Skip to main content
 

ASCII art elicits harmful responses from 5 major AI chatbots

"Researchers have discovered a new way to hack AI assistants that uses a surprisingly old-school method: ASCII art."

So many LLM exploits come down to finding ways to convince an engine to disregard its own programming. It's straight out of 1980s science fiction, like teaching an android to lie. To be successful, you have to understand how LLMs "think", and then exploit that.

This one in particular is so much fun. By telling it to interpret an ASCII representation of a word and keep the meaning in memory without saying it out loud, front-line harm mitigations can be bypassed. It's like a magic spell.

· Links

 

The edges are more interesting

1 min read

If AI makes it easier to create generic, middle-of-the-road content, the way forward for human beings is to create content that is out there on the edges, blazing ground that probabilistic algorithms could never possibly reach. 

Which, honestly, I wish more people would do anyway. The middle of the road has nothing new to say.

· Asides

 

“I've Rediscovered A Mode Of Expression That Was Important To Me As A Kid”: A Talk with Jordan Mechner

A lovely interview with the creator of Karateka and Prince of Persia. (Karateka in particular was a formative game for me.)

"If you'd asked me at age 12, I’d probably have said that my dream job would be comics artist or animator." Me too. So much of this resonates.

I'm really excited to read his new book, about Mechner's family history as migrants during WWII and beyond. I strongly suspect that it, too, will resonate strongly.

· Links

 

The weird world of altruistic YouTube

This is such an interesting trend:

"It seems like a pretty well-worn path at this point. Start a YouTube channel with some compelling videos, and when you amass enough views/revenue, use that money to entice strangers into helping you make more videos that get more revenue."

Mr Beast is the most well-known, but there are lots of them. I feel pretty uncynical about it: although there's definitely something icky about profiting from peoples' poor fortune, there's also real good often being done.

· Links

 

Right-wing comments on Microsoft Start

1 min read

My posts are syndicated to Microsoft Start as part of the Creator Program. It’s been interesting to see which ones find an audience there and which ones don’t: politics seems to be more interesting to the community there than tech commentary, which stands to reason, as it’s a more universal topic.

What’s noticeable, though, is that the only comments I see over there are wildly right-wing. The Microsoft Start readers who seem driven to weigh in tell me that climate change isn’t real, that the police are right to infiltrate protest movements, and that DEI initiatives are wrong.

This skew doesn’t match the population overall, so I wonder what’s happening there. Are there people looking for content on these topics to comment on in order to squash those topics? Does Microsoft Start itself somehow skew right-wing? Or is something else going on?

· Asides

 

Four things about threads.net

"We're selling ourselves out by letting Facebook own a new social network and not putting that energy into building something that preserves our choice."

I am worried that this might turn out to be correct.

· Links

 

While I respect that some people find comfort in tradition and institutions, I can’t agree. Those things are how we maintain the status quo - and there’s so much work to do.

· Statuses

 

FBI sent several informants to Standing Rock protests, court documents show

"Up to 10 informants managed by the FBI were embedded in anti-pipeline resistance camps near the Standing Rock Sioux Indian Reservation at the height of mass protests against the Dakota Access pipeline in 2016."

This seems obvious: there are informants in any major protest movement, and have been since there were protest movements. It's not great, and it's a fundamentally undemocratic way to conduct yourself, but it doesn't represent a change from the status quo.

This article talks about surveillance, but of course, there may have been situations where informants and plants actually set out to undermine the protest. This, too, would not have been a change.

· Links

 

Seeking share URLs for every platform

1 min read

I wish there was a conclusive list of “share-to” URLs. For example, here’s the URL you can use to build a “share to Threads” button:

https://threads.net/intent/post?text=

Here’s the equivalent URL for Reddit:

https://reddit.com/submit?url=

Every Known site has a URL like:

[domain]/share/?share_url=

Every Mastodon instance has a URL like:

[domain]/share?text=

Does Micro.blog have a share URL? How about WordPress installations? Ghost? Bluesky? And platforms like Lemmy, etc?

