Perplexity AI’s TikTok Merger Bid, DeepSeek V3’s Power, & Europe’s Call for AI Independence

Summary

AI Deep Dive Podcast Episode Summary
Title: Perplexity AI’s TikTok Merger Bid, DeepSeek V3’s Power, & Europe’s Call for AI Independence
Host: Daily Deep Dives
Release Date: January 19, 2025

---

Introduction

In the January 19, 2025 episode of the AI Deep Dive podcast, hosts A and B delve into significant developments within the artificial intelligence landscape. They explore a potential merger between Perplexity AI and TikTok, analyze the impressive capabilities of DeepSeek V3, discuss the Federal Trade Commission's (FTC) concerns regarding big tech collaborations, and examine Europe's push for AI independence in response to U.S. export restrictions.

---

1. Perplexity AI’s TikTok Merger Bid

The episode opens with a discussion on the audacious merger proposal between Perplexity AI, renowned for its AI-powered search engine, and TikTok, specifically its U.S. operations.

• Merger Details: Perplexity AI aims to integrate TikTok's video content into its search engine, potentially revolutionizing how users interact with search results.

• Political Implications: The merger comes at a tumultuous time for TikTok, with the U.S. government contemplating a ban. ByteDance, TikTok's parent company, has resisted selling TikTok, making Perplexity's bid a strategic sidestep of the broader regulatory issues.

• Quote Highlights:

  • At [00:53], Host A states, “Perplexity AI, the ones behind that super cool AI powered search engine, they want to merge with TikTok, specifically their US operations.”
  • At [02:01], Host B muses, “Imagine searching for something and instead of just getting links, you get TikToks.”

• Political Drama: The potential merger is further complicated by political uncertainties. President Biden has deferred decisive action to the next administration, potentially under the Trump administration, which may extend the deadline for TikTok's operations in the U.S.

• Quote Highlights:

  • Host A notes at [01:43], “President Biden basically passing the buck to the next administration...”
  • Host B adds at [01:45], “The Trump administration again.”

• Potential Outcomes: The hosts ponder the transformative impact of such a merger, questioning what a "Perplexity powered TikTok" would resemble and whether the merger could navigate the intricate political landscape to fruition.

---

2. DeepSeek V3’s Power

Transitioning from mergers to technological advancements, A and B explore DeepSeek V3, a Chinese AI model making significant strides in various applications.

• Capabilities: DeepSeek V3 excels in coding, translation, and email composition, outperforming prominent AI models from OpenAI and Meta, particularly in coding tasks.

• Performance Metrics:

  • Trained on 14.8 trillion tokens, illustrating extensive data exposure ([02:26]).
  • Boasts 671 billion parameters, which enhances its sophistication and performance ([03:10]).

• Quote Highlights:

  • Host B at [02:34]: “OpenAI Meta. Especially when it comes to coding tasks.”
  • Host A at [03:02]: “It's like every single book in the Library of Congress, but times a thousand.”

• Open Source Impact: As an open-source model, DeepSeek V3 allows developers worldwide to access and utilize its capabilities, fostering broader collaboration and innovation.

• Challenges and Limitations: Despite its prowess, DeepSeek V3 operates under Chinese regulations, limiting its ability to provide unbiased commentary due to programmed alignment with core socialist values.

• Quote Highlights:

  • Host B at [03:55]: “it's programmed to align with core socialist values...”

• Developer Insights: The model is developed by High Flyer Capital Management, a hedge fund leveraging AI for quantitative trading. Founder Liang Wenfeng champions open-source models as a means to challenge the dominance of closed models like those from OpenAI.

• Quote Highlights:

  • Host B at [04:27]: “...He believes that open source models like Deep Sea are the way to do it.”

• Infrastructure Investment: High Flyer has invested in substantial server infrastructure, including 10,000 Nvidia A100 GPUs, to support the development of what they term "super intelligent AI."

• Quote Highlights:

  • Host A at [05:07]: “10,000. Wow.”

• Open vs. Closed Source Debate: The hosts discuss the ideological clash between open-source transparency and the controlled development of closed models, highlighting the merits and drawbacks of each approach.

---

3. FTC's Concerns Over Big Tech Partnerships

A significant portion of the episode addresses the Federal Trade Commission's (FTC) apprehensions regarding collaborations between major tech companies and AI developers.

