OpenAI Eyes Ads, Tenstorrent Challenges Nvidia, and AI-Generated 3D Scenes

Summary

AI Deep Dive Podcast Summary
Hosted by: Daily Deep Dives
Episode: OpenAI Eyes Ads, Tenstorrent Challenges Nvidia, and AI-Generated 3D Scenes
Release Date: December 3, 2024

---

Introduction

In this episode of AI Deep Dive, hosts A and B explore the latest advancements and strategic moves shaping the artificial intelligence landscape. From OpenAI's potential foray into advertising within ChatGPT to groundbreaking technologies from World Labs and Cohere, and significant investments challenging industry giants like Nvidia, the discussion provides a comprehensive overview of current AI trends and their implications across various sectors.

---

OpenAI's Potential Introduction of Ads in ChatGPT

The conversation begins with OpenAI contemplating the integration of advertisements into ChatGPT. Hosts delve into the financial dynamics and strategic considerations behind this move.

• A (00:19): "It looks like OpenAI is thinking about ads in ChatGPT."

• B (02:03): "She's mentioned some downsides about becoming too reliant on ads, maybe even prioritizing advertisers over the users." Referring to OpenAI's CFO, Sarah Friar.

Key Points:

• Revenue vs. Research: OpenAI currently generates around $4 billion annually from ChatGPT, with 250 million weekly users. However, their development costs exceed $5 billion annually (A, 01:07), highlighting a classic dilemma between profitability and advancing research.

• Leadership Perspectives: CEO Sam Altman appears supportive of introducing ads, aiming to balance profit generation with sustaining research and development. In contrast, CFO Sarah Friar expresses caution, concerned about the potential negative impact on user experience and the platform's revolutionary status.

• Future Implications: The hosts ponder whether incorporating ads could transition ChatGPT from a cutting-edge tool to a conventional platform, potentially diminishing its "cool factor" (A, 01:58). Friar maintains that ads are not dismissed but are just one of several revenue streams being considered (B, 02:14).

---

World Labs' AI-Generated 3D Scenes Technology

Shifting focus, the hosts discuss World Labs' innovative technology that transforms a single image into an interactive 3D scene, revolutionizing digital content creation.

• A (02:39): "World Labs, they made this tech that can create interactive 3D scenes from a single image."

• B (04:17): "So it's like having a research assistant that just gets you."

Key Points:

• Technology Overview: Founded by AI pioneer Fei Fei Lai, World Labs has developed a method to convert any photograph into a fully navigable 3D environment. Users can explore these scenes from various angles, simulating depth of field and interactivity (A, 03:26).

• Applications: Potential uses extend across multiple industries, including video game development, virtual tours, architectural walkthroughs, and educational simulations, such as exploring historical sites like ancient Rome (B, 03:47; 04:03).

• Current Limitations and Future Goals: While the technology is impressive, it still faces challenges in generating large-scale scenes and eliminating rendering glitches (B, 04:19). World Labs aims to launch a commercial product by 2025, targeting a broad audience from developers to movie studios.

• Notable Quote: Justin Johnson, co-founder of World Labs, encapsulates their vision: "World models will let you not just get an image or a clip out, but a fully simulated, vibrant and interactive 3D world." (A, 04:52).

---

Cohere’s Rerank 3.5 AI Search Model

Next, the discussion moves to Cohere's latest AI-powered search model, Rerank 3.5, designed to enhance information retrieval across diverse data types.

• A (05:34): "Cohere and their new AI search model, rerank 3.5."

• B (07:16): "And it can handle both at the same time, no problem."

Key Points:

• Enhanced Search Capabilities: Unlike traditional search engines that rely on keyword matching, Rerank 3.5 leverages advanced AI to understand the intent behind queries, ensuring more accurate and relevant results (B, 06:12).

• Versatility: The model isn't limited to text; it can efficiently search through images, videos, and even code, making it highly useful for professionals dealing with large and varied datasets (A, 06:42).

• Multilingual Support: Rerank 3.5 supports over 100 languages, enabling seamless information retrieval for users worldwide, regardless of their language preferences (B, 07:05).

