What Perplexity's Raise Says About the State of AI Hype

0:00Today on the AI Breakdown, we're discussing what Perplexity's current and potentially future raise says about the state of AI hype. Before that on the brief, rumors from Apple, a Coca-Cola and Microsoft team up, and more features from Meta's AI glasses. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube, our Discord, and our newsletter. Welcome back to the AI Breakdown Brief, all the AI headline news you need in around five minutes. Another day, another little drip of information around Apple's forthcoming AI strategy.

0:34According to a post by the Weibo user known as PhoneChipExpert and re-shared by MacRumors, Apple is, quote, said to be developing its own AI server processor using TSMC's 3NM process, targeting mass production by the second half of 2025. They continue, Apple has ambitious plans to design its own artificial intelligence server processor. As to why they might be interested in this, they write, Apple's purported move towards developing a specialist AI server processor is reflective of the company's ongoing strategy to vertically integrate its supply chain. By designing its own server chips, Apple can tailor hardware specifically to its software needs, potentially leading to more powerful and efficient technologies.

1:09Apple could use its own AI processors to enhance the performance of its data centers and future AI tools that rely on the cloud. While Apple is rumored to be prioritizing on-device processing for many of its upcoming AI tools, it is inevitable that some operations will have to occur in the cloud. So two things. One, how credible is this source? Well, MacRumor points out that they have a number of previous accurate claims around Apple products. But when it comes to would this make sense with their strategy, it seems like self-evidently yes. Apple likes controlling everything end-to-end, and even if they're not able to do that before they start putting their AI strategy into market, you better believe they're going to be interested in moving towards that sort of model.

1:45Apple researchers also just released another new OpenLLM that they called OpenELM, which while the details are beyond the scope that we're going to get into here, again just shows how much activity is going in the AI space for this company. Over in OpenAI land, the company has announced what they call more enterprise-grade features for API customers. OpenAI announces PrivateLink, which they call a new way that customers can ensure direct communication between Azure and OpenAI while minimizing exposure to the open internet, i.e. enhanced enterprise-grade security, better administrative control that gives organizations more granular control over individual projects in OpenAI, and of course, an enterprise version of the new Assistance API, which has also been improved for developers more broadly.

2:24Finally, they announced more options for cost management. Now, I'm sure that enterprises will be interested to learn about these changes, but I think the Flowers from the Future account sums up many of our feelings when they write, We don't care about corporate quality of life nonsense. We want GPT-5. That same account, who has had some accurate leaks from OpenAI in the past, also interestingly tweeted, OpenAI not feeling threatened by the fact that other models come very close to GPT-4 obviously shows that they have already left the LLM game behind and don't really care anymore. GPT-5 isn't just a way better GPT-4.

2:55It's the next game after chat. They let the others fight for it now. Now, not everyone agrees. There are other reasons why OpenAI might not be racing to release GPT-5. However, it is an interesting thing to speculate on. Over in big company land, Coca-Cola and Microsoft have announced a new five-year strategic partnership to, quote, align Coca-Cola's core technology strategy system-wide, and basically, in so many words, leverage AI. They write that as part of the partnership, Coca-Cola has made a$1.1 billion commitment to Microsoft Cloud and its generative AI capabilities. Effectively, this is just a standard Microsoft Azure deal with a lot of emphasis on AI that's taking advantage of the fact that there is an AI component and a big fat juicy number along with it, as well as two very prominent names to get some buzz and make some hay.

3:39But then again, as the drumbeat of people wondering if businesses are actually going to see value from AI and going to commit deeper to it gets a little bit louder in the press, This type of story could be an interesting counterweight. Amazon has announced a new feature in Bedrock called the Custom Model Import. Basically, this allows companies working with AWS to import and access their in-house generative AI models, not just the models that are in Bedrock's library. Said the VP of generative AI at AWS, there have been AWS customers that have been fine-tuning or building their own models outside of Bedrock using other tools.

4:11This custom model import capability allows them to bring their own proprietary models to Bedrock and see them right next to all of the other models that are already on Bedrock and use them with all of the workflows that are already on Bedrock as well. In other words, this to me is really just an extension of Amazon's existing point of view that there will not be one model to rule them all or even a handful of models to rule them all. There will instead be tons and tons of custom individualized models, and they want to be able to support all of that. Lastly, it is not just the Meta suite of apps that are getting a Meta AI upgrade.