• Lock-In Risks: The FTC is worried that partnerships between big tech giants and AI developers could create a "lock-in" scenario, where businesses become overly dependent on a single AI provider, stifling competition and innovation.

• Quote Highlights:

  • Host A at [05:55]: “They're concerned these partnerships could stifle competition...”
  • Host A further explains at [05:54], “The FTC even used the term lock in...”

• Examples of Partnerships: Notable collaborations include Microsoft with OpenAI and Amazon and Alphabet partnering with Anthropic, which the FTC fears could give these companies disproportionate control over the AI ecosystem.

• Counterarguments: Microsoft defends its partnership with OpenAI, claiming it has spurred innovation and contributed to the success of one of the world's most prominent AI startups.

• Quote Highlights:

  • Host A at [06:33]: “They did?” referring to Microsoft’s pushback.
  • Host A at [06:34]: “...their partnership with OpenAI has actually helped create one of the most successful AI startups in the world and spurred innovation.”

• Balanced Perspective: Both hosts acknowledge the complexity of the issue, recognizing the potential benefits of such partnerships while also considering the legitimate concerns about market dominance and reduced competition.

• Quote Highlights:

  • Host B at [06:47]: “It's a complex issue...”
  • Host A at [06:54]: “It's a tough one, but it highlights this bigger question...”

---

4. Europe’s Call for AI Independence and US Export Restrictions

The final segment examines the U.S. export restrictions on advanced AI technologies to European countries and Europe's response advocating for AI independence.

• Export Restrictions: The Biden administration has imposed export controls on advanced AI chips from companies like Nvidia, AMD, and Intel to 17 European nations, citing national security concerns and the need to maintain technological supremacy.

• Quote Highlights:

  • Host A at [07:11]: “Chips to certain European countries.”
  • Host B at [07:13]: “And those are chips from, like, Nvidia, amd, Intel. Right?”

• European Response: European Commissioners Henna Vrkunin and Maruja Fischowicz expressed dissatisfaction, questioning why the U.S. treats its allies with such restrictions, thereby creating tension within transatlantic relations.

• Quote Highlights:

  • Host B at [07:21]: “European Commissioners Henna Vrkunin and Maruja Fischowicz...”

• Impact on EU's AI Ambitions: The restrictions hinder the EU's efforts to bolster its own AI capabilities, highlighting the intricate interplay between economic interests, technological advancements, and global politics.

• Quote Highlights:

  • Host B at [07:57]: “These restrictions could really slow them down...”

• Internal Division Within Europe: The export controls have also caused divisions within Europe, as some EU countries are exempted based on their status as key allies and partners, while others are excluded, raising concerns about the criteria and decision-making processes involved.

• Quote Highlights:

  • Host A at [08:13]: “It's dividing Europe itself...”
  • Host B at [08:21]: “Restrictions, the ones considered key allies and partners.”

• Future Implications: The hosts contemplate the longevity of these divisions and how shifts in U.S. administration could influence the future of these export policies, further emphasizing the dynamic nature of the global AI landscape.

---

Conclusion

The AI Deep Dive hosts provide a comprehensive overview of pivotal events shaping the AI domain. From high-stakes mergers and groundbreaking AI models to regulatory concerns and geopolitical tensions, the episode underscores the rapid evolution and complex interplay of factors influencing the future of artificial intelligence. Listeners are left with a nuanced understanding of the opportunities and challenges that lie ahead as AI continues to integrate into various facets of society and global affairs.

---

Key Takeaways:

• Perplexity AI’s merger with TikTok could reshape social media search functionalities amidst political uncertainties.
• DeepSeek V3 stands out in the AI race with its impressive capabilities and open-source approach, challenging established players.
• FTC's scrutiny of big tech partnerships highlights concerns over market dominance and the preservation of competitive landscapes in AI.
• U.S. export restrictions on AI technologies are straining transatlantic relations and pushing Europe towards advocating for greater AI autonomy.

Stay informed and ahead of the curve by tuning into future episodes of AI Deep Dive, where the latest AI breakthroughs and trends are explored in depth.

Transcript

[A]

Foreign. Let's kick off another deep dive today. We're taking a look at the AI Deep Dive newsletter.

[B]

Yeah, this one's popular, right?

