Okay, imagine this. An AI. Super smart. Smart enough to win a math Olympiad. Yeah. But it also internally triggers warnings. Yeah. For potentially helping make, you know, bioweapons, chem stuff, maybe even nuclear weapons. Yeah, exactly. That's Google's new DeepThink. And a version of it is kind of right there on your phone. Wild stuff.
Welcome to the Deep Dive. Today, yeah, we're going to unpack a whole stack of insights from AI revolution, Google's deep think and industry shifts and related AI news that's been popping up. Our mission really is just to cut through all the noise, find the surprising bits, the important facts, and help you get a real handle on where AI is heading quickly, clearly. We'll look at Google's new AI brain, this DeepThink, how other companies are racing like mad to build
even smarter systems. And definitely what all this rapid change might mean for, you know, your job, your future. It's going to be a proper deep dive. All right, let's start with that first big piece, Google's deep think. They didn't exactly shout about this launch, did they? No, it was kind of quiet. But make no mistake, this isn't just like another chatbot update. Right. This is a version, a variant of the AI model that actually got a gold medal at this year's International
Math Olympiad, the IMO. Yeah, think about that. The absolute smartest high school math kids on the planet. And an AI comes along and, well, wins gold. And a version of that same AI. Yeah. It performs at a bronze medal level, but it's right inside the Gemini app. Yep, right there. Better than most high school math geniuses. Running on your phone, it's accessible power. Something different. What I find really interesting is how it actually solves problems. Google calls
it a... parallel thinking techniques. Instead of just going down one path step by step. It like explores loads of different ways to tackle the problem all at once. Exactly. Like a super fast, really efficient brainstorm, trying out all these angles super quick, then locking in the best one. But OK, let's manage expectations. It's not about to replace a research physicist just yet. Softly. No, no, not quite yet. That full gold medal power that's still very much
kept in the lab under wraps. So the version in Gemini is constrained. Right. It's powerful for what it is, accessible, but it's not the full unlocked potential. Oh, and yeah, the access. Ah, the price tag. Yep. Only for AI Ultra subscribers right now. And that costs, wait for it, about $250 a month. Wow. Okay. Yeah. Really? Kind of shows you who they think the early adopters are, doesn't it? Or where the initial value lies.
So if we boil it down, what's the main thing to understand about DeepThink's power level that we can actually access right now? It's Olympia -level math smarts, accessible on your phone, but definitely not its full research power. Okay, now this next part. This is where it gets really interesting and maybe a little concerning. Mm -hmm. The internal red flags at Google. Yeah. This AI deep think it actually triggered their
own internal safety alerts. Right. Google themselves said it might have reached a, quote, critical capability level. That's their term. Yeah, their term. Yeah. For potentially helping bad actors with, like you said, biochem or even nuclear weapons knowledge. OK, so they're not saying it can do that. No, not that it crossed the line, but it got close enough that, boom, they added extra monitoring, tighter safety rules. Right. Two sec silence. Whoa. I mean, just imagine scaling
that. An AI that capable, potentially doing a billion queries a day. Yeah, that raises huge questions about how fast this is all moving and the responsibility that comes with it. And it's not like Google's alone in this. This isn't happening in some, you know, quiet corner. No, it's a full -on sprint out there. ByteDance, they just launched SeedProver. That's a silver medal IMO math model. And the key difference. It's open source. The
code is out there. Big deal. Meanwhile, OpenAI, they're expected to drop their own gold medal math AI pretty soon. But this time, Google got there first. They shipped. It's at the new bar. It is absolutely a sprint. Everyone wants that title, right? The smartest AI. The bragging rights are huge. But so are the risks. And it's not just math, is it? There's so much else happening. Google, DeepMind and Kaggle, they've launched this Kaggle game arena. Oh, yeah. What's that?