• Applications: One prominent use case is Retrieval Augmented Generation (RAG), where the model assists in generating more insightful and accurate responses by integrating real-time data (A, 07:27).

• Endorsement: Simon Lass, co-founder of Notion, endorses Rerank as essential for their AI features, highlighting its significance in information management (A, 07:55).

• Availability: Rerank 3.5 is currently accessible on Cohere's platform and major cloud services like Amazon Bedrock and SageMaker, with plans for expansion to other platforms soon (B, 08:05).

---

10storant Challenges Nvidia in the AI Chip Market with Bezos' Investment

The episode concludes with a discussion on 10storant, an AI chip startup receiving substantial investment from Jeff Bezos, aiming to disrupt the dominant position of Nvidia in the AI hardware sector.

• A (08:49): "Did you hear about Jeff Bezos' latest investment?"

• B (10:43): "He's the founder, the driving force. He's known for his innovation."

Key Points:

• Investment Details: Jeff Bezos, alongside Samsung, AFW Partners, LG Electronics, Fidelity, and others, invested $700 million into 10storant, valuing the startup at $2.6 billion (A, 08:56; 09:04).

• Strategic Positioning: 10storant aims to compete with Nvidia by focusing on open-source and off-the-shelf technologies, particularly leveraging the open standard RISC-V architecture to reduce costs and increase accessibility (B, 09:27).

• Leadership: The company is led by Jim Keller, a renowned chip designer with a storied career at Apple, Tesla, and AMD, bringing innovation and expertise to 10storant (B, 10:46).

• Market Impact: By avoiding direct competition with Nvidia's proprietary technologies and instead pursuing more flexible and affordable solutions, 10storant seeks to carve out a significant niche in the AI chip market. They have already secured nearly $150 million in contracts, signaling strong market confidence (B, 10:26; A, 10:34).

• Future Outlook: With substantial backing and a strategic approach, 10storant is positioned to potentially disrupt Nvidia's dominance, especially if their open and cost-effective models gain widespread adoption. Hosts note the competitive and dynamic nature of the AI chip market, emphasizing the excitement and unpredictability it brings (A, 12:08).

---

Conclusion

The episode of AI Deep Dive provides listeners with an in-depth analysis of significant developments in the AI industry. From OpenAI's strategic revenue considerations to groundbreaking technologies from World Labs and Cohere, and the emerging competition in the AI chip market led by 10storant, the discussion underscores the rapid innovation and competitive dynamics shaping the future of artificial intelligence. Hosts A and B effectively highlight the opportunities and challenges presented by these advancements, offering valuable insights for tech enthusiasts, developers, and industry stakeholders alike.

---

Notable Quotes:

• "World models will let you not just get an image or a clip out, but a fully simulated, vibrant and interactive 3D world." — Justin Johnson, Co-founder of World Labs (04:52)

• "It's like having a research assistant that just gets you." — B (06:37)

• "He's the founder, the driving force. He's known for his innovation." — B (10:43)

---

Stay informed and ahead of the curve by tuning into AI Deep Dive, where cutting-edge AI developments are unpacked daily to keep you connected with the future of technology.

Transcript

[A]

AI is everywhere these days. It seems like every day there's some crazy new thing popping up.

[B]

Yeah, it's really moving fast.

[A]

So today we're going to deep dive into. We got some juicy stuff from AI. Deep dive this article.

[B]

Oh, yeah.

[A]

It looks like OpenAI is thinking about ads in ChatGPT.

[B]

Interesting.

[A]

World Labs dropped this tech that makes 3D scenes from a single image.

[B]

Wow.

[A]

And Cohere's got this new AI search model, rerank 3.5.

[B]

So sounds like we got a lot to cover.

[A]

Yeah, we're going to unpack each of those and see what kind of earthquakes they're causing in the AI world.

[B]

Yeah, these are definitely some big changes. Could really reshape some industries like OpenAI and the ADS. That's a huge change. I mean, they're known for pushing boundaries, innovation, all that.

[A]

Right. And it kind of makes you wonder, is this the beginning of like, I don't know, like a trend or something? Are we going to see all this cutting edge AI become, you know, all about the bottom line? Well, I mean, is there revenue? It's already at like $4 billion annually.

