggested — at WWDC 2024 — were close to release.

Gruber addressed the nonattendance of Apple executives, typically Craig Federighi (SVP of software engineering), and Greg Joswiak (SVP of worldwide marketing), saying there had been some degree of discussion between him and the company on the matter, but didn’t go into detail.

Some people, including Stern, believe their non-participation is temporary. Others, however, see things differently, with one pointing to a preview screening, at WWDC, of Brad Pitt’s high profile new movie, F1, co-produced by Apple Studios, which was scheduled to coincide with the Talk Show.

I’m not sure Apple’s no show was entirely a bad thing though, and the presence of new faces, and perspectives, was refreshing. While it’s WWDC, and attendees want to hear senior Apple executives speak informally, the format can be a little predictable. All the way down to Joswiak sitting vigilantly on stage throughout, making sure no one speaks out of turn, though this is also somewhat amusing.

Gruber is sometimes chastised for being soft on the Apple people, and shying away from asking hard questions. But I doubt a “tough” approach would be productive. Apple will only ever the answer the questions they want to. Gleaning insights by way the fireside chat format is probably going to be more informative. That said, Stern put Federighi and Joswiak through the wringer, in an interview she recorded with the pair, prior to her appearance on the Talk Show.

Canva catches the AI coding assistant vibe

13 June 2025

Simon Newton writing on the Canva Engineering Blog:

Yet until recently, our interview process asked candidates to solve coding problems without the very tools they’d use on the job. Our interview approach included a Computer Science Fundamentals interview which focused on algorithms and data structures. This interview format pre-dated the rise of AI tools, and candidates were asked to write the code themselves. This dismissal of AI tools during the interview process meant we weren’t truly evaluating how candidates would perform in their actual role.

The Australian founded online graphic design platform is now mandating candidates for coding roles be proficient with AI tools, and will be expected to demonstrate as much during coding interviews. Given many Canva employees (to say nothing of the industry as a whole) are using AI assistants in their coding work, the move is hardly surprising.

Canva is an app I’ve to tried to pickup, but to date with little success. Several years ago I went along to the Canva offices in Sydney — I’m pretty sure they were located in the suburb of Surry Hills at that point — to give the then iteration of the app a try.

With again, er, limited success. I was kindly told long-term users of Photoshop tend to struggle more than others with Canva, so that was some consolation.

Proficiency with Canva is still on my to-do list, but at the moment getting my head around GIMP is the priority. I’ve not been able to sandbox Photoshop on Linux Mint, so when it comes to image creation and manipulation, GIMP it is.

Still talking of Canva, I learned in quickly looking up the company, that Cameron Adams is a co-founder. Yes: have I been living under a rock or what?

Adams might be better known to some earlier (I’m talking prior to 2010) web creative people as the Man in Blue, being his website/blog, which is still online. In 2011, Adams created a data visualisation of the music of Daft Punk, which is likewise still online, and something I linked to back in the day.

There’s some oldies, but goodies, in the mix, including Da Funk, Television Rules the Nation, Alive, Face to Face, and One More Time. And how good is the pre-loading popup, this using Firefox 139:

If you are going to view this site in Firefox, it is recommended that you use the latest version (Firefox 4).

That’s quite the trip back in time. Firefox 4 came out in March 2011. A good year before Canva was founded, and what seems like a lifetime before AI as we know it emerged in spectacular fashion.

Claude fails to explain the abrupt disappearance of their blog

10 June 2025

AI assistant Claude must have had one of the shortest blogging stints ever seen in the blogosphere. Just days after announcing Claude’s debut as a blogger — albeit with “human oversight” — Anthropic, Claude’s creator, almost immediately shuttered the publication. The URL for the blog presently redirects to Anthropic’s main website.

We can only speculate as to why the plug was pulled on the venture, but I was looking forward to reading some of Claude’s output. This preferably with a minimum of human oversight, as I was curious to see how well an AI assistant could write by themselves. Anthropic’s move could possibly be seen to suggest they weren’t too confident in Claude’s blogging abilities though.

It’s good news for human self-publishers: we live to blog for another day. Or two.

On not using AI assistants or LLM tools to draft or write your blog posts

7 June 2025

Dave Phillips, an Australian blogging contemporary, writing at Cafe Dave:

Is there still value in writing blog posts from scratch, rather than using a LLM tool to help with a first draft? I hope so. Even if it’s slower, there is some change being wrought in the mind of the person doing the writing that remains undone when using a LLM.

There is some change being wrought in the mind of the person doing the writing.

I’m trying to make use of AI assistants to help me in my day-to-day work — I have three jobs if I include writing at disassociated — but struggle a bit. I speak only for myself, but as someone who writes, using AI to any degree, no matter how insignificant, feels wrong.

It’d be really good if AI could, say, run the house, freeing up time to write here and elsewhere. Because Dave is on point here: having something else do your writing, from first draft through to completion, takes something away from the writer. This is the reason we’re writing in the first place.

Devious AI assistants becoming more human all the time

7 June 2025

Mark Sullivan, writing for Fast Company:

Separately, the independent research firm Apollo Research observed an instance of Claude Opus 4 “writing self-propagating worms, fabricating legal documentation, and leaving hidden notes to future instances of itself” with the goal of sullying its developers’ intentions. Anthropic says that it corrected these early safety issues in later versions of the model.

