Another Crazy Day in AI: The Invisible Army Behind AI
- Wowza Team
- 6 days ago
- 5 min read
Hello, AI Enthusiasts.
It’s that Thursday stretch where your mind’s halfway out the door, but there’s still time for a few good stories.
A new episode of The Neuron podcast peels back the curtain on the real people behind AI’s polished outputs. Caspar Eliot from Invisible Technologies shares how thousands of human trainers quietly power the world’s smartest systems. Real people label data, check responses, and quietly shape the way AI learns.
Meanwhile, Shopify’s seeing record growth from AI-powered shopping tools as “agentic commerce” starts turning chats into checkouts.
And Google’s expanding Chrome’s AI Mode to more countries and languages, making it easier than ever to dig deep into topics right from your phone.
Weekend’s close enough to taste.
Here's another crazy day in AI:
The people behind AI training
Shopify sees AI shopping surge in 2025
Google makes AI mode easier on iOS and Android
Some AI tools to try out
TODAY'S FEATURED ITEM: The Truth About AI Development
Image Credit: Wowza (created with Ideogram)
Who's actually teaching AI models how to respond intelligently?
There's a common assumption that AI models like ChatGPT and Claude learn everything by scanning the internet. The reality involves considerably more human effort than most people realize. In a recent episode of The Neuron podcast, hosts Corey Noles and Grant Harvey speak with Caspar Eliot from Invisible Technologies about the extensive human workforce involved in AI development. Invisible Technologies has worked on training 80% of the world's leading AI models, and Caspar walks through the actual processes that make these systems functional, from data labeling and response evaluation to the specialized expertise now required as models become more sophisticated. The conversation covers both the technical foundations and the practical business challenges companies face when implementing AI solutions.
A few things they dug into:
Training AI models requires three main components: giving them high-quality datasets to learn from, fine-tuning their behavior through targeted examples, and evaluating whether the changes actually improve performance
Behind every major AI model are people spending countless hours labeling data, rating responses, teaching models through feedback loops, and catching biases that automated systems can't detect
AI works by predicting the next most likely word or token, not by actually understanding content, which is why human judgment remains essential for determining if outputs are genuinely useful
The expertise required has evolved significantly. Early training relied on general annotators, but now companies need specialists in fields like healthcare, linguistics, or technical domains
Most businesses face foundational issues before AI can help them: poor data organization, processes they can't clearly explain, or attempting to automate tasks that weren't well-designed in the first place
Physical AI applications like self-driving cars and robotics will bring a new set of challenges around safety testing, legal responsibility, and getting people comfortable with the technology
Evaluating AI improvements isn't straightforward. A model might become more creative but also more prone to errors, making it hard to define what "better" actually means in practice
Valuable skills for AI work come from surprising places. The rapid decision-making and pattern recognition from competitive gaming, for example, translates well to prompt engineering and system design
One thing that comes through clearly is how different the day-to-day reality of AI is from what gets discussed in headlines. Caspar mentions he doesn't use AI tools much himself, mainly just for transcribing handwritten notes because he focuses better when writing by hand. That's a telling detail. His observation about model development potentially stopping without most users noticing for years suggests we might be paying attention to the wrong metrics when tracking progress. The example of analyzing basketball footage for the Charlotte Hornets shows how AI can do things that would be impractical manually, but someone still needs to decide what to look for and what the results mean.
The conversation also gets into something that doesn't always come up in AI discussions: implementation is messy. Companies often have data spread across disconnected systems or workflows that nobody can fully explain. Adding AI to that situation doesn't solve the underlying problems. Caspar's take on job displacement is more nuanced than the usual predictions. He points out that people don't behave like purely rational economic actors, and there will probably always be value in human judgment and interpersonal work. Whether AI ends up replacing jobs or just changing them likely depends on choices people and organizations make, not just on what the technology can theoretically do. It's an honest look at where things actually stand rather than where we imagine they might be headed.
Watch it on YouTube here.
Listen on Apple Podcasts here.
Listen on Spotify here.
OTHER INTERESTING AI HIGHLIGHTS:
Shopify Sees AI Shopping Surge in 2025
/Sarah Perez, Consumer News Editor, on TechCrunch
Shopify reported a major rise in AI-driven shopping, with traffic from AI tools increasing sevenfold and orders up elevenfold since January. The company credits partnerships with OpenAI, Perplexity, and Microsoft Copilot for helping shape a new era of "agentic commerce" where AI agents assist users directly in chat. Shopify President Harley Finkelstein said AI now drives both customer tools and internal decisions. With most shoppers open to using AI when making purchases, Shopify is positioning itself at the forefront of this transformation in online retail.
Read more here.
Google makes AI Mode easier on iOS and Android
/Nick Kim Sexton, Senior Product Manager, Chrome, on Google Blogs – The Keyword
Google is expanding access to AI Mode in Chrome to make it easier for mobile users to explore complex topics. The new update adds an AI Mode button under the search bar in new tabs on iOS and Android, allowing for deeper and more intuitive queries. The feature is currently available in the United States and will soon reach 160 additional countries and languages including Hindi, Japanese, and Portuguese. This update strengthens Google’s effort to make AI-powered search seamless across all devices.
Read more here.
SOME AI TOOLS TO TRY OUT:
That’s a wrap on today’s Almost Daily craziness.
Catch us almost every day—almost! 😉
EXCITING NEWS:
The Another Crazy Day in AI newsletter is on LinkedIn!!!
Leveraging AI for Enhanced Content: As part of our commitment to exploring new technologies, we used AI to help curate and refine our newsletters. This enriches our content and keeps us at the forefront of digital innovation, ensuring you stay informed with the latest trends and developments.
Comments