[B]

It's a lot of money.

[A]

It's crazy with, with 250 million people using Chat GPT every week.

[B]

Every week. Wow.

[A]

But they're also burning through over $5 billion a year developing these models.

[B]

So it's like they're trying to find that sweet spot between profit and research.

[A]

The research and development. The profit. Yeah.

[B]

Classic dilemma.

[A]

It is.

[B]

Well, OpenAI CEO Sam Altman, he seems to be warming up to this whole ad idea.

[A]

Oh, really?

[B]

But the cfo, Sarah Friar, she's a bit more cautious.

[A]

Makes sense.

[B]

She's mentioned some downsides about becoming too reliant on ads, maybe even prioritizing advertisers over the users.

[A]

Ooh, that's, that's a sticky one.

[B]

Yeah, that's, that's a tough one.

[A]

So, like, we got this, we got this tension, right, between like pushing AI to its limits and keeping the lights on.

[B]

Yeah.

[A]

You know, and I wonder, will ads make ChatGPT feel less revolutionary?

[B]

It's a good question.

[A]

More like just another platform, you know, trying to sell us something.

[B]

Right. Yeah.

[A]

Like, is it going to lose its cool factor?

[B]

I mean, Friar hasn't completely ruled out ads though.

[A]

Okay.

[B]

She said I don't preclude. But for now, she said there's lots of low hanging fruit in the way we are doing things. So they're, they're definitely thinking of different ways to make money from ChatGPT. Yeah, ads might just Be one part of it.

[A]

That low hanging fruit line. That really gets the wheels turning, doesn't it?

[B]

It does. It makes you wonder, you know, what are they thinking?

[A]

Exactly. Okay, let's switch gears. Let's talk about something totally different.

[B]

Okay.

[A]

It feels like straight out of a sci fi movie. World Labs, they made this tech that can create interactive 3D scenes from a single image.

[B]

Yeah, this is really, really cool technology.

[A]

It's crazy. So World Labs, it was founded by this AI pioneer, Fei Fei Lai.

[B]

That's right.

[A]

They figured out how to turn any image into a 3D world that you can explore.

[B]

Like literally walk around in.

[A]

It's wild. Yeah. So if you have a picture, let's say, of your living room, you could, you could potentially step into a 3D version of it, walk around, look at things from different angles.

[B]

That's wild. Like, almost like stepping into a photograph.

[A]

It's mind blowing. Yeah. And the coolest part, the scenes, they're not just like static 3D models.

[B]

Right.

[A]

They're interactive.

[B]

Oh, wow.

[A]

You can control the camera. You can zoom in, zoom out, even simulate that depth of field effect.

[B]

So it's like you're really there.

[A]

It's like you're really there. Yeah. I saw their demo online and it's. It's so realistic looking. It's incredible.

[B]

Yeah. It's pretty amazing what they've done.

[A]

So what does this mean for things like video games, you know, virtual tours, all that?

[B]

Oh, the potential is really limitless.

[A]

Yeah.

[B]

I mean, imagine architects, they could make 3D walkthroughs of buildings.

[A]

Oh, yeah.

[B]

Before they're even built.

[A]

That would be incredible.

[B]

Or students, they could explore like historical sites.

[A]

Yeah. Like ancient Rome or something.

[B]

Yeah, exactly. In these immersive environments. Wow. The possibilities are really endless.

[A]

It feels like one of those advancements that could totally change how we use digital content.

[B]

Absolutely. It's a game changer for sure.

[A]

But it's got its limitations. Right.

[B]

It's still early days, of course. World Labs has said they're still working on generating like really large scenes.

[A]

Okay.

[B]

And sometimes there are little glitches in the rendering.

[A]

Just like any new tech, it's going to have its kinks.

[B]

Exactly. Yeah. But they're aiming to have like a real commercial product ready by 2025.

[A]

That's not that far off.

[B]

Yeah. So. And they're targeting, you know, everybody from video game developers to movie studios.

[A]

That's a wide range.

[B]

It is. It's a very wide range.