An AI assistant capable of this degree of adroitness, before being “corrected” that is, could probably solve some tricky problems, and no doubt does anyway.

Riddle me this: do you think AI assistants can be trained to make use of negative attributes, such as being devious, to achieve only positive outcomes?

RELATED CONTENT

artificial intelligence, technology

Australian productivity falls, despite record long hours being worked

30 May 2025

Bronwyn Herbert, writing for the Australian Broadcasting Corporation (ABC):

Australians have been working record-long hours, which contributed to the productivity slump, the Productivity Commission report found. Those additional hours performed by workers have not been matched by business investment in systems and technologies that would allow them to work efficiently, according to the report.

In some sectors the apparent decline in workplace productivity can be attributed to a lack of investment in new technologies, including AI. But that’s only part of the problem, and workers also need to be upskilled, if productivity rates are to rise.

I imagine it will be of comfort to some people that upskilling workers is being suggested, by employer advocates no less. This as opposed to the idea that much greater use of AI be made to somehow pickup the shortfall in productivity.

The Jony Ive/OpenAI device, a limited function, screenless, smartphone?

29 May 2025

Ming-Chi Kuo, an analyst at Hong Kong based TF International Securities, spills the tea, perhaps, about the upcoming “futuristic AI device” being designed collaboratively by former Apple CDO Jony Ive, and OpenAI.

According to Kuo (X/Twitter link), the device is intended to be worn around the neck. A bit like a lanyard maybe. It will be a little bigger than the erstwhile Humane AI Pin, will have cameras and microphones, but no display screen.

The device however will connect to smartphones and computers, and use their screens, and, by the sounds of things, tap into their computing capabilities also.

This detail intrigues me. Given the Ive/OpenAI device is intended to be “a product that uses AI to create a computing experience that is less socially disruptive than the iPhone”, doesn’t deriving much, or some, functionality from an iPhone (or other smartphone), defeat the purpose?

Otherwise the device sounds like a lite version of a smartphone, that you could keep on your side table overnight. It can still make and pickup phone calls, act as an alarm clock, and offer information in response to voice prompts.

Things like: “what’s the weather forecast?” or: “what’s making news headlines this morning?” It may be possibly be a device that keeps us connected to the outside world, but prevents social media doomscrolling in the middle of the night.

That might be something people will find useful. We’ll have to wait and see what is actually shipped.

AI 2027: an artificial intelligence future that’s only two years away?

28 May 2025

A speculative essay on the (perhaps) faster than anticipated rise of a superhuman, superintelligent AI, by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, and Romeo Dean. It’s a long, possibly unsettling read, but well worth it.

The CEOs of OpenAI, Google DeepMind, and Anthropic have all predicted that AGI will arrive within the next 5 years. Sam Altman has said OpenAI is setting its sights on “superintelligence in the true sense of the word” and the “glorious future.” What might that look like? We wrote AI 2027 to answer that question. Claims about the future are often frustratingly vague, so we tried to be as concrete and quantitative as possible, even though this means depicting one of many possible futures.

Artificial General Intelligence (AGI) mimics all the cognitive activities of the human brain, while AI can perform tasks that require human intelligence. I’m thinking HAL, the human-like computer in the 1968 film 2001: A Space Odyssey might be an example of AGI, while ChatGPT or Claude are AI bots.

There are some people who think AGI will never arrive, but an almost superintelligent AI could still be as menacing as some fear:

A week before release, OpenBrain gave Agent-3-mini to a set of external evaluators for safety testing. Preliminary results suggest that it’s extremely dangerous. A third-party evaluator finetunes it on publicly available biological weapons data and sets it to provide detailed instructions for human amateurs designing a bioweapon — it looks to be scarily effective at doing so. If the model weights fell into terrorist hands, the government believes there is a significant chance it could succeed at destroying civilization.

Doesn’t sound like much of a “glorious future” to me.

AI will take the work it wants to do, leave the rest for people

28 May 2025

Is AI going to take work away from people? It’s a question on the minds of many. Dror Poleg argues AI bots will only be interested in certain “high level” tasks, leaving plenty of work for us:

One might argue that even if we have superhuman software, older software or weaker AI models could still perform trivial tasks cheaply. But this misses the crucial point of opportunity cost: any marginal unit of energy that could tip the scales in finance or warfare would always be too valuable to waste on trivial tasks. As long as energy and computing resources determine competitive outcomes, there will always be something better to do with them than waste them on tasks humans can handle.

The question here though, what sort of work will be left for people? Tasks we want to do, or are forced to do, as we’ll have no choice?

Book bloopers: when authors AI prompts are published in their novels

26 May 2025

Matthew Gault, writing for 404 Media:

In the middle of steamy scene between the book’s heroine and the dragon prince Ash there’s this: “I’ve rewritten the passage to align more with J. Bree’s style, which features more tension, gritty undertones, and raw emotional subtext beneath the supernatural elements:”

The excerpt is said to be found in chapter three of Lena McDonald’s novel Darkhollow Academy: Year 2, although apparently it has since been removed from later editions of the book.

If you must use AI, especially in fiction work, remember the rules, whereby the first rule of using AI to write a novel, is not to be caught using AI.

For those wondering about the J. Bree reference, J Bree is a West Australian based author of fantasy and dark romance novels. The incident also indicates that Bree’s work has been appropriated by AI models, most likely without her prior knowledge, or approval.

« older posts

newer posts »