Welcome all you cool cats to The Neuron! I’m Pete Huang

Today, we’re talking one thing and one thing only: the upcoming launch of ChatGPT Search, and why it’s striking at the right moment in Google’s history.

It’s Saturday, May 4th. Let’s dive in!

OpenAI has a problem with leaking.

When you think about it, of course they do. OpenAI is the leading AI company and it’s not even close.

Of course there’s gonna be a ton of people trying to figure out what their next move is. People do that for public companies, for no-name companies that suddenly become targets for mergers and acquisitions, for companies that have no name at all but mysteriously do important work.

OpenAI is high-profile, incredibly consequential and literally everything they have released has changed our understanding of what is possible with AI.

Even worse for OpenAI, they’ve got 700 people in a small city of San Francisco, population of under a million, where a constant stream of AI meetups, tech parties and very well-connected networks means that if you’re in the tech community here, not knowing someone who knows someone that works at OpenAI is actually way less likely than you think.

Still, OpenAI has to do what they can to crack down on the possibility of leaks. Internally, they’re highly restrictive about who knows what. If you’re not on the technical side, there’s a very small chance you even know a project exists until it’s time for launch.

They also have hand readers and fingerprint scanners on their key entrances. And they’ve been doing their best to track down who leaks. In April, they fired two researchers for leaking information.

But, despite all of this, you have anonymous Twitter accounts like jimmy apples, who have correctly called upcoming product releases and exact timing for OpenAI and other AI companies like Anthropic and Google. Nobody knows who they are. Nobody knows if they work at OpenAI or just hear a lot of things.

We just know they have access to information that we don’t.

So here we go. April 30th, from jimmy apples:

“10am, 9th of May for an Openai event apparently, might not be model release but search engine announcement. 

Guess they can’t help themselves to upstage Google I/O “

This is an exciting tweet. But it’s not enough for anyone to make firm proclamations. After all, Jimmy is great, but Jimmy is an anonymous Twitter account who’s not citing any sources.

But the tweet coincides with another group of hobbyists that love trying to figure out what’s next. These folks scan the Internet looking for new web domains that OpenAI has set up.

And that sounds very hard, but even you and I can come up with that list. You know how some web domains start with https instead of http? When you want to create one of those, it publishes a certificate online with that web domain. That includes subdomains, meaning not just apple.com but podcasts.apple.com and whatever.apple.com. In fact there are tools where you can literally just enter a site and see all the subdomains.

As early as April 24th, people figured out that OpenAI created search.chatgpt.com, which really increases people’s confidence that a search product was finally in motion.

The fact that OpenAI was working on search at all isn’t news. In fact, The Information reported this in February of this year.

And CEO Sam Altman, who is incredibly well trained at talking about things at the right time, went on the Lex Fridman podcast and openly talked about search. Here’s that clip:

“The intersection of LLMs plus search, I don’t think anyone has cracked the code on yet. I would love to go do that. I think that would be cool.”

And indeed he has.

So what will the search product include?

Other investigators think it’ll have a standard chat-driven search experience, image search, tools like weather and sports news and finance news and timezone difference.

It’s almost like a new Yahoo homepage, which for many years was one of the most popular pages on the internet.

OpenAI is angling to become your new default homepage when you access the internet.

All of this is hitting Google at an awkward time. The Mountain View-based tech giant has had a spat of bad-looking news recently that dovetails with a years-long trend of growing resentment of Google search.

I’ll just say it: searching Google sucks. But I don’t think a lot of people realize just how much it sucks, mostly because we haven’t really had a viable alternative.

This week, I was doing some reading on California bill 1047, which I talked through on our last episode. But when you search California bill 1047, you really don’t get what you’re truly looking for. I was looking for a breakdown of the bill, the key arguments for and against.

But when I searched California bill 1047, the first link I clicked on is the official California state legislature page. Great, definitely correct, tells me nothing in plain English. The second link is a press release from the bill’s main sponsor. Not what I’m looking for. The third link is from a law firm. Useful, but incomplete.

And so on.

It’s way worse if you’re ever trying to buy things or get advice. If you search “best running shoes for men”, you get a flood of suggestions, filled with random items from Google Shopping or all these pay to play buying guides from struggling media outlets trying to make an extra buck off of you.

If you want travel advice or a recipe, just be prepared to wade through 3,000 words of family history or random stories before you actually get to the good stuff.