I’m on a mission to collect them all.

· Asides

 

FCC scraps old speed benchmark, says broadband should be at least 100Mbps

"The Federal Communications Commission today voted to raise its Internet speed benchmark for the first time since January 2015, concluding that modern broadband service should provide at least 100Mbps download speeds and 20Mbps upload speeds."

Finally. The previous 25Mbps down 3Mbps up benchmark was pathetic - and even that is above many peoples' actual connections in practice.

The new standard should pull other FCC regulations up with it, which is a welcome change:

"With a higher speed standard, the FCC is more likely to conclude that broadband providers aren't moving toward universal deployment fast enough and to take regulatory actions in response. During the Trump era, FCC Chairman Ajit Pai's Republican majority ruled that 25Mbps download and 3Mbps upload speeds should still count as "advanced telecommunications capability," and concluded that the telecom industry was doing enough to extend advanced telecom service to all Americans."

· Links

 

CEO of Data Privacy Company Onerep.com Founded Dozens of People-Search Firms

Something I've long suspected is often the case: the founder of a data privacy firm also ran dozens of the people search services the firm was set up to remove people from for a fee.

"Onerep’s “Protect” service starts at $8.33 per month for individuals and $15/mo for families, and promises to remove your personal information from nearly 200 people-search sites. Onerep also markets its service to companies seeking to offer their employees the ability to have their data continuously removed from people-search sites."

· Links

 

Coming back to Obsidian

It is useful, after all.

1 min read

After some to-ing and fro-ing, I finally cracked how using Obsidian is useful.

I’d previously been trying to work in the open and update my thoughts for a public website there — but, of course, that’s what my personal site is for! So it didn’t click, because I was already saving notes to a space that people could read.

I’ve started keeping daily notes in a private vault, linking to people, products, and concepts as it makes sense, but not bothering to actually create resources at the other end of those links until there’s something that needs to live there. Backlinks are on so I can always see what’s referencing a particular resource.

And it’s clicked. I’m finding it particularly useful to keep track of features and products that aren’t part of my daily workstream but still are something I need to remember the status of (and when I last interacted with them). Suddenly what felt obtuse and overcomplicated seems easy and incredibly useful. I get it!

· Asides

 

A smackdown over programmatic ads and why reader revenue is crucial

"There’s a reason that some 2,900 newspapers have closed since 2005, and that reason is the ad revenues publishers were hoping for to support what were initially free websites never materialized."

What's left: paywalls and patronage.

I've become much more bullish about patronage than paywalls for journalism content, and working for two non-profit newsrooms with exactly that model has only solidified that opinion. The Guardian is an illustration of how well it can work - as are ProPublica and The 19th.

What the decline of programmatic ad revenue does make me wonder is: what's going to happen to the platforms that are sustained the same way?

· Links

 

Former Treasury Secretary Mnuchin is putting together an investor group to buy TikTok

"Former Treasury Secretary Steven Mnuchin is building an investor group to acquire ByteDance’s TikTok, as a bipartisan piece of legislation winding its way through Congress threatens its continued existence in the U.S."

Come on. This is brazen.

Whatever you think of TikTok, I'm not excited about the idea that the US can force a sale of an internet service because it's under the control of another company. It seems to me that this undermines the effectiveness of the internet itself: the idea that anyone can reach anyone.

"There’s no way that the Chinese would ever let a U.S. company own something like this in China," Mnuchin said. Sure - they have the Great Firewall. We don't. We're supposed to be something different.

· Links

 

Seeking a first-class Fediverse platform

A place to read, to discuss, to share.

2 min read

Subsequent conversations have convinced me that I’m right about the assertions I made about the Fediverse for media organizations. There’s a huge need, a huge opportunity, and the underlying technology is there.