[A]

It's going to be interesting to see, you know, how they keep Refining this and what people do with it.

[B]

Absolutely.

[A]

I love this quote. It's from Justin Johnson, one of the co founders of World Labs. He says, unquote, World models will let you not just get an image or a clip out, but a fully simulated, vibrant and interactive 3D world. That's a pretty big statement.

[B]

Yeah, it is. It's a bold vision and it really highlights their ambition. I mean, they're not messing around. They want to change how we experience digital content, period.

[A]

Yeah, it's a big swing for sure. So we got OpenAI maybe going into the ad world. World labs pushing 3D content creation further than ever before. Now, let's talk about something that might seem a little more, I don't know, practical.

[B]

Okay. Yeah.

[A]

But it's just as revolutionary. And that is Cohere and their new AI search model. Rerank 3.5.

[B]

Oh, yeah, this one is really interesting.

[A]

So Rerank 3.5. It's tackling this problem. We all have it, you know, trying to find the exact information you need.

[B]

Oh, yeah.

[A]

In these mountains of data, it's a.

[B]

Needle in a haystack.

[A]

Sometimes it is. Right. You spend so much time going through search results trying to find that one little nugget of information. So Cohere is trying to make that way more efficient.

[B]

They are. And I think they might be onto something really special here.

[A]

I hope so.

[B]

Yeah.

[A]

What makes it different than all the other search engines? What's the. What's the secret sauce?

[B]

It's not just matching keywords. This thing is built on a really powerful AI model that actually it understands what you're asking. Oh, wow. It can grasp the intent behind your search.

[A]

That's incredible.

[B]

So it's not just finding documents with certain words.

[A]

Okay.

[B]

It's finding the stuff that answers your question.

[A]

Really gets what you're looking for.

[B]

Exactly. Yeah.

[A]

So it's like having a research assistant that just gets you.

[B]

Yeah, it's like a mind reader almost.

[A]

And here's another cool thing.

[B]

Okay.

[A]

Rerank 3.5. It can work with all kinds of data.

[B]

Oh, that's interesting.

[A]

It's not just for text. It can search through images, videos, even code.

[B]

That's. That's huge.

[A]

It's crazy.

[B]

Especially for people who work with, you.

[A]

Know, huge amounts of data, like scientists, researchers, analysts.

[B]

Yeah, it could be a game changer.

[A]

It really could.

[B]

And it gets even better.

[A]

Okay, come on, what else?

[B]

It's multilingual.

[A]

Oh, wow.

[B]

This thing can work across over 100 languages.

[A]

So you could have someone, like, in Japan searching in Japanese and someone in Brazil, searching in Portuguese.

[B]

Exactly.

[A]

And it can handle both at the same time, no problem. That's incredible. And it's not just about finding the information, it's understanding it.

[B]

That's right. It's about extracting meaning from the data.

[A]

So one of the big applications for this is something called retrieval, augmented generation or rag. So that's where AI uses real time data to give you better, more accurate responses.

[B]

Exactly.

[A]

So it's like, it's not just a search engine, it's like a, it's like a research super assistant. It finds the information, understands it, then uses it to help you get insights.

[B]

Even generate creative content.

[A]

Yeah, that's, that's mind blowing.

[B]

It is pretty amazing.

[A]

And so to back this up we've got Simon Lass, he's the co founder of Notion and he said that Coheres re ranker is essential for their AI features.

[B]

Yeah, that's a pretty big endorsement coming.

[A]

From a company like Notion who's all about managing information. That's, that's a big deal.

[B]

It is, yeah. And the good news is rerank 3.5 is available right now on Coyer's platform, but also on Amazon, Bedrock, Amazon, SageMaker, and it's coming to other cloud platforms soon.

[A]

Wow. So it's going to be accessible to just about everybody?

[B]

Pretty much, yeah.

[A]

It's really incredible how AI is changing how we, how we use and interact with information.

[B]

It's transforming everything.

[A]

It is, it's transforming everything from finding a single document in a mountain of data to getting real insights from complex data sets.

[B]

It's a whole new world.

[A]

It's a whole new world. AI is changing the game.

[B]

It really is.

