Everyone interested in discussing this topic should listen to the latest Minnmax podcast with Erik Wolpaw. They start talking about AI around an 1 hour and 10 minutes in. Wolpaw works for Valve currently, and used to work for Double Fine. He’s also half of Old Man Murray for all you old people out there. He’s worked on Portal, Portal 2, Psychonauts, Half Life: Alyx, etc. The discussion was good.
4 Likes
1 Like
The guide writing scene is already hanging by a thread. It’s night and day the difference between looking up info about a game released recently and a game released before, say, 2020 (somewhat arbitrary, the problem has been a long slide without a single clear inflection point).
Something I have run into countless times is a resource from a gaming website that did SEO well enough to rise to the top of page 1 Google search results, but when I click through to the guide it is either:
-
It’s only half-written - like a guide to all the collectible items that only covers the first 5 chapters of a 12-chapter game - with a note that the author/site will add to it as they discover more… but the game came out 2 years ago and it’s blatantly obvious nobody is ever coming back to finish this guide.
-
It’s not even half-written. It’s a template that has places for information to be added, but no information has been added to any of the places yet. I don’t know if the intention is for these pages to be crowd-sourced, or if it’s similar to #1 and the site is pretending that they were ever going to fill in the information even though they’re obviously never going to do it.
-
Duplicates. 8 different gaming sites that have collectible lists covering only the first 5 chapters of a 12-chapter game, because they just scraped the half-assed effort made by #1 and called it a day. Often these copycats don’t even have a note saying more will be added later.
In our attention economy and 24-hour news cycle, being first to publish is money. The financial incentive is to get maximum eyeballs on content, and “publishing first” maximizes eyeballs more than “publishing well.” While spending a few days after release making the content better might boost social sharing of your content, it probably won’t boost it by enough to be more profitable than cranking out low-quality content about two new releases, each on their release day, before the web is flooded with competition.
People don’t want to hear this (frankly I don’t want to hear it), but at this point I’ve seen enough to know it’s the truth: Journalism, if it’s to be a business and not a hobby, needs subscribers to produce high-quality content. A subscription model makes a media site less desperate to monetize attention at any cost and instead incentivizes the cultivation of a good journalistic reputation.
But I grimace as much as the next person when I click a search result and hit a paywall, and often immediately hit the back button to look for a free source of information instead. It’s really hard for a subscription service to compete with free, even when it’s infinitely higher quality or more useful. I do have a few subscriptions to different newspapers and magazines, but it has to be something I read/would read regularly to feel worth it.
People can click into way more different sites than they can afford to subscribe to - there’s just less room in the market for multiple subscription-based publications to thrive. Every site that can’t sustain itself with subscribers is left with no choice but to try to monetize fleeting attention.
Well I see that I just wrote a small novel here, kudos and appreciation to anyone who read it all…
TLDR: AI is just accelerating the death of journalism that was already set in motion some time ago when SEO and ad views supplanted reputation and subscriptions as the primary currency for writers.
3 Likes
1 Like
Triple whammy of Sora shutting down, Disney torching the OpenAI deal, and Baltimore suing xAI a few days ago was hilarious. Wish it made up for the ecocidal data centers and economic destruction.
5 Likes
At work I recently had a quarterly business review with our email marketing platform account rep. This includes sharing some of their upcoming roadmap. She shared that they’ve pulled back on some genAI they’d tried to integrate into the platform, because nobody was really using it. They did some client surveys to understand why and the two things that emerged were, 1) because their platform is used by enterprise-tier businesses, that means highly specialized roles, which means the person who sets up an email in the email platform is very likely not the person who writes the email, and 2) “What we heard is you all don’t want AI to write the emails for you. You do want it to be able to give you feedback and suggestions you can consider, but you want to be writing the messages yourself.”
So now that’s what’s on their roadmap, a less intrusive AI that you can bounce ideas off and get suggestions from, rather than an AI that’s coming for your exact job.
