Apple May Ditch Google Search, Netflix Adds AI Search, and Mistral Launches Medium 3

Summary

AI Deep Dive: Episode Summary

Title: Apple May Ditch Google Search, Netflix Adds AI Search, and Mistral Launches Medium 3
Host/Author: Daily Deep Dives
Release Date: May 8, 2025

---

Introduction

In this episode of the AI Deep Dive Podcast, hosts A and B navigate through the latest advancements and strategic moves in the artificial intelligence landscape. From potential seismic shifts in search engine dynamics to groundbreaking AI applications in streaming and robotics, the discussion provides listeners with a comprehensive overview of how AI is rapidly reshaping various industries.

---

Apple Considers Replacing Google Search with AI-Powered Alternatives

The conversation kicks off with major developments surrounding Apple's potential integration of AI search engines into Safari. Hosts A and B delve into the implications of Apple possibly moving away from Google as the default search provider.

A [01:00]: "Ah, yes, and the names involved are the big ones, aren't they? OpenAI, perplexity, anthropic."

This speculation isn't merely hearsay; it gained traction during Eddy Q's testimony in a significant Justice Department lawsuit against Alphabet over Google's default search deal in Safari, valued at a staggering $20 billion annually.

B [01:36]: "He directly linked it to the rise in AI usage. People finding answers differently, I guess."

Eddy Q highlighted that Safari searches have declined recently, attributing this downturn to the increasing reliance on AI for information retrieval. The hosts discuss the possibility of AI search providers potentially overtaking traditional search engines like Google.

A [02:00]: "That's a huge statement."

While Apple isn't making an immediate switch, the ongoing discussions with companies like Perplexity indicate a serious exploration into AI-driven search solutions. This could herald a more conversational and nuanced search experience for users, moving away from keyword-based queries.

---

Netflix Embraces Generative AI for Enhanced Search Functionality

Transitioning to the streaming giant Netflix, the hosts reveal that the company has launched a new AI-powered search tool utilizing OpenAI's ChatGPT. This advancement aims to facilitate a more natural and conversational search experience for users.

A [02:43]: "Ah, yes. Yeah, Netflix. They've launched a new search tool powered by generative AI."

Instead of simple keyword entries like "comedy," users can now ask for more nuanced content, such as "something scary but not too scary, and maybe a little bit funny, but not haha funny." This approach allows the AI to understand and interpret user preferences with greater sophistication.

<blockquote> > **B [03:19]:** "So testing the waters." </blockquote>

The rollout begins with an opt-in beta for iOS users, following initial tests in Australia and New Zealand. While other streaming platforms like Amazon and Tubi have experimented with similar AI search features, Netflix's cautious yet strategic implementation suggests a dedication to refining the technology for optimal user engagement.

B [03:41]: "True. Adoption is key."

Additionally, Netflix is exploring other AI applications, including personalized title cards and multilingual support, aiming to enhance the overall user experience through subtle yet impactful AI-driven personalization.

---

Mistral Launches Medium 3: A Cost-Effective AI Model with Robust Capabilities

The discussion then shifts to Mistral, a French AI startup that has been making significant strides since its inception in 2023. The hosts highlight the launch of Mistral Medium 3, a new AI model that emphasizes efficiency and cost-effectiveness without compromising performance.

A [04:09]: "And the claims are pretty strong, focusing on efficiency."

Mistral Medium 3 reportedly matches or exceeds 90% of the performance of Anthropic's Claude Sonnet 3.7, while being 90% less expensive. It also outperforms other recent open models like Meta's Llama 4 Maverick and Coheres Command A.

B [04:25]: "Coheres command A. Yeah, that's the claim."

Priced at 0.4 cents per million input tokens and 2 cents per million dollar output tokens, Medium 3 offers a cost-effective solution for businesses seeking powerful AI capabilities. Its flexibility allows deployment across cloud environments and self-hosted setups, catering to diverse operational needs.