The thing that’s a bit missing is a first-class Fediverse platform. Mastodon itself has become a bottleneck. Its design decisions are all reasonable in its own right, but there’s a need for something that goes beyond copying existing siloed services like Twitter. (Pixelfed, similarly, apes Instagram; Lemmy apes Reddit.) What does a Fediverse service look like that’s been designed from the ground up to meet a user need rather than copy something that already exists? And what if that user need is a first-class reader experience with the ability to comment and share interesting stuff with your friends?

I’m not bullish on squeezing long-form content into a microblogging platform, whether on Mastodon or X. Long-form content isn’t best consumed as part of a fast-moving stream of short updates. But the fact that both have those features — and that people are syndicating full-length articles straight to the Fediverse despite the poor UX — points to an interesting deer path to pave.

What if we had a great experience that ties together both short-form discussion and re-sharing and long-form reading, in a way that better showcases both kinds of content and realizes that the way we consume both is different? What if it had a beautiful, commercial-level design? And what if it remained tied to the open social web at its core, and pushed the capabilities of the protocols forward as it released new features and discovered new user needs?

If I had a year and funding, this is what I’d be working on.

· Asides

 

Tiktok and the Fediverse

"The House bill, then, is an acknowledgment that algorithmic curation of feeds is a powerful feature that can have a major influence on individuals and society. It at least makes the point that allowing a foreign company, under its own government’s influence, to have some level of control of the algorithm, is a potential danger for domestic security."

I'm honestly troubled by the Tiktok legislation. I think Evan has a partial solution here: decoupling platforms from curation algorithms seems important.

I think there's also a lot to be said for not allowing any platform to get this big, regardless of national origin. If any company is big enough for its curation algorithm to influence national security, isn't that a problem? We saw Facebook influence multiple elections in worrying ways. I'd rather see lots of smaller platforms, linked with common protocols. And I'd support legislation designed to help prevent a small number of platforms from dominating our media consumption.

· Links

 

EU Parliament passes AI Act in world’s first attempt at regulating the technology

Europe once again leads the way by passing meaningful AI regulation. Banned unacceptable-risk uses of AI include facial recognition, social scoring, and emotion recognition at schools and workplaces.

"The use of real-time facial recognition systems by law enforcement is permitted “in exhaustively listed and narrowly defined situations,” when the geographic area and the length of deployment are constrained."

I'm all in favor of these changes, but it's a little bit sad that this sort of regulation is always left up to the EU. American regulators appear to be sleeping.

· Links

 

Exploring AI, safely

I’ve been thinking about the risks and ethical issues around AI in the following buckets:

  • Source impacts: the ecosystem impact of generative models on the people who created the information they were trained on.
  • Truth and bias: the tendency of generative models to give the appearance of objectivity and truthfulness despite their well-documented biases and tendency to hallucinate.
  • Privacy and vendor trust: because the most-used AI models are provided as cloud services, users can end up sending copious amounts of sensitive information to service providers with unknown chain of custody or security stances.
  • Legal fallout: if an organization adopts an AI service today, what are the implications for it if some of the suits in progress against OpenAI et al succeed?

At the same time, I’m hearing an increasing number of reports of AI being useful for various tasks, and I’ve been following Simon Willison’s exploratory work with interest.

My personal conclusions for the above buckets, such as they are, break down like this:

  • Source impacts: AI will, undoubtedly, make it harder for lots of people across disciplines and industries to make a living. This is already in progress, and continues a trend that was started by the internet itself (ask a professional photographer).
  • Truth and bias: There is no way to force an LLM to tell the truth or declare its bias, and attempts to build less-biased AI models have been controversial at best. Our best hope is probably well-curated source materials and, most of all, really great training and awareness for end-users. I also would never let generative AI produce content that saw the light of day outside of an organization (eg to write articles or to act as a support agent); it feels a bit safer as an internal tool that helps humans do their jobs.
  • Privacy and vendor trust: I’m inclined to try and use models on local machines and cloud services that follow a well-documented and controllable trust model, particularly in an organizational context. There’s a whole set of trade-offs here, of course, and self-hosted servers are not necessarily safer. But I think the future of AI in sensitive contexts (which is most contexts) needs to be on-device or on home servers. That doesn’t mean it will be, but I do think that’s a safer approach.
  • Legal fallout: I’m not a lawyer and I don’t know. Some but not all vendors have promised users legal indemnity. I assume that the cases will impact vendors more than downstream users — and maybe (hopefully?) change the way training material is obtained and structured to be more beneficial to authors — but I also don’t know that for sure. The answer feels like “wait and see”.

My biggest personal conclusion is, I don’t know! I’m trying not to be a blanket naysayer: I’ve been a natural early adopter my whole life, and I don’t plan to stop now. I recently wrote about how I’m using ChatGPT as a motivational writing partner. The older I get, the more problems I see with just about every technology, and I’d like to hold onto the excitement I felt about new tech when I was younger. On the other hand, the problems I see are really big problems, and ignoring those outright doesn’t feel useful either.

So it’s about taking a nimble but nuanced approach: pay attention to both the use cases and the issues around AI, keep looking at organizational needs, the kinds of organic “shadow IT” uses that are popping up as people need them, and figure out where a comfortable line is between ethics, privacy / legal needs, and utility.

At work, I’m going to need to determine an organizational stance on AI, jointly with various other stakeholders. That’s something that I’d like to share in public once we’re ready to roll it out. This post is very much not that — this space is always personal. But, as always, I wanted to share how I’m thinking about exploring.

I’d be curious to hear your thoughts.

· Posts

 

A Seattle Airfield Offers a Rare View of ICE Deportation Flights

"The observation room at Boeing Field offers what is arguably America’s best real-time window into our vast network of privately run deportation flights, a system that has generated troubling reports of passenger mistreatment and in-flight emergencies."

Important work from some pretty brave activists that sheds light on what's being done in our name. Sunlight will hopefully help improve the conditions these immigrants are forced to endure. Ideally the flights would stop completely.

· Links

 

AI news that's fit to print

"What I thought would be helpful, instead, is to survey the current state of AI-powered journalism, from the very bad to really good, and try to draw some lessons from those examples. I'm only speaking for myself today, but this certainly reflects how I'm thinking about the role AI could play in The Times newsroom and beyond."

A pretty good roundup, including the mistakes, folks using AI for pattern-recognition, and newsrooms that are actually using generative models.

· Links

 

How do we make progress in America?

Every American deserves to live well.

2 min read

Someone I follow posted this weekend about how the progressive wing of the Democratic Party was stupid because it consistently pushed for projects that would require higher taxes. I don’t like the framing, and as a self-identified progressive I’m not particularly excited about being called stupid. But there is an underlying political reality about America’s inability to raise taxes which I can grudgingly accept.

I think, though, that a lot of this is about what you get for those taxes. When I moved to the US from the UK, the percentage I paid out of my paycheck in taxes went down (although not by as much as you’d think, given the rhetoric). The amount I had to pay out of pocket for living expenses skyrocketed. It’s far more expensive to live in America than in Europe. Consumer prices are lower, sometimes by a lot; healthcare is free at the point of use; in most places you don’t need to own or run a car.

American taxes don’t seem to be used on infrastructure that most people can actually use. Part of that is the bananas military spending, for sure: a wartime economy instead of one that builds domestically. Part of that is solid opposition from the Republicans, whose modern incarnation appears to hold an Ayn Randian opposition to any kind of policy that could actually help regular people. Part of that is a solid neoliberal streak from the Democrats themselves. All of which is informed, in part, by American public sentiment.