4:41Meta's Ray-Ban smart glasses are getting it as well. The company writes, we're rolling out Meta AI with Vision so you can ask your glasses about what you're seeing and get helpful information completely hands-free. The company continues, we started testing a multimodal AI update in December so you can ask your glasses about what you're seeing and they'll give you smart, helpful answers or suggestions. That means you can do more with your glasses because now they can see what you see. Starting today, we're rolling this functionality out to all Ray-Ban Meta smart glasses in the US and Canada in beta.

5:08So an example, if you've got a menu in a different language in a different country, smart glasses can use the built-in camera and Meta AI to translate the text. So far, Meta's Ray-Bans have been the surprising AI AR wearable breakout, and this feels like just another feature that could keep that momentum going. For now though, that is going to do it for today's AI Breakdown Brief. Up next, the main AI breakdown. Attention, AI breakdown listeners. Consensus 2024 marks the 10th gathering for all things crypto, blockchain, and Web3. However, importantly, this year's agenda will also dive deep into AI-driven transformation.

5:41And the speaker lineup includes the leading minds and innovators at the forefront of this digital renaissance. Don't miss the Consensus AI Summit to cut through the hype to find where true transformation and opportunity lie. Listeners to this show can get 15 % off registration with the code AIBreakdown. Visit Consensus2024.coindesk.com to learn more. Some of the folks who will be at Consensus this year include Guillaume Verdun, aka Beth Jezos, founder and CEO of Xtropic, as well as spiritual leader of the accelerationist movement, Neil Stephenson, co-founder of Lamina One, and Brendan Eich, the CEO of Brave Software.

6:13Again, go to consensus2024.coindesk.com to learn more and get 15 % off registration with the code AIBreakdown. Before we get back to the AI Breakdown, I want to share something fun we have coming up on Superintelligent next month. Super Intelligent is, of course, our new platform for teaching people how to use AI in a way that is much more fun, fast, and practical. The platform has hundreds of short tutorial videos, each of which is paired with a set of step-by-step instructions that get you using AI tools in minutes, not hours, and certainly not days. For those of you who haven't signed up yet but want to check it out, in May, I am running a special NLW cohort.

6:47What this means is that people who sign up with the code NLWMay will get$5 off their first month, but they'll also have access to a private channel in our Discord with me. I'll be handpicking tutorials each week that I think are the most useful to start with, and I'll also be available for questions, advice, and feedback from this group. Spots for this cohort are limited, so if you want to be a part of it, again, sign up at besuper.ai with code NLWMay. That's besuper.ai with code NLWMay. Welcome back to the AI Breakdown. Usually, I would not make an entire main part of an episode about a fundraising announcement.

7:24However, nominally, that's what today's is. AI Google search engine competitor Perplexity is now officially a unicorn, and as I'll mention, might be raising at an even higher valuation already. We will cover all of the actual investment news, but what's interesting is also what the investment says about, one, the state of venture capital in AI right now, And two, the whole debate around proprietary technology versus quote-unquote rappers. Let's start with a very representative quote from T. Fallabilist on Twitter, which says, Google's search is now annoyingly bad, and after tweaking my search habits a bit, perplexity is refreshingly good.

7:58And yes, even for serious tasks like finding key info for investment research. For those who aren't familiar, perplexity is, like I said, an AI-powered search engine. When you search on perplexity, it gives you what looks more like a dossier than Google's classic set of links. It provides an AI curated answer that sums up what it found, but it also cites its sources. Not only does it link to its sources, it also explicitly ties them to the parts of its answer where they are relevant. So for example, when searching for the firewood that burns the longest, the first sentence of the answer is oak is the firewood that burns the longest, for which perplexity is pointing to three sources.

8:33There is also a side panel where you can search videos, and if you have a pro account, you can generate images. Now already this is a great interface and something that people are really responding to. However, they also have the ability to ask follow-ups. For example, the related follow-ups that they suggest are things like how to properly season firewood. What's the difference between hardwood and softwood? Let's click that one. And as you add more of these searches, all of which are part of the same overall user experience, you're building a little knowledge base for yourself that then becomes saved as part of your library.

9:03Now, Perplexity has a bunch of other features for pros, including collections, which allow you to associate a bunch of searches together, which is particularly valuable for things like planning a family vacation, where you might have lots and lots of different types of searches. So that is perplexity in a nutshell. Many people have had this same experience of finding themselves just naturally shifting to perplexity and away from Google. Enough people have had that experience, in fact, that the company has just announced a raise of$62.7 million, led by Daniel Gross, but also including notables like Stanley Druckenmiller, who is a famous hedge funder and extremely influential investor on Wall Street, NVIDIA, Jeff Bezos, Gary Tan from Y Combinator, Andre Karpathy, Naval Ravikant, and more.