[A]

Oh, absolutely. It's like the go to source for professionals. You know, those folks who are all about leveling up. Bigger paychecks, working smarter, not harder. Maybe even squeezing in a bit more vacation time.

[B]

I like the sound of that.

[A]

Don't we all. And today's newsletter, it's a doozy. We've got a potential shakeup in social media land. A Chinese AI company that's making waves. The FTC getting all concerned about big tech.

[B]

Oh, and the impact of US policies on Europe. I saw that.

[A]

Yeah, that too. We've got a lot to unpack.

[B]

Looking forward to it.

[A]

All right, so first up, let's talk about this merger that could completely, I mean, completely change the social media game.

[B]

Perplexity AI and TikTok.

[A]

You got it. Perplexity AI, the ones behind that super cool AI powered search engine, they want to merge with TikTok, specifically their US operations.

[B]

It's a bold move, for sure.

[A]

You think?

[B]

Well, I mean, especially when you consider everything that's going on with TikTok right now.

[A]

Right, right. That whole looming ban in the U.S. exactly.

[B]

And ByteDance, TikTok's parent company, they've been pretty resistant to selling it off.

[A]

So how does Perplexity even fit into all of this?

[B]

Well, their merger proposal, it's like they're trying to sidestep the wholesale issue. However, at the same time, they'd be bringing video content into their search engine, which could be a game changer.

[A]

Wow. And then you throw in all the political drama on top of that.

[B]

Oh, yeah. That adds another layer of complexity.

[A]

President Biden basically passing the buck to the next administration, which happens to be.

[B]

The Trump administration again.

[A]

Yep. And then you've got Trump potentially extending that deadline and TikTok saying they might have to, you know, go dark.

[B]

It's a lot of moving parts.

[A]

Makes you wonder, could this merger actually happen? And if it does, what on earth would a perplexity powered TikTok even look like?

[B]

Hmm, that's a good question. Imagine searching for something and instead of just getting links, you get TikToks.

[A]

Wow. Okay, let's shift gears for a minute. Let's talk about DeepSeek V3, the Chinese AI model. Yeah, it's been causing quite a stir. This thing can code, it can translate, it can even write emails.

[B]

And it's open source, right?

[A]

Yep. Meaning developers everywhere can get their hands on it. Pretty impressive. Stuff.

[B]

It's not just impressive, its performance is what's really turning heads. It's been outperforming some big names in the AI world.

[A]

Like who?

[B]

OpenAI Meta. Especially when it comes to coding tasks.

[A]

Really?

[B]

There's this test, the ATOR Polyglot test. It measures how well an AI can generate new code that fits into existing systems. And Deepseek V3 just blew the competition out of the water.

[A]

Wow, so we're talking like David versus Goliath, but in the AI world.

[B]

Exactly. And the scale of this thing, it's mind boggling.

[A]

I was looking at the numbers. It was trained on 14.8 trillion tokens.

[B]

Tokens? Think of them as pieces of words, sentences, code. It's massive.

[A]

It's like every single book in the Library of Congress, but times a thousand.

[B]

That's a good way to put it. And then there's a parameter count. Deepseek v3 boasts 671 billion.

[A]

671 billion parameters. What are those? Exactly.

[B]

So, parameters. They're like the knobs and dials that control an AI's behavior. More parameters mean more nuance, more sophistication. But it also means you need way more data and computing power to train the darn thing.

[A]

Makes sense.

[B]

Deepseek v3 has significantly more parameters than something like Metislama 3.1, for example.

[A]

Oh, okay, so it's powerful, it's versatile, it can automate tasks, boost creativity, even break down language barriers. Sounds almost too good to be true. What's the catch?

[B]

Ah, well, there is one thing. Deep Seek v3. It operates under the rules and regulations of China.

[A]

Meaning?

[B]

Meaning there are certain topics it steers.

[A]

Clear of due to censorship.

[B]

You could say that it's programmed to align with core socialist values, as they call it. So, you know, while it might be amazing at writing code or translating languages, don't expect it to win any awards for unbiased commentary.

[A]

Okay, that makes sense. So who are the brains behind this game changing AI?

[B]

Well, Jamseed comes from a company called High Flyer Capital Management.

[A]

High Flyer. What do they do?

[B]

They're a hedge fund that uses AI for quantitative trading. Pretty high tech stuff.

[A]

Interesting. And they created this AI model.

