Imagine AI not just doing tasks for you, but actually catching cyber attacks live as they happen. Or picture it running its own chemistry lab, you know, making discoveries in just days that used to take years. So today we're digging into a source that, well, it really shows AI is moving much faster than we thought, making real impacts right now. Welcome to the deep dive. Yeah, the source we're looking at today is absolutely
packed with the latest in AI innovation. It's a fascinating look, really, at what's happening on the ground. We're going to cover AI going up against finance heavyweights, some really surprising new uses, and just the incredible speed this is all moving at. And our goal here, as always, is to cut through the noise, pull out the key insights, and we'll leave you feeling like you really get the picture. Okay, let's dive into the first big topic. Anthropix Claude
for financial services. This thing is taking aim directly at the Bloomberg terminal. It is a big move. I mean, the Bloomberg terminal, it's been the standard in finance for decades, right? It costs, what, like $25 ,000 a pop? Exactly. So Claude's coming in as this powerful, potentially disruptive alternative. So what's actually different about it? How does it work? Well, it does the expected thing, pulling in external market data. Yeah. You know, S &P, Faxet, that kind of stuff.
Okay. But the key part is... It then merges that with a company's own internal data, stuff sitting in Databricks or Snowflake, for instance. Ah, so it connects the outside view with the inside view? Precisely. It's like building with these different data Lego blocks, basically. And I saw mention of something called Clawed Code. Yeah, that's important. It lets analysts run their own custom models or simulations right
there. So no more... Fiddling with complex scripts or trying to jury rig different APIs together. Right. No more Frankensteining your APIs, as they put it. It makes that whole process much smoother. Okay. And they've boosted the token and query limits, which is huge. We think it can handle bigger chunks of information. Exactly. Analysts can feed it massive documents like a whole 10K filing or complex M &A docs without hitting those frustrating token walls where the
AI just stomps. And how does it actually perform? Numbers wise. Pretty impressively, actually. On the finance agent benchmark, it scored 44 .5 percent accuracy. OK. And how does that compare? Well, that puts it right up there neck and neck with OpenAI's best models. Most of the other tools tested were like under 30 percent. Wow. That's a significant difference. It really is. And the impact on businesses seems real. Breedwater apparently saw a 20 percent productivity jump.
Norway's sovereign wealth fund saved. Get this, 213 ,000 hours. That's staggering. And AIG cut its underwriting review time by five times. And we should be clear, these aren't just simple lookups, like what's the stock price? No, this is complex stuff, like multi -step analysis of SEC filings, the kind of thing that, yeah, would
make a junior analyst break a sweat. That integration with internal data, though, it sounds powerful, but does the source mention any... security concerns, giving an AI access to sensitive company info? That's a really good point. The source mainly focuses on the efficiency wins. But yeah, it definitely implies you need really solid security around these systems. Makes sense. And just quickly, it also mentioned perplexity is sort of quietly emerging as another strong tool for analysts.
The Coinbase CEO, Brian Armstrong, he called their stuff a 10x unlock for AI in finance. Okay, so pulling this together. What does this really signal for the future of those traditional finance tools, maybe even how finance teams work? I think it suggests AI is making that high end analysis faster and maybe more accessible to beat. All right, let's shift gears. The next section was called Today in AI. Kind of a rapid fire look at different developments. Yeah, this part got
really interesting. A lot going on simultaneously. On one hand, you see this creative expansion, like Runway's new AI motion caption model. Oh yeah, that sounded cool. It tracks face, hands, body, and then lets creators just drop that motion into any scene. Pretty amazing for visual stuff. Definitely. And then there's that Google business calling agent. Right. The one that calls shops for you. Yeah. It'll phone up a local store, ask if they have something, check the price,
and then just report back. So no more slightly awkward phone calls for basic info. Exactly. It could be a real time saver for those little tedious tasks. And speaking of innovation, China's approach to data centers. Putting them in the ocean. Yeah, isn't that wild? Using the seawater for natural cooling. It's clever. The first phase apparently has enough power, like 198 servers, to train GPT -3 .5 in just one day. That's serious capacity and a really neat solution to the heat
problem. And AWS is also pushing forward, putting another $100 million into their generative AI innovation center. Right, helping more companies build actual AI agents. You know, AI that can handle... complex workflows, not just single tasks. But it's not all smooth sailing and cool tech, is it? There's a flip side mentioned. No, definitely not. A pretty stark finding was that 75 % of S &P 500 companies now list AI as a risk factor. 75 %? That's huge. What kind of risks?
Everything, really. Ethical issues, potential for operational screw -ups, regulatory worries, security vulnerabilities, the whole gamut. And just to hammer that point home, there was another chat GPT outage mentioned, right? Around July 16th. Yep. Global users hit with login problems, blank screens, lost work. It's a good reminder that this tech is still maturing, still has hiccups.