All that has happened because Google set up its algorithm to incentivize that. I swear I don’t hate travel bloggers. I love that they’re really good at their game. I think the Google algorithm is broken.

The man we have to thank for that appears to be Prabhakar Raghavan, a senior vice president at Google who is in charge of a majority of Google’s revenue: ads, maps, payments, Google Assistant, and yes, search.

In fact, there’s a recent article from tech PR and writer Ed Zitron titled “The Man Who Killed Google Search” who details how Raghavan, then the Head of Ads, pushed a growth at all costs mindset. That mindset played foil to his peer Ben Gomes, the Head of Search, who wanted the team to focus on ways to grow the number of search queries without damaging user experience.

So ads, the money people, are pushing for more, while search, the product people, are pushing for solving user needs. It’s a classic tension, but at this scale, it can get really hairy. And political.

As corporate mindsets would have it, Raghavan eventually wins and even kicks out Gomes to take his place as the Head of Search. And Gomes, well, he gets shoved to the side and becomes the SVP of Education, a completely unrelated business line.

Over the next few years, that growth at all costs mindset leads to declining satisfaction with what you’re able to find on Google.

In fact, have you ever noticed yourself adding the word “reddit” after your search? Or when you do search for something, that “reddit” almost always appears as the first autocomplete suggestion?

It’s because people don’t trust the results on Google, so much so that all they use Google for is to search Reddit.

Raghavan knows Google search is in a tough spot. In a recent all-hands meeting, Raghavan spent more than half an hour trying to rally his team to work harder, but without specifics about what was going wrong.

He said things like: “It’s not like life is going to be hunky-dory, forever” and “We need to twitch faster, like the athletes twitch faster”

The only tangible thing he had for his org was that they would have less time to complete projects.

The takeaway from Raghavan isn’t necessarily that he’s a bad person. It’s that his behavior is exactly why there’s an opportunity to disrupt Google.

Search and ads continues to rake it in. In the first quarter of 2024, Google search and ads pulled in $46 billion in revenue. That’s up from $40 billion a year ago.

It’s the core of Google’s business. And with shareholders looking for continued growth, they have to keep squeezing Google in its current form, try to find new and better ways for people to engage.

Meanwhile, all this generative AI stuff happens, and it’s really hard to thread the needle between properly investigating and innovating with new technology while maintaining your market position. You are naturally incentivized to only do the latter, to protect your golden goose.

This is called the Innovator’s Dilemma. Startups use these emerging technologies to chip away at small edges. Meanwhile, the big incumbents need to keep growing sales in their current model to keep shareholders happy. Until the startups product gets good enough that it starts to topple the big guys. The big guys just didn’t react quickly enough.

Let’s talk about who these startups are and where this new ChatGPT search fits in.

The two most notable startups are called Perplexity.ai and You.com. If you’ve been following the newsletter, you know that we talk about Perplexity a lot.

Because it’s good. We like it.

And in particular, we like it because it reminds us not just of how broken Google is, but the fact that most people don’t even realize how broken it is because they’re so used to it

In Perplexity, you lead by chat. It’s basically a ChatGPT but if it was plugged into the Internet. So in the case of California bill 1047, instead of having to click through a bunch of links to figure out the right thing, you can just say “explain california bill 1047” and it will.

No need to read all the links anymore. Just let AI do it for you, then you can click into the link if you want to read it yourself.

Great product, that’s it right? Battle’s won? Not quite.

We have to remember that Google is still Google. As terrible as it is, they can get away with it because people still use it. For the past few years, in the US, Google has controlled 87-90% of all searches on the Internet.

90%! And that’s through the period where Prabhakar Raghavan is in charge of Google search.

That’s what happens when you’ve got more than 15 years of dominance in the market. You have an entire generation who has used nothing but Google.

For Perplexity to make a dent, they really need to shoot big. They need billions of searches. Literally. Google has anywhere between 4 and 8 billion searches a day. And Perplexity is racing against Google to get there.

Google, despite all of its challenges at its size, is working on AI search, something called Search Generative Experience, where it shows you an AI-generated summary at the top of the page before showing you the links. For Google SGE, cost was a major issue, it was just very expensive to process that search and would make Google way less profitable if they rolled it out as it is today. But this week, Google also announced that they’ve cut that cost by 80%.