Mistral has secured substantial funding, exceeding $1.1 billion, and boasts a clientele that includes major corporations like BNP Paribas and AXA. The company is also developing services such as Lechat Enterprise, a customizable chatbot solution designed for business integration.

B [06:02]: "Lets businesses customize it more, connect it to things like Gmail, Google Drive, SharePoint."

This model is currently in beta testing across sectors like finance, energy, and healthcare, where it's being utilized for customer service, workflow automation, and complex data analysis. With its wide availability through platforms like Amazon SageMaker and upcoming integrations with Microsoft Azure and Google's Vertex AI, Medium 3 is positioned to become a widely adopted AI solution.

---

Amazon Introduces Vulcan: A Tactile Warehouse Robot Enhanced by AI

The final major topic covers Amazon's latest foray into robotics with the introduction of Vulcan, a warehouse robot equipped with a sense of touch. Hosts A and B explore how this innovation represents a significant advancement in robotic interaction and efficiency.

B [06:44]: "That's the one. It's a big step for robotics. It has force sensors, so it can actually feel when it makes contact with something."

Vulcan features force sensors that enable it to physically sense and interact with objects more delicately and accurately than purely vision-based systems. The robot comprises two arms: one for rearranging items within storage bins and another equipped with a camera and suction cup for item retrieval.

A [07:09]: "Force feedback, touch feedback. It basically learned by doing, by feeling."

Trained on extensive physical interaction data, Vulcan can handle approximately 75% of Amazon's stock, demonstrating its capability to manage diverse items efficiently. It is currently deployed in Spokane, Washington, and Hamburg, Germany, where it has already processed half a million orders.

B [07:37]: "So it's not just experimental?"

Amazon maintains that robots like Vulcan are designed to enhance safety and efficiency rather than replace human workers, aiming to create a collaborative environment where robots and humans work together seamlessly.

---

Conclusion: AI's Expanding Footprint Across Industries

The episode wraps up with a reflection on how AI is transitioning from theoretical constructs to tangible tools that influence everyday activities. From search engine dynamics and streaming experiences to cost-effective AI models and advanced robotics, AI technologies are deeply embedded in various facets of modern life.

A [08:30]: "They're all different areas, but they paint a picture, don't they? AI is moving fast on many fronts."

The hosts pose thought-provoking questions about the future trajectory of AI, such as how AI search will evolve and the potential for robotics to transform multiple industries. The rapid advancements suggest that features once confined to science fiction are becoming increasingly realistic and integral to our daily lives.

B [08:57]: "Exactly. It leaves you with that question, maybe with AI moving this fast, well, what other sci fi tech is actually way closer to being just normal than we think?"

---

This episode of AI Deep Dive underscores the swift and multifaceted integration of AI technologies across diverse sectors, highlighting both the opportunities and challenges that lie ahead as AI continues to evolve and expand its influence.

Transcript

[A]

It's almost every day now, isn't it? Some huge AI development pops up. Yeah. And just trying to keep track feels like, well, a whole job in itself.

[B]

Totally. It's why we do this. Right?

[A]

Exactly. Welcome to the deep dive. We sift through all that noise, pull out the really important stuff, the things you actually need to know.

[B]

Think of it as your filter. We're looking past the headlines to what it means, you know, for you. A shortcut to staying informed without getting buried.

[A]

And today we're focusing on a whole batch of recent AI news. We've looked at a roundup of top stories to give you a, well, a concise overview of what's really happening.

[B]

Right. The goal is pretty simple. You walk away understanding the key moves without feeling overwhelmed by, like, all the details.

[A]

Okay, let's dive in. Where should we start?

[B]

Let's kick things off with Apple. There's talk, pretty serious talk about them possibly adding AI search engines into Safari.

[A]

Ah, yes, and the names involved are the big ones, aren't they? OpenAI, perplexity, anthropic.

[B]

Those are the ones mentioned.

[A]

And this isn't just rumor, is it? It actually came up in Eddy Q's testimony, you know, for that big Justice Department lawsuit against Alphabet.