So taking all these different pieces, the creative tools, the efficiency gains, the infrastructure challenges, the risks, how does this paint a picture of where AI is right now? It feels like AI is booming, spreading out fast. Yeah. but also hitting some real growing pains, like dealing with major risks and needing massive infrastructure. Beat. Okay, moving on. The source also touched on more practical applications, specific tools
people are using. Yeah, and it made a point that not every AI idea is about, you know, instant riches. Right. One actual business model they mentioned was using AI to auto -generate videos for popular affiliate products. Kind of clever niche application. And then there was LeVart, described as the first real AI design agent. Yeah, that one sounded interesting. It supposedly uses models like ChatGPT and Runway to build out whole campaigns, websites, branding, all
from just one prompt. That's quite a leap for creative automation, if it works well. For sure. And for the builders out there, the tech folks, there's a guide mentioned for using N8n templates. That's one of those workflow automation tools. Okay, what was the guide about? Basically, how to use their templates without running into common errors, covered risks to watch out for, set of steps, debugging. You know, the practical stuff. And I have to admit, I still wrestle with prompt
drift myself sometimes. Oh, yeah. Where the AI kind of wanders off. Exactly. It starts strong, then veers off course from what you originally asked. So guides like that, super helpful, honestly. So looking at these examples, automated videos, AI design agents, workflow tools, what's the common theme here about how AI is being used practically? It seems AI is really becoming a go -to tool for... creative generation and for automating some pretty complex workflows. Now
this next section, this really grabbed me. AI progress might be moving much faster than expected. Yeah, this part was eye -opening. The thinking was, you know, 2025 would be the year AI agents just started handling repetitive tasks well. Right, basic automation. But what we're seeing now is AI making actual discoveries, catching live threats. It almost feels like we've jumped ahead, maybe skipped a step in how these intelligent systems evolve. The source gave specific examples,
right? Like Google's big sleep AI agent. Yes. That one apparently spotted and stopped a live software exploit before it could be used maliciously. Before it was even deployed in the wild. That's what Google claims. The first real world case, they say, of AI intercepting an active cyber threat like that. That's a pretty big milestone. Huge. And then there's that self -driving chemistry lab at NC State. The self -driving lab? Yeah, an AI -powered lab that runs its own chemical
experiments. It collects something like 10 times more data than human researchers. Okay. And the potential payoff. It could slash the time it takes to discover new materials from years, literally years down to days. Wow. Whoa. Imagine scaling that. A billion queries running experiments. That's just incredible potential. And there's data backing up this feeling of acceleration.
Mm -hmm. Data from NTR apparently shows AI is doubling the length and complexity of tasks it can manage across nine different benchmarks roughly every 4 .5 months now. Every four and a half months. How does that compare to before? Just six months ago, that doubling rate was about seven months. So the pace itself is picking up speed. So it's accelerating faster. Exactly. Which leads to that question they posed. Are we maybe moving into a post -agentic AI phase?
And just to clarify that term post -agentic AI. It basically means AI moving beyond just following instructions, right? Yeah. Starting to autonomously discover things. Yeah, precisely. Not just executing tasks we define, but finding new solutions, new capabilities on its own, moving beyond just being an agent doing our bidding. So if this acceleration is real, this almost exponential improvement, what could that mean for innovation, for solving
big problems? It really suggests we might be right on the edge of AI fundamentally changing how discovery happens, compressing what took. research years into maybe just AI weeks in fields like science, software, security. Two sec silence. Mid -roll sponsor read. Okay, let's try to pull this all together. Reflecting on everything we've covered from the source today, what's the big picture? Well, it paints a picture of incredibly rapid AI innovation, doesn't it? Things are moving
fast. Yeah. From seeing AI challenge huge established players like Bloomberg and finance. To actually stopping live cyber threats and speeding up fundamental scientific research. Feels like AI is shifting gears. Shifting from just automating the simple stuff. To being a genuine force for like fundamental change. It's not just about efficiency anymore. And we're seeing these really powerful new tools pop up, design agents, better financial analysis.
But at the same time, there are these really crucial conversations happening and needing to happen around the risks, the ethics, the sheer amount of infrastructure required. Right. It's not just capability. It's also about control, safety and access. So the main takeaway, I think, is crystal clear. AI development isn't just fast, it's accelerating. And it's changing what's possible fundamentally right now, today. That acceleration point really stands out, which leads to a question
for you listening. Given this rapid, maybe even accelerating pace, how do you see AI's progress impacting your own field or maybe just your day -to -day life, even in the next few months? Yeah, that's something worth thinking about. What part of this deep dive resonated most with you? What caught your attention? Definitely something to reflect on. Thank you for joining us on The Deep Dive. We really hope this quick tour through the latest AI happenings gave you some aha moments,
maybe some things to chew on. Until next time, EOTROMusic.