How do we get to the good stuff? Universal healthcare, high-speed rail, integrated public transit, a welfare system that catches people who fall through the cracks, well-funded public education, renewable energy a renewed investment in the arts, public science infrastructure, parks, bike lanes, shared spaces, real programs for the homeless … and so on? Let alone gun control, anti-trust reform, and all those more contentious tasks that seem insurmountable. These all seem important prerequisites for everyone being able to live well, which surely should be the goal. And yet they seem completely, hopelessly unreachable.

Is there hope for the American experiment? And if so: where?

· Asides

 

Doing weeknotes

"Weeknotes are well suited to teams that want to communicate about their work to colleagues or management. But they’re useful in other circumstances, too, such as individuals communicating to the teams they’re part of, or leaders communicating to the people they lead."

This is a pretty great introduction to weeknotes - something that I have to admit I've implemented only sporadically at work, and never on my own site. This page has me reconsidering and thinking about buckling down.

There's a ton of value in both reflecting on and communicating what happened over the last week. Some of my favorite product managers I've worked with, in particular, have done this very well, and it was always a level-setter for the whole team's knowledge.

Something to consider?

· Links

 

AI gives the news you need

I can't share a quote from this one without ruining it. But you should go read it.

· Links

 

The internet, addiction, and me

Sometimes it's not okay to look down from the world.

9 min read

I used to have a night-time routine. I would help my mother up the six stairs from the living room to her bedroom, give her a hug, and set her up in bed. Sometimes, if she was feeling particularly weak, I would bring her toothbrush to her with a mug of water, so that she could brush her teeth in bed.

I could hear the rolling stand that held her food pump against the hardwood floor as she moved around at night, to go to the bathroom. My dad had all the carpeting removed when they bought the house — carpets harbor dust and fungus that could inflame her lungs.

Years out from a double lung transplant, it was no longer the pulmonary fibrosis that was causing her pain: it was the anti-rejection drugs. The operation had saved her life, but it was far from a magic bullet. For eight years, she seemed to go from near-death experience to near-death experience: operations to remove scarring on her lungs, fungal infections, feeding tubes, inability to eat, nausea, pain. In 2019, we spent eleven straight weeks by her bedside. In 2020, the silver lining of the pandemic was that I no longer had to go into an office, and could spend most of my time helping to care for her. In 2021, on an awful Sunday evening in June, we lost her.

She fought for over a decade. Even at the end, she said she wasn’t ready to say goodbye. She still had life left. She didn’t want to leave us.

There are so many things I want to tell her; so many things I want to talk through with her. There’s so much I want to apologize for, too: she had told us, over and over again, that she didn’t want to die in a hospital. In some of her last lucid moments, she tried to remove the tubes on her arms. “This is not okay,” she said. Palliative care, which is supposed to be about making her as comfortable as possible, seemed in the end to be about making us as comfortable as possible. They starved her. I watched as my sedated, unconscious mother starved to death in a hospital bed.

This is not okay.

I feel compelled to go back to that hospital room, as if she’ll be waiting for me there. When I was still in San Francisco, I’d walk by the hospital and look up at the corner room, facing the trees on the hillside, hoping to see her silhouette.

I wish she would show up in my dreams, so I could at least talk to a version of her, even if I intellectually know it would just be my own projection. She hasn’t shown up there once, except as a brief staccato “oh my god, you guys” that came out of nowhere and woke me up like a nightmare.

The morning she died, I collapsed into Erin; I’m not ready, I said, over and over, as if it could change anything.

I’m not ready.

I will never be ready.

I came back to Britain for my friend’s wedding a year after her lung transplant. I didn’t stay long: whenever I went anywhere, there was always the fear that something would happen. But I’d ripped my life apart to come to California to be with her, and returning there made me feel at least a little bit connected to what my life had been. I saw my friends, I saw the places that used to be home to me. But rather than slotting back, there was a bittersweetness to everything. It had all changed, my life and theirs, and this couldn’t be home to me anymore. I was severed.