It's basically where AI models, GPT, Claude, Gemini. Even Grok will compete against each other in strategy games. Kicks off August 5th. Huh. Like an AI Olympics for strategy. That's actually pretty cool. Yeah. A way to benchmark how they handle complex changing situations beyond just, you know, text. Pushing them in different ways. But, you know, it's not all just impressive demos and competitions. There's friction, too. Perplexity AI. Ah, the search engine thing. Right. They're
being accused of some. Let's call it sneaky scraping. Making millions of requests daily, even from sites that specifically block bots like theirs. And Cloudflare called them out? Cloudflare delisted them. But Perplexity is apparently just refusing to accept that action. Sticking to their guns. Wow. Yeah, that's quite a stance, isn't it? It really highlights that tension. AI needs data. But what about website rules? Ethics .ip. Big questions there. And on the creative front, XAI.
Oh yeah, Grok Imagine. Dropped officially. Turns simple text or even an image into a 15 second video. With audio. Just like that. Text to video. Pretty much. Imagine whipping up quick explainers or social clips. The speed of progress in generative AI is just wild. From words to a mini movie. And then there's the classic tech rivalry. Google took a direct shot at Apple. Oh! How so? With a Pixel 10 ad. Yeah. Basically mocking the fact that Apple's big Siri AI update is still, you
know, coming soon. Ah, playing the we have it now card. Yep. The ad pretty much says, if Apple's still coming soon, maybe you should just change your phone. Ouch. Yeah. Classic competitive spirit. The gloves are definitely off. So, okay. We've got these super smart AIs, rapid progress, intense competition, internal safety flags, data scraping rows. When AI gets this smart this fast. What's the core challenge everyone's grappling with?
It's balancing that incredibly fast innovation push with critical safety needs and figuring out the ethical lines as you go. Right. Let's maybe shift gears slightly to how AI is actually becoming more practical, more accessible day to day. They're known for quantum computing, right? They just launched an AI toolkit. It's already usable. Quantum AI, that's pretty cutting edge. Yeah, and they have a working demo already, generating images using quantum principles somehow.
They're going to present it all at the AI Research Summit next year. Wow. Connecting quantum research to something tangible like image generation, that's a big step. Using quantum mechanics for faster AI calculations. Neat. And the money flowing in. It's huge. Observe Inc., for instance. They just raised $156 million. For what exactly? For their AI observability platform. Basically AI watching other AI systems. Ah, okay. Like quality control for AI? Sort of. Helps companies see
how their AI is doing, fix problems. And get this, they're processing over 150 petabytes of data every month with help from OpenAI. 150 petabytes. That's just... astronomical, shows you the scale these systems operate at. And yeah, observability is crucial if you're deploying complex AI. You've got to know what it's doing. And it's not just big companies and huge investments. We're seeing guides pop up for building no -code AI. Meaning you don't have to be a programmer? Exactly. Like
there's a guide on building a rack system. Using tools like NAN and Supabase, anyone can try it. Right. R -drying, retrieval augmented generation. That's a really useful technique. Explain that. It's basically where the AI doesn't just rely on its general training data. It combines that with specific documents you give it. So like you could ask it about your company's vacation policy and it pulls the answer directly from your actual policy document. Much better answers.
