Reading tech-news has been a central part of what I do for a living for 20+ years. This results in numerous intriguing links to share. On Facebook, on Twitter, on my personal blog. Sometimes in a more structured way, mostly more ad hoc.
When I started this site, link-blogging was a part of the ambition. The 'l' in myttl – "my thoughts, tools, and links".
But I haven't come around to it. Until now.
Thanks to Apple Shortcuts, I have finally a process that I believe will allow me to maintain some consistency with link-blogging here moving forward.
My goal is to publish one post every weekend. With things I've encountered over the week. Important, fun, useful. A lot of technology, from both a end-user and societal standpoint. Perhaps a little photography stuff. Etc.
(I share most of the links that will be included in my weekend posts on my Mastodon account throughout the week. Feel free to follow me there!)
Ethan Mollick's blog One useful thing has become one of my favourites right now. A lot of good posts on AI and society, but also delves into the practical aspects of LLMs and more.. In this post, he takes a step back from the news cycle and peers into the future, where AI not only can “think” (LLMs), but also have eyes, ears, and a voice.
It’s hard to not be impressed with what multi-modal models will be (are!) capable of.
But it’s also clear that we’ll need to guard our agency over the technology.
Speaking of news cycles. This list of tools for fact checking and investigative journalism is probably useful for others than reporters.
I do a lot of LLM explorations at the moment, not the least using the Text Generator plugin together with Obsidian. Primarily not for generating text, but assisting me in my own writing. Having a LLM as an editor is increasingly becoming a part of my writing workflow. And in that process, getting inspiration from prompting is key. This is a good collection of prompt tips...
...but the question is for how long the ordinary user will have to think about prompting at all. There are already tendencies where the LLMs become integrated into the tools we already use, Microsoft Copilot in Office 365 perhaps the best example at the moment. Kevin Roose and Casey Newton explores the future of prompt engineering together with Scale AI's Riley Goodside in this episode of Hard Fork.
This is also hints at a future were the LLMs are integrated in new ways: OpenAI considering becoming a hardware company?
Ending this first link-blog with something completely different: Nature’s gradients. Wonderful pictures by Om Malik, but also a scientific explanation on the science behind the colors in the sky.