9:42This is technically a Series B1 round, following on from a Series B in January. As part of their announcement, Perplexity said that they've grown to serve 169 million queries per month, hitting more than a billion over the last 15 months. They also announced that they're seeing increased demand from global telcos who want to bring AI to their mobile subscribers and from enterprises. They noted that they have partnerships now with SoftBank and Deutsche Telekom, which together will distribute Perplexity to a combined total of over 116 million users. And they announced Perplexity Enterprise Pro as their first B2B offering.

10:12Some of the customers include Databricks, NVIDIA, Zoom, Cleveland Cavaliers, Latham & Watkins, Universal McCann, Paytm, Vercel, and more. Bloomberg had the exclusive on the announcement and comes to a fairly obvious conclusion. Perplexity's valuation has doubled in three months in a clear sign of investor enthusiasm. So what additional information did we get from this Bloomberg announcement? The biggest thing was probably that the company has been growing fast, not just in terms of use, but also in terms of revenue. I think previous numbers that they had suggested that they were making something like$3 million, but now they're up to$20 million in annual recurring revenue.

10:44Part of that, it seems, comes from the company's new enterprise efforts. So alongside the announcement, Perplexity announced Enterprise Pro. And so what does that actually mean? How are companies using this? Well, luckily, they gave a lot of examples. And by the way, for any AI startups out there, being clear about how others are using your product can be very, very valuable. So for example, in this announcement, they list that product teams at Zoom are using it for targeted search, HP Salesforce is using it for in-depth prospect research, data teams at the Cleveland Cavs are researching ticket sale trends as well as doing partnership prospecting, and other things like that.

11:16The CEO of Databricks estimated that Perplexity was helping their team save around 5 ,000 working hours monthly. Additionally, they announced that they were looking to expand their user base to other places outside of America. To do so, they touted new partnerships with Japan's SoftBank and Germany's Deutsche Telekom, which combined have more than 300 million users worldwide, and which perplexity suggests will help them reach 113 million new users. So I said that I wanted to talk about what the implications were for the state of AI, specifically AI hype. There are two dimensions of that that I want to address.

11:45The first has to do with the idea of AI wrapper companies. There has long been a sense that if you weren't building your own model, you were going to have a hard time defending your business. A lot of the discussion in the second half of last year was about that. A number of the companies that were struggling with layoffs were companies who had just basically tried to wrap ChatGPT and package it for a specific audience, and who were finding that kind of difficult. Perplexity shows, however, that pursuing a very specific product, in their case, of course, Search, might lead to some different dynamics.

12:15Now, I don't want to minimize the technology that Perplexity has built themselves to do the summarization associated with sources, etc. But they are not building their own models. They are pulling from existing models. And in fact, you can go to Perplexity Labs and test out different models against each other. Point being, though, that this is a product company, not a model company, and yet they're getting a lot of traction. In other words, as we see models themselves get more commoditized than they are now, will the emphasis shift back towards product companies that actually find product market fit around a specific use of AI?

12:46Might we see a complete inversion of that idea that it's all about the model underneath to instead it all being about the product that sits on top? I'm not sure, and it's very likely that it's not that dramatic a switch, but it's something really interesting to consider. Second, in terms of what it says about hype from a fundraising perspective, it's also pretty hard to say. On the one hand, this is a company who raised just a few months ago and now are raising at double the valuation, which seems pretty hype-y. But then again, the company is clearly growing. And in fact, one could argue that the economics of this most recent investment look a little bit better from a forward revenue perspective than did the last investment.

13:21And the most interesting part of this, however, is that TechCrunch is reporting that in addition to this raise, which was just confirmed, the company is out again raising another$250 million or more at a$2.5 to$3 billion valuation. TechCrunch writes, We understand from multiple sources close to the company that perplexity is raising a further round. NEA and IVP, both previous backers of the company, are among those looking to invest in this larger round, according to sources. A partner from an existing investor said, Now when it comes to why jump on this now, I actually think that TechCrunch sums it up really nicely.

13:55They say, perplexity's reason for raising again so soon? Yes, perhaps to capitalize on customer and investor interest at what one investor described as a zeitgeist moment for the startup, but also because of the mechanics of building any kind of AI service right now. Said that investor, compute is very expensive, so they may need to raise for that reason alone. In other words, while perplexity may be having a moment and leveraging that moment to move a little bit faster on fundraising than they might otherwise have, it can't be far from their minds just how expensive it is to compete in this AI space.

14:24In that environment, you kind of want to take the money when you can get it. So like I said, I think this is a really interesting story, not just from perplexity's point of view, but also from what it says about other trends in the AI space. However, that is going to do it for today's AI Breakdown. Until next time, peace.