2 Likes
2 Likes
2 Likes
This thread on the Claude source code leaks is a goldmine of hilarity:
1 Like
1 Like
“in cases of uncertainty many people would rather be told to do something than to choose to do it” reminds me of this, which I just noticed hasn’t been posted here!
The 4-sentence summary:
When Maria Park, KRAFTON’s head of corporate development, told Kim a “dismissal with cause” would not rid the company of its $250 million bonus obligation without exposing the company to “lawsuit and reputation risk,” Kim looked toward an AI chatbot for guidance.
Kim, spooked by what he privately called a “pushover” deal, bypassed his own legal team and turned to ChatGPT for help. When the AI chatbot responded that the earnout would be “difficult to cancel,” the ruling read, Kim didn’t accept the answer. He pushed further—and the chatbot obliged with a detailed, multi-stage corporate takeover strategy dubbed “Project X.”
This guy wanted to be told to do… the thing he already wanted to do, so he kept asking for orders until he got the orders he wanted to follow. I don’t know how else to explain that other than he must have actually believed that ChatGPT couldn’t be wrong, and that if it gave him a plan, the plan would work.
This whole “agentic” chapter of AI is just incomprehensible to me. The amount of control people are happily giving up, and for what?
I work in marketing for an organization with a very buttoned-up image and low risk tolerance, and marketing is one of the fields that is most overrun by AI being shoehorned into our tools. Some of them are occasionally useful, but by far most of the features these tools try to push on me are effectively asking me to put my employer’s name (externally) and my name (internally), not just on work that I didn’t do, but on work that I can’t effectively review/QA. Who has that kind of risk tolerance, to just let stuff be done in your name without you even being aware of it??
Asana has this new AI Studio that they’ve turned their automated rule editor into, that tries to default you into using AI by prefilling various fields with AI tokens that you have to actively delete if you want to actually set the values of those fields yourself. Apparently their vision for how this tool works is I create a rule like, “When a Task is added to this Project, create Subtasks.” Then I can tell the AI to determine how many subtasks there should be and what they should be called and I’m assuming probably what should be in their descriptions, as well. Once I’ve set up this content-less skeleton of a workflow, I then type instructions into a chatbot window in the corner where I explain in plain English what I want to accomplish.
You know, if this was just an assistant that would help me understand how to make custom rules by walking me through the steps after getting my instructions, that would be moderately useful and I might even use it time to time.
But it doesn’t do that. If it took my instructions and then used them to fill all those fields in the rule editor with the appropriate content based on what I’m looking for, I would probably not use it, but if I was feeling really stuck on something I might decide it would be worth the hassle of letting it write the rules and then have me go in and try to understand how it turned my instructions into rules, so I could do this for myself in the future.
But it doesn’t do that either. It just shows little “Use AI” tokens in all the fields and it fucking generates the field values/contents on the fly, in real time, as tasks are added to the project. Even though I would have the ability to QA and edit these, it could be hours or even days, if I’m out of office, before I notice that some nonsense task has been created and has assigned nonsense work to other people, who by that point may have already acted on the nonsense assignment. How can anyone think this is a viable way of doing business? These things don’t know what truth or reality are, why in the world is anyone letting them make decisions?! How are people comfortable turning something like that loose?
2 Likes
I did share a Bluesky post from Rami Ismail about it, but I always welcome more sharing of this story because it’s hilarious in its stupidity.
It’s truly mind boggling to me. I can’t imagine releasing control over to something like that, how could I ever respect myself, or expect anyone to respect my work if I did?
2 Likes
I haven’t been as emotionally engaged as I want to be lately enough to talk about AI like I want. But I saw this video by Caelan, which have been completely escalating in severity and desperation and I can’t help but feel horrified at just how powerless we are to stop horrific things corporations like OpenAI have been encouraging for the sake of engagement and profit.
2 Likes
1 Like
You couldn’t make a satire this good:
2 Likes
Asking a robot what to drink seems…idiotic:
1 Like
1 Like
1 Like
Polygon is now using AI to summarize the comments under each article. As if the sale to Valet wasn’t bad enough, but I guess AI is par for the course when it comes to Valet owned sites.
1 Like