AI Daily news June 25: 🏀AI is changing the way NBA teams evaluate talent 💻 Google launches Gemini CLI: a free, open source coding agent 📊OpenAI’s Workspace, Office comp ⚖️Judge rules Anthropic AI book training is fair use 📚 AI Help with Grading
14:08 Share

Okay, let's unpack this. We've got just a whirlwind of AI innovation landing on our desk. Time for a deep dive. But hey, before we jump right in, just a quick note. This deep dive is a brand new episode of AI Unraveled. It's created and produced by Etienne Newman. He's a senior engineer and passionate soccer dad up in Canada. Please remember to like and subscribe to AI Unraveled. You'll get loads more insights like these into the wild world of AI.

Right. So our mission today, we're going to cut through all the noise, give you a shortcut to being really well informed. We're focusing on this fascinating daily chronicle of AI innovations from literally just yesterday, June 25th, 2025. It's kind of amazing.

How much can happen in 24 hours? It really is. I mean, what jumps out immediately from the sources we looked at for this deep dive is just the sheer breadth of what's happening. You know, in just one day, we're seeing major court rulings, completely new applications, and even people grappling with these big existential questions about AI. It just powerfully underlines how fast this is all moving across pretty much every sector. Moving so fast.

So let's maybe zoom in a bit from that big picture. Let's look at where AI is making an immediate impact in our daily lives, thinking particularly about work and education. It feels like it's becoming less about AI taking over jobs and much more about AI being like a really powerful co-pilot. Take Walmart, for instance. Huge company. They've just unveiled a whole suite of AI-powered apps for their 1.5 million associates. We're talking 10%.

tools designed to streamline things like onboarding new staff, managing schedules, which must be incredibly complex, and even giving real-time advice for customer support. How significant is that really? Seeing AI adopted on that kind of massive scale. Well, that's a critical point, and it definitely...

points to a bigger trend. If you step back and look at the bigger picture, these applications are fundamentally about empowerment, not really displacement. For a company like Walmart, it's about, yes, boosting efficiency, but also importantly, maybe improving employee satisfaction. You know, by taking away some of those tedious, repetitive tasks, it lets the human staff focus on, let's say,

higher value interactions, solving trickier problems. It's not just speed, it's shifting the focus. Right, shifting focus from the mundane to the more complex. Yeah. And it's not just giant corporations either. We're seeing educators finding real value here too. Teachers are using AI for personalized feedback on student work, helping with grading, even drafting initial lesson plans. A lot of them are saying it's a significant time and actually improves the quality of their teaching. It's like having an extra pair of hands, an efficient assistant. Exactly. For teachers, AI is becoming this vital technology

teacher's assistant. It can directly help reduce burnout, which is a huge issue. And at the same time, it can enhance instruction, allowing for more personal

more personalized approaches for different students. So this really brings up that important question. How does AI best support human roles? How does it augment them rather than just replacing them? These examples, both in retail and education, they seem to heavily lean into that support role, redefining that human AI partnership. OK, so AI is definitely enhancing roles in places we know well, but it's also pushing boundaries, isn't it, into new areas, sometimes surprising ones.

And at the same time, it's raising these tricky questions about what it can really do. This is where it gets super interesting. So on the power side of things, Google just launched something called Gemini CLI. It's a free open source coding agent. It uses their pretty powerful Gemini Pro 2.5 model. And get this, developers get 1,000 daily requests free of charge.

For anyone less familiar, CLI is command line interface, basically. Makes it easy for developers to talk directly to the model. Plus, it's got this extensibility architecture, meaning devs can plug in other services, add new features. That sounds like a big deal for solo developers or small teams, right? Oh, definitely. That lowers the barrier to entry quite a bit. And then you've got OpenAI. Reports are they're developing productivity tools right inside ChatGPT. Tools that sound a lot like Google Workspace or Microsoft Office.

Things like real-time document collaboration, chat for multiple users, a record mode for making transcripts, uploading files directly to projects, and connectors to pull in data from places like Teams, Google Drive, Dropbox. It sounds like they're building a whole work platform in there. Yeah, this push by OpenAI is...

