2025-05-20 05:45:00
Joe Wilkins writing for Futurism: AI Chatbots Are Becoming Even Worse at Summarizing Data
Alarmingly, the LLMs' rate of error was found to increase the newer the chatbot was — the exact opposite of what AI industry leaders have been promising us. This is in addition to a correlation between an LLM's tendency to overgeneralize with how widely used it is, "posing a significant risk of large-scale misinterpretations of research findings," according to the study's authors.
Yesterday I posted about how we don't really understand why (not "how", but "why") LLMs work as well as they do, and this study showing that the most advanced the chatbot gets, the more likely it is to make certain errors around data summarization is fascinating. In early 2023 AI boosters were like "if we see LLMs get better at this rate forever, they're going to be insanely smart in no time!" Yet, since GPT-4's release 2 years ago, it really feels like the advances are pretty minor. Sure, the tooling is getting better, but most of these gains seem to be to be us finding more effective ways to use the models, not the models themselves getting meaningfully better. Even the "thinking" models we've seen post Deepseek R1 have been largely been customized models that are to to use chain-of-thought responses which we've found helps them get to better answers.
I'm not denying that models are improving, but I just look at my own experience and stories like this, and it really feels like we had a massive breakthrough in late 2022 that lead to ChatGPT's historic launch, we had some meaningful improvements in the 6 months after that, and ever since then the model gains have been marginal. I'm sure an expert could tell me all the ways the new models are better, but I just don't feel it from the user side.
I guess what I'm getting at is I'm still not convinced that LLMs are on an unstoppable march upwards. Maybe something will change and we'll understand "why" they work or some other major insights that enable us to make some breakthroughs, but it just feels to me like we're finding new and interesting ways to use models of a general performance level to enhance their outputs rather than seeing the models themselves approach anything remotely close to "AI". And honestly, LLMs are still a massively impactful technology, even if they never get much better for day-to-day things than they are today.
2025-05-20 05:30:00
Johnny Ryan: EU ruling: tracking-based advertising by Google, Microsoft, Amazon, X, across Europe has no legal basis
Dr Johnny Ryan said "Today's court's decision shows that the consent system used by Google, Amazon, X, Microsoft, deceives hundreds of millions of Europeans. The tech industry has sought to hide its vast data breach behind sham consent popups. Tech companies turned the GDPR into a daily nuisance rather than a shield for people."
Man, 2025 is the year of malicious compliance getting the old side eye from regulators.
2025-05-20 05:15:00
Werner in the Things blog: A Swift Cloud
We’ve rebuilt Things Cloud from the ground up. The system that keeps your to-dos in sync is now faster, built with modern technology, and ready for the future.
This is gonna sound like a burn, and I don't mean it that way, but is there anything more iconic than Cultured Code working on a major update that improves Things in significant ways that most users will never notice? I know many are eagerly awaiting a major update to Things, and I too will be excited when it comes, but honestly, what I love about Things is that it works great, is perfectly reliable, and just works in the way I need it to. I'm not dying for it to get any significant updates.
2025-05-20 05:00:00
Glenn Fleishman writing for Six Colors: Can We Still Love Apple? Should We Ever Have?
Maybe it is the right time for this love affair to come to an end. Not the end of my love for what I can do with Apple stuff, but creating boundaries, something good for any relationship. From Tim Cook down, executives—Schiller excepted—have proven themselves unworthy of our trust. As shepherds of the company, they have revealed themselves. I may still love the concept of Apple, but certainly the company no more.
This rings very true to me. It's the dichotomy of having a strong emotional connection to the products that a company makes without thinking about the company in those same terms. As I wrote last year:
I’m sure Apple will continue to be very successful for many years to come and I expect to buy many products in the future as well (after all, Microsoft and Google don’t feel much better). I’ll surely even give some of those products glowing reviews on this very blog. And yet, I do wonder if the Apple enthusiast crowd as we know is in permanent decline.
That piece was about how the Apple relationship with its fans was always different than most other companies and that relationship was feeling more bog-standard transactional. In the year since then, I really think that piece holds up in terms of predicting the vibe shift in the Apple enthusiast community that's occurred since then.
I'm absolutely looking forward to WWDC because it's where I'll get to see how the products in my life are going to get better in the coming year, and I'm excited because it's a unique time where everyone in my Apple-centric online circles is particularly locked in and chattering all about nerdy stuff. It'll be fun, it'll be exciting, but it won't be quite what it once was.
Near the end, he said this, which I resonated with as well:
We are often more frank about things we love in describing their flaws than those we hate because we care enough to want them to improve.
There are so many topics that I never discuss on this blog. It's not because I think they're perfect and free from criticism. It's more that I don't think about them enough to have a strong enough opinion to share anything here. I feel like a cliché parent here, but truly, I'm critical because I care.
2025-05-19 09:01:42
This is such a good, quick video explaining why we don't really understand why LLMs work the way thy do.
2025-05-19 06:51:39