I gave a presentation about the indieweb at an Edinburgh TechMeetup where my laptop had frozen up and needed to be hard-rebooted halfway through. Afterwards, we all gathered at a nearby pub, and a prominent member of the Edinburgh tech scene said to me, “I wouldn’t have gone. I would have said, ‘sorry, Mum, you made the choice to move there’.” I couldn’t understand, and I still can’t. She had never met my mother. She would never understand who my mother was. And she misunderstood me if she thought I would ever say that. (Did I do the wrong thing?, I asked myself that night, and for years afterwards, over and over.)

Ma’s illness was genetic. We’ve lost five members of our family — people we dearly loved. Researchers were finally able to figure out how to identify the relevant mutation in the TERT gene, which eventually led to my sister and I getting cleared. But, of course, the science is evolving; there’s no complete guarantee that we are actually cleared. It will hover over us forever either way: we lost people we dearly love to this thing as recently as this summer, so any relief we might have felt was painfully hollow.

Holy shit, did it fuck me up.

I remember my first experience of really feeling different when I was around eight years old; the dawning understanding in my third-culture mind that people saw me as some kind of other. One boy used to drag me into the ditch at the side of the school playing field and just jump on me, as if he was trying to break my legs. The teachers at my school mocked me for having a German name; forty years later, the war still weighed heavily for them. I have wondered if they would have acted differently if they’d known my Jewish heritage, but honestly, I don’t think it would have mattered. I wasn’t one of them, was the thing; I was Other.

When I was a teenager, I became so tall that I often loomed over people. My new presence attracted yet more attention, and I grew to hate the looming hugeness of my body, this bounding form that people found it necessary to laugh at. I wished I could have disappeared. I wished I could have been normal. I fantasized that there was a magic word that other people knew that I didn’t, and if I could only figure out how to invoke this special incantation, I would finally feel like I was okay.

So when this happened, when I tore my life to bits at the hands of this terrible terminal disease, I felt like I deserved it. I didn’t feel like Ma deserved it; I didn’t feel like my dad deserved it; I didn’t feel like my sister deserved it; I didn’t feel like the other members of my family deserved it. Intellectually, I don’t believe in fate or karma. Nonetheless, I deserved it. Of course I did.

The internet, though. Here was a place where I could write something, or take a photo, or build some software and release it, and the world would respond. Every response was a distraction from what was actually happening. This other world, not so much a backchannel to real life as a parallel universe with its own culture and rules, could take me away, just as it had when I was a teenager. Even then, I would check for new messages relentlessly, dialing up to Demon Internet and logging in many times during a long, after-school evening. Now, decades later, the web seemed infinite, and there was always something new to say, to get involved in. It was a balm, and then an addiction, and then a distraction. A way to feel less worthless. And whereas my teenage self had needed to dial up from the desktop computer in his bedroom after school, the iPhone gave me access to it anywhere.

I wrote recently about needing to pull back from social media. It’s not the first time I’ve written a post like this: it’s been a cycle of addiction. But I don’t think I’ve ever written in depth about why I needed that back-channel. It’s sometimes easier to look down at the device in your hands than take life squarely in the face.

But that doesn’t mean the escape is reasonable, or healthy, or right. There were times over the last fifteen years when I needed to be present in the moment and I just wasn’t: when I was racking up points sharing links rather than facing up to what was happening. It made me avoidant and disconnected; untethered me from the world. It dulled my empathy and let me remove myself from it all.

This isn’t a “screens are bad” post. It is a post about being forced to reckon with what actually matters, even when it’s hard to look at.

Maybe what I’m saying doesn’t make sense to you. But when I say I want to remove myself from social media, when I don’t think it’s good for me, and when I keep coming back, this is what I mean. This is what’s happening.

Which means the indieweb isn’t just a technology movement to me. It’s a way of reclaiming more of myself. And in that light, perhaps I should just own my mind and switch it all off for good.

· Asides

 

It’s snowing again. That’s it, I’m moving to Spain.

· Statuses