Henk van Ess's article DeepSeek: The Real Price of China's AI investigates DeepSeek, a Chinese AI company, whose seemingly inexpensive ChatGPT competitor masks a sophisticated system of state control. DeepSeek's AI, while appearing cost-effective on the surface, is demonstrably limited in independent thought and strictly adheres to Chinese Communist Party propaganda. Today’s episode guides are, in style, the virtual hosts of NotebookLM.

Ever wish a chatbot could actually see what’s in a PDF instead of just skimming the text? Well, Claude—the AI chatbot—is now able to analyse the visual information.

Imagine the potential. From financial analysts spotting hidden profit drops in charts, to journalists exposing fake health diagnoses a bit quicker with the help of chatbot-aided discoveries. Suddenly, AI isn’t just reading documents; it’s seeing the whole picture (literally).

In this episode, I’ll break down why this shift is a deal for researchers, journalists, and anyone who’s ever spent hours dissecting official documents to find the story in the data. Well, it’s not me speaking, but my two AI-assistants.

(Below you see what the problem is with analyzing PDF’s with chatbots - but not anymore.)

Don’t forget to read the full article about VisualPDF:

Paul Myers (BBC Verify) and me, Henk van Ess sat down in London to discuss a pressing issue: the shitification of Google. Yes, that’s right—Google has slowly been losing its touch. But hold on—Paul wants me to clarify that these are his own spicy hot takes, and in no way represent the fine, polished opinions of the BBC!

Shitification—yes, that’s a thing now! It’s the slow, painful, infuriating decline of services we once loved and depended on, now reduced to hollow shells of their former glory. Enter: Google.

First, they murdered Google Cache. Once upon a time, you could view a saved copy of a website that’s vanished from the web. And now? That magical button has gone to the place where hopes and dreams die. Why? Because Google said, “You don’t need it, darling!”

But wait, it gets worse. How many search results do you get now? 300. Yes, just 300 out of possibly millions! So unless you’re looking for Disney’s Mickey Mouse, your obscure little website is living in Google’s black hole, probably with those old Yahoo! email accounts we all forgot about.

And don’t even get us started on reverse image search. Once a glorious feature, now bastardized by Google Lens, which, let’s be honest, is just a glorified game of “What’s that blurry thing in your photo?” It’s useful for identifying dog breeds, but can be terrible for hardcore OSINT researchers like us who want to verify images based on sites and dates.

Google is a bit like a high-end coffee shop that suddenly decided to serve instant. You still go there, but you’re deeply disappointed. Shitification has arrived, folks, and Google’s the poster child. Paul and me discuss possible solutions. Enjoy!

Welcome to Digital Digging, I’m Henk van Ess, and today I talk about summarizing with AI chatbots. You’ve probably tried it—just typing “Summarize this” into ChatGPT and hoping for the best. Don’t! It’s like asking for a full meal and getting a side salad.

In this episode, I cover some highlights from my article “Do Not Use ‘Summarize This’” and break down how to get the summary you actually need. Whether you’re trying to condense War and Peace—which, spoiler alert, is more than just ‘Russia had a rough time, some people fell in love’—or customizing your prompts for better AI results, I’ve got you covered!

Following two AI-generated episodes, I am delighted to announce that this one is entirely human-created - taped via my phone.

Join me and Chris Lunday from Business Insider for a 15-minute discussion on integrating AI with research, particularly in the field of open-source intelligence (OSINT). This conversation is ideal for individuals interested in conducting web research with large language models.

In this podcast, we hear from Lunday, a 29-year-old reporter from Berlin, how you can waltz with Google and AI to find the right answers to your research. Chris , like many of us, has been exploring the world of AI-assisted investigation.

Open Source Intelligence, or OSINT for those who love a good acronym, is all about finding publicly available information online. AI is making this process easier than ever before.

Chris and his fellow students from the Axel Springer Academy of journalism & technology in Berlin discovered that they could supercharge their Google searches in ways that would make even the most tech-savvy person jealous.

One of the most intriguing examples they tackled involved finding services in Germany that help people keep loved ones' ashes at home (which, surprisingly, isn't allowed there).

Using ChatGPT, they crafted the perfect Google search to find these underground urn-keepers. The secret? Focusing on WhatsApp numbers, because nothing says "totally legit business" like advertising your services via instant messaging.

But the real showstopper was a challenge to find a specific location based on a photo of a Tesla charging station near an Ulta Beauty store. It's like a nerdy version of "Where's Waldo?" but with electric cars and makeup.

Chris used ChatGPT to generate super-specific Google searches to find all the Tesla charging stations and Ulta Beauty stores in Phoenix. The AI wasn't perfect – it initially missed a few locations – but with some gentle prodding (read: relentless questioning), it eventually coughed up the correct information.

So, what do we learn from this conversation?

1. AI isn't replacing human smarts; but it can make our Google searches exponentially better.

2. Always question your AI. It's like dealing with a very knowledgeable toddler – impressive, but prone to making stuff up.

3. The future of problem-solving might involve less head-scratching and more chatbot-wrangling.

Now, if you'll excuse me, I need to go ask ChatGPT how to end this article with a witty one-liner.

As we waltz into AI-assisted sleuthing, remember: the key is in the delicate dance between human intuition and machine precision.

Why is the flight so fast? At first glance, it seemed simple – tailwinds, right? We unravel the mystery of a New York to Phoenix flight with the help of ChatGPT through a maze of weather maps, AI analyses, and flight data. We'll challenge common assumptions, crunch some numbers, and even corner a few pilots.

What I learned:

* Hypothesis Hyperdriver: AI can cook up initial theories faster than you can say "jetlag," tapping into a massive database of factors like weather patterns and flight stats.

* Pattern Recognition : AI's got eagle eyes for spotting anomalies in historical flight data, flagging anything funky that might explain our unexpectedly speedy journey.

But watch out for these AI airsickness moments:

* Garbage In, Garbage Out: The quality of AI's insights is only as good as the questions we ask. Too narrow, and we might miss the forest for the trees (or the jet stream for the tailwind, in our case).

* Real-World Reality Check: While AI can crunch numbers like a champ, it might struggle with nuanced real-world concepts. .

* Bias Turbulence: AI models can inherit biases from their training data, potentially leading us down a wrong flight path if we're not careful.

* Not a Solo Pilot: AI's great, but it can't replace good old-fashioned human expertise. Always cross-check with the pros (like our friendly neighborhood pilots) for the full picture.

Combine AI's data-crunching superpowers with human critical thinking and expert knowledge, and you’re good to go.

In this first episode of Digital Digging, we explore how to spot AI-generated content and the tools that can help in this process. Henk van Ess, known for his expertise in digital research, takes a step back and allows AI to do the heavy lifting—his voice and the episode’s content have been fully generated using AI technologies. Two AI co-hosts, Castor and Pollux, guide listeners through the discussion, bringing together insights from Henk’s recent article on the topic.