[B]

Yeah, and their founder, Liang Wenfeng, he's got this vision. He wants to push the boundaries of AI, and he believes that open source models like Deep Sea are the way to do it. He's basically challenging the dominance of those closed models developed by companies like OpenAI.

[A]

So it's like a clash of ideologies brewing in the AI world.

[B]

You could say that it's open source versus closed, transparency versus control.

[A]

And High Flyer is putting their money where their mouth is, right?

[B]

Absolutely. They've invested heavily in their own server clusters. Get this, they boast 10,000 Nvidia A100 GPUs. That's a serious commitment to achieving what they call super intelligent AI.

[A]

10,000. Wow. Okay, so open source versus closed. What do you think? Which side will come out on top?

[B]

It's tough to say. Both approaches have their pros and cons. Open source promotes collaboration, transparency, anyone can benefit from advancements. But then closed models, they might offer more control, more security.

[A]

It's like choosing between two paths. Both lead somewhere, but you're not quite sure where.

[B]

A good analogy. Yeah. And it's not just about who develops the AI, it's about how it's used, how it's controlled.

[A]

Which brings us to the FTC and their worries about big tech getting too cozy with AI developers.

[B]

Oh, yeah, that report raised some eyebrows.

[A]

They're concerned these partnerships could stifle competition, create a situation where a few powerful companies call all the shots in the AI ecosystem.

[B]

It's a valid concern.

[A]

The FTC even used the term lock in, meaning basically, companies become so dependent on a single provider for those essential AI services that they can't break free. It limits choices, it limits innovation. It potentially hurts businesses and consumers. I see they called out a few examples. Right. Microsoft backing OpenAI, Amazon and Alphabet partnering up with Anthropic.

[B]

Right. The concern is that those big companies have an unfair advantage.

[A]

Access to all those resources, computing power, the best talent, mountains of data, it's all there.

[B]

And that makes it harder for smaller companies to get a foothold in the market.

[A]

To be fair, Microsoft did push back on those concerns.

[B]

They did?

[A]

Yeah. They said their partnership with OpenAI has actually helped create one of the most successful AI startups in the world and spurred innovation. So what do you think? Good thing, bad thing? Somewhere in between.

[B]

It's a complex issue. There are definitely valid points on both sides, like weighing the potential benefits against the potential risks.

[A]

Right. It's a tough one, but it highlights this bigger question, which is how AI is going to impact the world on a global scale.

[B]

Ah, yes, the geopolitical implications.

[A]

Precisely. And, well, that brings us to the Biden administration's export restrictions on advanced AI.

[B]

Chips to certain European countries.

[A]

17 of them to be exact.

[B]

And those are chips from, like, Nvidia, amd, Intel. Right, the ones crucial for AI development.

[A]

Yeah. Basically the US is trying to control who gets their Hands on those technological building blocks.

[B]

The US Government's stance is that it's a national security thing to maintain our.

[A]

Technological edge, prevent adversaries from getting access to sensitive technology.

[B]

Right. But, you know, not everyone's happy about it, especially in Europe.

[A]

You know, European Commissioners Henna Vrkunin and Maruja Fischowicz, they weren't too thrilled. They basically said, we're your allies. Why are you treating us like this?

[B]

It's definitely created some tension. The EU has been trying to boost.

[A]

Its own AI capabilities, and now they're facing these roadblocks.

[B]

These restrictions could really slow them down. It highlights how interconnected everything is. You know, economic interests, technological advancements, global politics. It's all part of the AI landscape now.

[A]

And it's not just creating tension between the US and the eu.

[B]

Oh, you mean how it's dividing Europe itself?

[A]

Exactly. Some EU countries are exempt from these.

[B]

Restrictions, the ones considered key allies and partners.

[A]

Right. But others are left out in the cold.

[B]

It raises questions about how these decisions are being made, what criteria are being used.

[A]

It's like drawing a line in the sand. On one side, you've got the in group, on the other side, the out group. But even for those in the in group, who knows how long that will last?

[B]

It's true, things could change quickly, especially with the new administration coming into power in the U.S. well, we've covered a.

[A]

Lot of ground here. Social media mergers, international power plays. It's clear that AI is evolving at warp speed. Thanks for joining me on this incredible journey exploring the latest developments in AI and to our listeners, thanks for tuning in. Keep those minds curious, and we'll catch you on our next Deep Dive.