Google isn’t the only one who has to worry about money. Perplexity does, too. For one, they’re spending a ton, so much so that they’re now looking for another $250 million in fundraising, on top of its existing $165 million.

And they’re trying to figure out how to make money, too. Perplexity currently charges $20/month for an upgraded version of the product. They don’t do ads like Google does. But they’re also not sure if this is the best way to do it. Earlier in April, Perplexity’s chief business officer told the media that they’re going to try ads, which is a pretty big departure from earlier statements.

Finally, here’s the most awkward part about Perplexity’s situation. Perplexity uses Google data. Like, they’re literally trying to take down Google but they’re also like “wait we need you” at the same time. That must be a fun negotiation.

All put together, Perplexity knows they’ve built a good product. The early adopter types have really picked this up, and it’s used across Silicon Valley. Everyone here knows it.

But they also know that time is of the essence. They’re in a mad scramble to get more users, trying every tactic they can to get more eyeballs on them. That includes their short news podcast, Discovery Daily by Perplexity, which has amassed quite a following - I’ve heard the numbers but I can’t share - but they’re hoping that the podcast audience converts to people using Perplexity.

It includes partnerships with pretty much anyone they can. For people who buy the Rabbit R1, the new AI device, they’re giving away a 1-year Perplexity subscription. And we’ve seen them do this partnership with a number of companies, so clearly the revenue from those subscriptions is less important than getting people to love the product.

Which brings us back to ChatGPT search. People already use ChatGPT for information. It was one of the early use cases, and even today, when I see someone use ChatGPT for the first time, the first thing they do is type in something they would search on Google.

ChatGPT Search can avoid some of the same issues that Perplexity has. For one, OpenAI has over $10 billion in fundraising. They have a lot of money. And frankly they could probably get more if they wanted to.

They also have way more users. While Perplexity has reported about 10 million people who use Perplexity at least once a month, ChatGPT has over a hundred million people using it at least once a week.

As for the data, because of OpenAI’s partnership with Microsoft, they can use Bing’s search data instead of Google’s. They also have their own programs that can crawl the internet, so they don’t even have to rely on Bing or Google, depending on how that’s deployed.

All this leads to May 9th, the currently estimated date for OpenAI to launch ChatGPT search, which should be available at search.chatgpt.com. By the way, I’ll be at the Google I/O conference, which will be the following week on May 14th.

Depending on how that launch goes, if it goes at all, we might be in for a bit of an awkward time at Google I/O. I’ll be reporting back with the hot gossip if that’s the case.

Your big takeaway for ChatGPT Search:

The allure of toppling Google is simply too attractive to pass up.

People have been calling for Google’s downfall since ChatGPT launched, and we might just be witnessing the start of a real battle for one of the largest prizes on the Internet.

Throughout the 2000s, Google became the dominant search engine with an algorithm that produced better search results. That algorithm beat Yahoo, AltaVista and many others now in the tech graveyard.

And it hasn’t really had a challenger since. Market share has stayed at sky high levels for more than 15 years, approaching 90% in the US. And while other Big Tech companies like Apple could have challenged them by building their own search engine, Google just paid them off in order to not compete. For Apple, that means $20 billion a year for Apple to use Google as the main search engine instead of anything else or building their own.

15 years is a long, long time to be dominant. In the age of the Internet, when new upstarts can take over in an instant, Google now has the challenge of trying to reboot its ability to innovate and disrupt itself. Or else ChatGPT Search, Perplexity and others may act on an existential threat.

This is not a battle that will happen overnight. Google isn’t going anywhere tomorrow. And they have lots of ways to defend.

But at the minimum, we’re seeing the first real moment of weakness for Google, and startups just can’t help themselves.

Some quick hitters to leave you with:

• Online investigators are now wondering if GPT 2 chatbot, which I talked about last time, is be a modified version of GPT-4 that’s specifically used for ChatGPT search. Sorry folks, it’s not GPT-5 or even GPT-4.5
• Ever wonder what Warren Buffett says about AI? Here’s a quote: “If I was interested in investing in scamming, it’s gonna be the growth industry of all time and it’s enabled, in a way by AI”
• Users of X are going to find news summaries inside the app. These will be powered by Grok, the AI model built by xAI.

‍

This is Pete wrapping up The Neuron for May 4th. I’ll see you next week.

‍