Well, it's a profoundly significant strategic move. It really is. Sam Altman, their CEO, he warned last year they'd sort of steamroll startups, remember? But now they seem to be stepping directly onto Microsoft's traditional turf. You know, Microsoft, their major investor. This aggressive play, it kind of highlights the incredible revenue projections they must have, too. Apparently, business subscriptions brought in $600 million in 2024, and they're projecting something like $15 billion by 2030, mostly from these enterprise deals.

It's a serious bid for market dominance in productivity. Makes you wonder how the rest of the tech world is going to react. Yeah, it really does. But OK, on the flip side of all this power, we're also getting a clearer view of AIs, maybe puzzling limits. There was this Apple paper titled The Illusion of Thinking, revealed something quite surprising. Apple researchers found that these large reasoning models, LRMs they call them, they do OK on simple or medium difficulty puzzles.

But their accuracy just collapses like really sharply as the puzzles get more complex. Even when they should have enough token capacity, enough sort of computational room to work, the models just seem to give up. It suggests their reasoning might be brittle, limited. That is a fascinating finding. But, you know, it does raise an important question. Are these true limitations in reasoning ability or are they just

Or are they maybe just artifacts of how we're testing them? Critics are already arguing that Apple's findings might actually reflect engineering issues, not real reasoning limits. For example, maybe some output token limits were hit causing that collapse, or maybe they used puzzle versions that were actually unsolvable, which isn't really fair to the model. And what's really compelling is that when they reformulated the puzzles, like

asking for a different kind of answer, the models often did much better. So the tension here is pretty profound, isn't it? It's between what looks like reasoning and what might just be a result of how we measure it. It just underscores how critical it is to have really robust, carefully designed ways to evaluate AI. We need to be so careful about saying what AI can and can't do. Well, yeah, it's truly fascinating exploring these nuances, the capabilities and the limitations. And actually, for those of you listening who are thinking, okay, I want to understand this stuff on a deeper level.

or maybe even I want to start building with AI. Well, Etchen Newman's resources are designed for exactly that. You can find his AI certification prep books, things like the Azure AI Engineer Associate, the Google Cloud Generative AI Leader Certification, AWS Certified AI Practitioner Study Guide, Azure AI Fundamentals, Google Machine Learning Certification. They're all at djamgat.com. They're really built to help anyone get certified in AI, maybe boost your career.

Plus, you should check out the AI Unravel Builders Toolkit. It's packed with stuff like AI tutorial PDFs, certification guides, even audio and video tutorials to help you actually start building. All the links, they're in the show notes. Okay, so moving beyond the office, beyond the classroom, AI is popping up everywhere, transforming sports, mental health, even our basic understanding of biology. But naturally, this huge expansion also brings up some critical ethical questions, even existential ones. Let's talk surprising applications first, though.

Did you know NBA front offices are using AI pretty heavily now? For scouting, drafting, player development. Like there's this data scientist, Sean Farrell, who presented a model predicting NBA success just based on how players talk in interviews. His model used like 26,000 transcripts, got 63% accuracy with language alone, but jumped to 87% when they added other contexts. And apparently players who spoke in simple present-focused terms tended to do better. Isn't that wild?

That is wild. Language patterns predicting success. Right. And teams are also using LLMs to make sense of complex scouting notes, using AI platforms like Autostats and SkillCorner for analyzing player movement in detail, even health data.

There are tools like Springbok Analytics assessing muscle quality from MRI scans using AI. It's becoming this really comprehensive approach. It really is a holistic view powered by AI. And then there's mental health. Reid Hoffman, the LinkedIn co-founder, he just led a $12 million investment in a company called Sanmai Technologies. They're developing an AI-guided ultrasound helmet for treating conditions like anxiety and depression. The really fascinating part, they're aiming for in-home use at a price under $500.

It's a non-invasive alternative to things like Neuralink, you know, which usually involves actual brain surgery. What's the significance of something like that in mental health care? Oh, that's genuinely striking. I mean, what stands out immediately is how AI is moving past simple data crunching. It's getting into these deeply nuanced areas like human communication patterns in sports and now precise non-invasive medical tools for the brain. For

For mental health specifically, having an affordable in-home non-invasive option could be a real game changer for accessibility, for patient comfort. It just highlights the incredible versatility, the potential impact of AI across these really diverse and sensitive fields. Definitely. And Google's also made this huge leap in genetics.