[A]

So speaking of Amazon.

[B]

Yes.

[A]

Did you hear about Jeff Bezos latest investment?

[B]

I did, I did hear about that.

[A]

He's putting some serious money into the future of AI chips.

[B]

Yeah, he's betting big.

[A]

He is. So Bezos along with Samsung, they just put $700 million into 10storant.

[B]

Wow.

[A]

That's a startup that's making AI chips. And get this, that investment values 10storant at 2.6 billion.

[B]

That's, that's a lot of money.

[A]

That's serious confidence. So what's the deal with ten Star? Are they, are they taking on Nvidia, the big dog of AI chips?

[B]

Seems like it, yeah.

[A]

It's a bold move.

[B]

It is, it's a very bold move. But they have a, they have an interesting strategy.

[A]

Okay.

[B]

They're, they're focusing on open source and off the shelf technology.

[A]

Okay.

[B]

To keep their costs down.

[A]

Smart.

[B]

And they're, they're really into riscv.

[A]

Riscv, what's that?

[B]

It's an open standard processor architecture. Some people think it could. It could be a rival to arm.

[A]

Oh, wow. So they're really trying to disrupt the market, huh?

[B]

They are, yeah.

[A]

Building AI chips that are more, More accessible.

[B]

Yeah. And potentially more affordable.

[A]

Yeah, that's, that's a, that's a good approach.

[B]

They're. They're betting that this open and flexible approach will give them an edge.

[A]

Makes sense. You know, Nvidia's built their empire on. On proprietary technology, so.

[B]

Right. They've got their own thing going.

[A]

It's interesting to see how this whole, this whole AI chip landscape is changing.

[B]

It's a really dynamic market right now.

[A]

It is. It's not just about, you know, raw power anymore.

[B]

Right.

[A]

It's about finding the right balance, you know, performance, cost, accessibility.

[B]

Absolutely. It's a multifaceted problem.

[A]

So 10storant, definitely a company to keep an eye on. They've already got almost $150 million in contracts.

[B]

Wow, that's a good start.

[A]

It's a good start.

[B]

Yeah. I mean, it's still. It's still nothing compared to Nvidia.

[A]

Yeah, yeah, Nvidia's a giant, but.

[B]

But it's a good sign they got a secret weapon.

[A]

Oh, who's that?

[B]

Jim Keller.

[A]

Oh, wow.

[B]

The legend himself. He's worked at Apple, Tesla, amd.

[A]

He's a. He's a big deal.

[B]

He is.

[A]

What's he doing with ten Storant?

[B]

Well, he's the founder, the driving force. He's known for his innovation. Yeah, he pushes the boundaries of chip design.

[A]

So he's, he's the brains behind the operation.

[B]

He is. And he's got a clear vision for how ten Storant can win. He said you can't beat Nvidia if you use hbm because Nvidia buys the most HBM and has a cost advantage. But they'll never be able to bring the price down the way HBM is built into their products and their sockets. So they're not trying to go head to head with Nvidia.

[A]

Right. They're trying to outsmart them.

[B]

Yeah, it's like David versus Goliath. It's a classic underdog story and it's.

[A]

Going to be interesting to see how it all plays out. So this funding round, it was led by AFW Partners, Samsung securities, and there were others involved too. Bezos Expeditions, LG Electronics, Fidelity. You know, some big names.

[B]

Yeah, that's a pretty impressive list.

[A]

It is. They're all betting on 10storant to disrupt the AI chip market.

[B]

And, you know, it's a sign that this market is. It's ready for a shakeup. Nvidia's been on top for a long time.

[A]

Yeah.

[B]

But these new competitors, they're coming in with fresh ideas.

[A]

Yeah. And a lot of money behind them.

[B]

That never hurts.

[A]

It never hurts. That's what makes following the AI industry so exciting right now.

[B]

Absolutely.

[A]

There's so much innovation, it's impossible to predict what's going to happen next.

[B]

It's like the Wild west out there.

[A]

It really is. Yeah. Thanks for joining us on this deep dive into the fascinating world of AI. We hope you found it informative and thought provoking. Until next time, stay curious.