[B]

Exactly. The lawsuit over the Google default search deal in Safari. Which is what, like $20 billion a year?

[A]

A staggering amount. 20 billion. It really shows how valuable that default spot is.

[B]

And right in the middle of that testimony, Q drops this observation. Safari searches, they've actually declined.

[A]

Recently declined. Wow. And he thinks it's because of AI.

[B]

He directly linked it to the rise in AI usage. People finding answers differently, I guess.

[A]

Well, that makes a certain kind of sense, doesn't it? If an AI just gives you the answer.

[B]

Scroll through pages of links. Right. And Q even suggested these AI search providers, they might eventually just take over, become the main way people look for info online, maybe even replace Google in that sense.

[A]

That's a huge statement.

[B]

Replacing Google, it's definitely looking ahead. But Apple isn't flipping a switch tomorrow. They seem to think AI search still needs, well, improvement.

[A]

Right. It's not quite ready for that default spot yet.

[B]

Not yet. Needs refinement.

[A]

But they are talking seriously with perplexity, according to reports. So the exploration is real?

[B]

Oh, yeah. The talks seem pretty concrete.

[A]

So for you listening, this could mean the way you search on your phone or computer might get a lot more conversational down the line. Less about keywords, maybe.

[B]

Could be a different kind of interaction entirely. Okay, so from searching the web. Oh, how about searching for what? To watch Netflix.

[A]

Ah, yes. Yeah, Netflix. They've launched a new search tool powered by generative AI.

[B]

Right. Using OpenAI's ChatGPT basically to let you search in a more natural conversational way.

[A]

Exactly. So instead of just typing, say, comedy, you could ask for, I don't know, something funny and upbeat.

[B]

Or get really specific. Like that example. Something scary but not too scary. And maybe a little bit funny, but not haha funny.

[A]

Yeah, exactly. It's about understanding that kind of nuance.

[B]

Which is pretty cool. And how are they rolling this out? It's starting as an opt in beta. IOS users first this week. They did test it before in Australia and New Zealand, apparently.

[A]

So testing the waters.

[B]

Cautious, probably smart. Other streamers have tried AI search. You know, Amazon has the voice search on Fire tv. Tubi had a chatgpt search for a bit.

[A]

Oh, Tubi had one. I didn't realize.

[B]

Yeah, but they stopped it. So it's not a guaranteed hit, this kind of feature. Will people actually use it? That's the question.

[A]

True. Adoption is key. And Netflix is looking at other AI stuff too, right? Like title cards and different languages.

[B]

Yeah, using generative AI for that too. So you might see more subtle personalization creeping in for you at home. Maybe finding that next binge watch gets a bit more intuitive.

[A]

Interesting. Okay, let's shift gears. Let's talk about the AI models themselves, the engines driving all this. There's a French startup, Mistral.

[B]

Mistral. Yeah. Making waves. They've just put out their latest model, Mistral Medium 3.

[A]

And the claims are pretty strong, focusing on efficiency.

[B]

Right. They're saying its performance is at or above 90% of Anthropic's Claude Sonnet 3.7.

[A]

Which costs more 90% of Sonet 3.7. And better than some recent open models too, like Meta's Llama 4 Maverick and.

[B]

Coheres command A. Yeah, that's the claim. Benchmarks are always tricky, but it points to strong performance for the cost.

[A]

Okay, and speaking of cost, how are they pricing it?

[B]

It's priced per token, 0.4 cents per million input tokens, 2 per million dollar output tokens. And a million tokens is what, about 750,000 words? Gives you a rough idea.

[A]

Okay. And it's flexible, can run in different places.

[B]

Seems like it. Cloud environments, even self hosted setups if you've got like four GPUs or more. They're aiming for that cost effective but powerful niche.

[A]

And Mistral itself? They only started in 2023. Right. But they've got serious funding over 1.1 billion.

[B]

Yeah, they're a frontier model lab. Big ambitions. They already have customers like bnp, Paribas, axa.

[A]

Wow. Major companies.