You have a new AI called Alpha Genome. This thing predicts how single DNA changes variants impact how genes work. It analyzes DNA sequences up to a million letters long. It predicts where genes start and end, how RNA gets processed, even how much RNA is produced. That's groundbreaking for understanding our own biology. Maybe for finding new drugs, too. Absolutely. Fundamental research potential there. And on the AI for good side, psychologists are exploring how AI can make play-based learning better for kids.

Think AI companions that adapt to a child's mood, their progress. Prompting curiosity to help with vocabulary, understanding things better, or even developing this framework called Play Stands for Purpose, Love, Awareness, and Yearning. The goal is to use AI to help kids get into that flow state while learning. It's not just about drilling facts. It's about nurturing engagement, creativity. Though importantly, they do caution against letting AI take too much control over the learning environment.

finding that balance is key. Right. That balance is crucial. And that play framework, it's more than just a teaching technique, isn't it? It's sort of a profound statement about where AI and education could go, recognizing that real learning isn't just information dump. It's about emotional engagement, curiosity. It challenges the idea of AI as just a tutor and pushes towards AI as a facilitator of human potential. All these innovations, they show AI's incredible ability to

augment what humans can do in ways we're probably only just starting to grasp.

from genetics to learning states. But with all this incredible progress, inevitably come the challenges. The critical ethical stuff, the existential questions. We just saw a huge ruling involving Anthropic, the AI company. A U.S. district judge ruled that training their models on books they legally acquired, that's fair use, meaning it's seen as transformative, doesn't just copy the originals, doesn't harm the market for the books. That's a pretty big win for AI companies using data.

However, and this is the kicker, the judge also found Anthropic downloaded millions of books from pirate sites, like Books 3, Library Genesis. And that, the judge said, was copyright infringement. So now they face a trial in December. Potential damages could be up to $150,000 for each pirated book. Ouch. Ouch, indeed. And this ruling, it really gets to the heart of a massive question. How do we balance innovation with intellectual property rights now, in the age of AI?

The judge drawing that clear line, legally acquired data versus pirated sources, that's a really critical precedent, especially with all the copyright lawsuits piling up against AI firms right now. It basically clarifies that, OK, maybe the training itself is transformative, but where you got the data from matters a lot. It's definitely not a free for all for getting training data. This ruling is going to have huge ripples across the industry affecting how everyone sources their data going forward.

No doubt. And finally, we have to touch on this pretty sobering statement from Google CEO Sundar Pichai. He openly acknowledged that the possibility of AI leading to human extinction is pretty high. Though he did add he's optimistic that humans can act collectively to prevent that. Yeah. And hearing that from such a central figure in the AI world,

openly admitting existential risk. It just profoundly underscores the urgency, doesn't it? The need to establish global safety frameworks, robust AI governance, seriously robust. It's a stark reminder that with this kind of incredible power comes just profound responsibility.

The conversation really needs to move beyond just what can it do to focus squarely on safe, ethical development, making sure humanity's long-term future is front and center. So after all that, what does it all mean? I mean, today's deep dive showed us an AI landscape just bursting with innovation, empowering workers, teachers, transforming sports, mental health, even our understanding of genetics. It's incredible. Yeah. But it also throws up these really complex challenges.

Understanding AI's real limits, navigating the legal minefields, grappling with those huge existential questions. It's moving so fast. Staying informed feels more critical than ever. Indeed. And the journey with AI, it isn't just about the tech advancing. It's really about how we as a society understand it,

apply it, and crucially, govern it responsibly. There's always more to learn, isn't there? And hearing multiple perspectives, different viewpoints, that really enriches our understanding, especially with something so transformative. Absolutely. So as you go about your day, maybe think about this. With AI changing so rapidly, how we work, how we learn, even how we evaluate talent, what new skills do you think will become essential for really thriving in this AI-powered future that's unfolding around us?

And hey, as we wrap up, if today's deep dive sparked your curiosity, maybe made you want to go from just listening to actually building with AI. Remember those fantastic resources from Etienne Newman over at djamkes.com. You've got his AI Cert Prep Books, the Azure AI Engineer Associate, Google Cloud Generative AI Leader Certification, all those we mentioned, plus the AI Unraveled Builders Toolkit. That toolkit has tutorial PDFs, guides, audio, video, everything to help you get started building. Seriously, check the link in the show notes.

Thank you so much for joining us on this deep dive. Until next time, keep learning, keep questioning, and definitely stay curious.