Very practical. We're also seeing specific business strategies emerge like this value first AI lead strategy. Fiverr and Upwork are apparently using it. How does that work? Seems like using AI to find and connect with customers who already know exactly what service they need. High intent, kind of pre -sold leads. Smart. Using AI for better targeting. And if you're thinking about skills, there's even a guide out there. Seven AI skills to make you rich in 2025. Covers AI
design, development strategy. It works. Yeah, the message there seems to be. Don't just learn about AI. Learn how to use AI to solve actual business problems. That's where the value and the money is. Think of AI as your super -powered assistant. And the tools themselves. They're getting so specific and powerful. SciSpace AI automates research tasks for academics, like 150 different tools. Wow. Furbyte helps writers get publish -ready content faster. Tregno builds
entire SEO strategies. Gulpo creates AI videos like explainers. It's amazing how specialized they're becoming, each one tackling a really specific pain point, automating stuff that used to take humans ages. Right. It means you can potentially do more, faster, maybe even better quality in some cases. Absolutely. More leverage. So looking at all this practical stuff, the tools, the investment, the no -code movement, what's the big takeaway? How is AI evolving for... you
know, us regular users or businesses. I'd say AI is becoming way more specialized, definitely more accessible, and it's getting woven really deep into workflows everywhere. Okay. Now for the topic that probably touches everyone listening, jobs. The big one. We looked at these charts, fastest shrinking jobs versus safest jobs from AI 2025, 2030. And honestly. The picture is pretty stark. Yeah, it's clear. If your job is mostly repetitive tasks, administrative stuff, it's
at risk. High risk. But if it involves a lot of human interaction, complex problem solving in the real world, emotional intelligence, it seems much safer. For now, at least. Definitely. Anything clerical, just moving data around, stuff tied to old systems that's being automated fast. Bots, self -checkouts, predictive software, they're taking over those roles. It's a fundamental shift. But like you said, it's not all doom and gloom. Not every job is on the chopping block. No, thankfully
not. Jobs needing real -time judgment, empathy, creativity, fixing physical things. Think therapists, electricians, artists, good manager. They're still very much needed. Right. Those uniquely human skills. But, and it's a big but. Don't get too comfortable thinking AI can't handle complexity. Remember the McDonald's AI drive -thru fail? Kept adding bacon to everything. Slightly amused. I do remember that. Hilarious. Or, less hilariously, Amazon's Just Walk Out
tech, the cashierless stores. Turns out, behind the scenes, there were reportedly like a thousand people in India watching feeds to make sure the charges were right. So a lot of the shiny AI automation we see still have a whole lot of humans behind the curtain, kind of like the Wizard of Oz. Pretty much. It's not always as automated as it looks. It reminds me, you know, I still wrestle sometimes with prompt drift myself. Yeah.
When you're having a long chat with an AI and it kind of forgets the original point or instruction. Oh, yeah. It makes me wonder how these more complex jobs, the ones needing sustained focus and context, how they'll really evolve as AI gets better, but maybe still imperfect for everyone. Yeah, it's basically sorting jobs into two piles, isn't it? The easy to automate pile and the hard to automate pile. And you definitely want to be in the second pile. Absolutely. Which means doubling
down on those human skills. Creativity, critical thinking, collaboration, empathy, adapting is key. And doing it quickly. So bottom line for people listening, how should we really think about our careers, our skills in this this AI driven world that's changing so fast? Foken hard on those uniquely human, non -repetitive skills and be ready, really ready to adapt and learn how to work with these new AI tools. Sponsor. OK, that was a lot. So let's try to pull the
big threads together. What does all this really mean? We've seen AI pushing boundaries like crazy. Google's DeepThink is a prime example, incredible power, Olympiad level smarts, but also these serious safety questions popping up internally. Yeah. And the race to build the next big AI model. It's incredibly intense, super dynamic, super competitive. New stuff is landing constantly, almost redefining state of the art week by week. We also saw how AI isn't just lab experiments
anymore. It's becoming really practical, specialized. New tools, new strategies are changing how businesses work, how individuals create and research. More accessible, more integrated. And crucially, all of this is shaking up the job market big time. It's speeding up that shift away from repetitive tasks towards roles that really need those human skills, the ones AI can't easily replicate yet. So this whole deep dive today really shows technology
is moving at lightning speed. But it's forcing us, all of us, to wrestle with some really complex ethical questions, rethink data, privacy, and fundamentally rethink our relationship with work itself. So turning it over to you listening, what stood out most to you from all this? How do you see these AI shifts impacting your world, your job, even just, you know, your day to day
life? Yeah, we really encourage you to maybe dig a bit deeper into some of these areas because the pace of change is, well, it's kind of astonishing. It really is. Thank you for joining us for this deep dive. Until next time, keep digging into what's truly interesting. Otiro Music.