[B]

And they're building services like their chatbot, lachette.

[A]

So what's Medium 3 good at?

[B]

They're highlighting coding, STEM tasks and multimodal understanding. So handling text and images, that kind of thing.

[A]

Yeah. It's being tested already?

[B]

Yeah, Beta testing in finance, energy, healthcare. Using it for customer service, automating workflows, analyzing complex data.

[A]

Okay, so real world applications are already happening. That's significant. And where can people access it?

[B]

It's on Mistral's own API. Amazon SageMaker now, and they say it's coming soon to Microsoft Azure, AI Foundry and Google's Vertex AI.

[A]

So pretty wide availability eventually. And didn't they just release another modder recently? Mistral Small.

[B]

Yeah, Mistral Small 3.1 came out in March and they're already teasing an even bigger model is on the way. Plus, they launched lechat Enterprise.

[A]

Right, the business version of their chatbot. With, like an agent builder.

[B]

Yeah. Lets businesses customize it more, connect it to things like gmail, Google Drive, SharePoint. It just became generally available after a private preview.

[A]

Okay. Lots happening there.

[B]

One more thing on that. Lechat Enterprise will support MCP soon. That's anthropic standard for connecting AI assistants to data systems. Other big players are adopting it too.

[A]

So more interoperability, maybe. Interesting for you. Listening. This just means these powerful AI tools are getting better, cheaper, and easier to access. Likely sparking more innovation everywhere.

[B]

Definitely. Okay, one last piece of news. Moving from the purely digital to the physical robots.

[A]

Amazon's new warehouse robot, Vulcan. The one with a sense of touch.

[B]

That's the one. It's a big step for robotics. It has force sensors, so it can actually feel when it makes contact with something.

[A]

Feel. Wow. How does that work in practice?

[B]

It gives it a more nuanced way to interact with items compared to just using cameras.

[A]

Two arms, right? One moves stuff around.

[B]

Yeah. Rearranges things in the storage bins. And the other arm has a camera and a suction cup for grabbing items.

[A]

And it learned this?

[B]

Yeah.

[A]

Trained on physical data.

[B]

Yeah.

[A]

Force feedback, touch feedback. It basically learned by doing, by feeling. Amazon says it can handle about 75% of their stock now.

[B]

75% is a lot. And it gets better on its own, Apparently.

[A]

Yeah. Self improving, which is pretty key for robots dealing with all sorts of different items in a warehouse. Where are they using it now?

[B]

It's deployed in Spokane, Washington. And Hamburg, Germany. And they say it's already processed half a million orders between those sites.

[A]

Half a million. So it's not just experimental?

[B]

No, seems like it's doing real work. It's the latest in Amazon's, you know, huge fleet of warehouse robots.

[A]

Of course, when you talk robots in warehouses, people worry about jobs.

[B]

Always comes up. Amazon's line is that these robots, like Vulcan, they enhance safety, make things more efficient, but aren't primarily about replacing people.

[A]

Right, that's the official stance. But for you listening, it's another example of how AI is giving robots increasingly sophisticated ways to, well, interact with the physical world. Huge implications beyond just warehouses, really.

[B]

Definitely. So just to wrap up this deep dive, a quick summary. We saw Apple looking hard at AI, changing the search game.

[A]

Netflix trying out AI for finding shows.

[B]

Is Stroll pushing out a really potent but efficient new AI model.

[A]

And Amazon building robots that can essentially feel, Right.

[B]

They're all different areas, but they paint a picture, don't they? AI is moving fast on many fronts.

[A]

It really is. It's less theoretical now, more embedded, affecting how we find information, how we watch stuff, even how goods get moved around. It's touching everything.

[B]

Which makes you think, right? How is AI search going to change your day to day online? What new tricks will these AI models learn next month, next year?

[A]

Yeah. And how far will robotics go in transforming industries?

[B]

Exactly. It leaves you with that question, maybe with AI moving this fast, well, what other sci fi tech is actually way closer to being just normal than we think?