B2B Wins #10: Google is dead. Long live Google.
Google is way ahead of ChatGPT. Reinvention is coming. For its sake I hope so.
The first time I used ChatGPT, I did a search. Why did I do a search? I don't know. It probably has something to do with all the searches I do these days as part of the market research I'm doing. But there's something else. The Google search user experience
sucks is awful. So is Siri's. So are most of them.
As a search user, I’m tired of scrolling past a sea of ads to find the information I need. As a company with content, I’m weary of the pay-to-play model. As a knowledge seeker, I’d like the answers presented directly without all the scrolling and clicking. So, my first interaction was a search. I yearn for a better search user experience.
I did a search in a domain where I have expert knowledge: Enterprise Search. It would be a search that any CIO-type would do if they were trying to understand the enterprise search landscape. The results that ChatGPT gave me were not wrong though they were misleading. You can see the entire interaction in the images posted at the bottom of this article. You can see the same prompt put to Google.
While the results were lackluster, the user experience was magical. I asked a question in plain English. I clarified the question. I provided more information as context. The search results got better as I tightened it up. This is the search I'm looking for. Google's days are numbered. Or maybe not.
We hate chat
“…80% of consumers say customer service chatbots increased their frustration level…”
Am I right about people wanting to converse with computers in plain English?
According to Forbes, we appear to hate chatbots. They're everywhere, especially in customer service. And they answer so few of our questions. Well, sort of.
I was at a conference two weeks ago and it was clear that companies are using chatbots to great effect. Answering simple questions, which are mind-numbing for customer service reps to answer and costly for companies to fund, is where chatbots shine.
According to vendor sales reps, this is freeing up those customer service folks to answer the questions that stump chatbots. Of course, this also leads to frustration when people spend a few minutes conversing with a chatbot only to have to punch out to a real person. Why not just start with a person they ask?
Those questions are far too complex for me to answer. What I do know is we talk to our electronic devices all the time. The responses are not always satisfying. But it's how we're wired to work. I have no doubt that chatting, conversation, and question-answering are game changers for digital interaction. And for search.
“I found this for you on the web” -Siri
"Ready it for me!"-Everyone
It would seem that Google got blindsided by the launch of ChatGPT. To be fair, we all were. OpenAI deployed ChatGPT unexpectedly with well-calculated fanfare. It was one of the most buzzworthy product launches of the past decade. It was a stroke of marketing genius.
It’s not as if Google hasn’t been working on conversational technologies related to knowledge discovery. Everything that appears on Google’s search results page requires some understanding of content in a way that can begin to mirror conversation.
We first saw Google engaging us in a conversational manner with their “People also ask” box that appears in some searches. It shows actual questions and their respective answers. Imagine that, you don’t have to click on a blue link. You just get an answer.
This user experience is far less magical than ChatGPT. I believe that was intentional on the part of Google. Today Google is essentially constrained to only provide answers that exist within a single web page.
I’m not sure why they limit themselves in this manner but I suspect it’s because they’re concerned with ensuring that there is clear attribution for the answer. A large part of Google’s value is that they’re able to demonstrate that the answer actually comes from a website that has domain authority. In a chat context, at least so far, we haven’t seen any attribution. In my “search” with ChatGPT it never told me where it was getting its answers. If I wasn’t a domain expert, I may have been inclined to believe its answers.
In May 2021, Google announced its LaMDA conversational language model. It’s every bit as cool as ChatGPT. What it isn’t is available to anyone to use. Google has been very careful to limit access. For over a year it was only available to select academics. Recently, they’ve begun to make it more available to the developer community.
Will Google take advantage of this moment to accelerate the evolution of its platform?
Can we afford this?
“I estimate the cost of running ChatGPT is $100K per day, or $3M per month” - Tom Goldstein
One of the concerns for this technology is just how much computing power is needed to deliver conversations. Tom Goldstein’s estimates were based on the volumes we see today for ChatGPT. If there are a million users, are they getting 1M questions a day? Maybe. Seems high.
OpenAI has not declared any pricing for ChatGPT but Sam Altman said on Twitter that he felt that each chat was “single digits cents”. Let’s say it’s five cents. So how much would it cost to replace Google with a ChatGPT-based search engine?
There are 8.5 billion searches done on Google each day. Assuming that all that shifts to a conversational model like ChatGPT that’s $425M per day just for computing.
For the full year 2021, Google’s expenses related to generating revenue were $110B. Just the cost of running ChatGPT on all those searches would be $155B.
And that’s just the computing power. You’d need all the personnel required to manage the technology. Serious optimization is going to be needed in order to make running these conversational models as search engines affordable. Or the economic model is going to have to change. Would you pay a penny for a search that was done conversationally?
Google can’t afford to ignore it
Alphabet cash on hand for the quarter ending September 30, 2022 was $116.259B - Alphabet Quarterly Report
“Yahoo squandered its massive head start and let each wave of new technology in search, social, and mobile pass it by”— Forbes
Google is in a bit of a bind. If someone can figure out the architecture that makes ChatGPT more truthful and can optimize operations to make it cost-effective a search engine based upon ChatGPT could very quickly dominate the search industry. This isn’t just speculation. There is a cautionary tale from twenty years ago.
For those of us who roamed the internet in the early days, Yahoo was the search engine of choice. Technically, it was pretty dull. But it worked well enough. Unfortunately for Yahoo, when Google launched in 1998 they failed to respond in a meaningful way to the threat. They didn’t innovate. It cost them dearly.
Yahoo, worth $125B in 2000 was sold together with AOL in 2021 for $5B. Yahoo and its subsequent owners missed many opportunities to claw back market share from Google. Their big problem was a failure to respond with any innovative search technology that could deliver results as good or better than Google’s. So is Google in a better position than Yahoo was in 1998?
Google has plenty of cash. That’s in the plus column. In the minus column is Google’s corporate culture. Google is sluggish and secretive. They’re intensely data-driven so all changes are measured to ensure they’re improving the customer experience. That customer experience is focused on whether Google believes questions are being answered correctly. That good. It’s also slow.
Secrecy is also a hallmark of their business. They don’t tell anyone how their tech works. That’s mostly to prevent folks from gaming their ranking algorithms. It’s also to prevent some of their emerging tech from getting into the hands of potential usurpers. This culture may be the thing that kills Google.
Google has demonstrated their LaMDA models are likely every bit as good as ChatGPTs. So what are they waiting for?
The dustbin of history provides some lessons. Blackberry, who created and owned the smartphone market, disappeared when Apple launched the iPhone in 2007. Digital cameras killed Kodak. Kodak invented the digital camera! Most companies in this dustbin failed to adapt because they were trying to protect legacy cash flows. It’s a story as old as time.
Google is dragging its feet because it hasn’t figured out how to ensure that the transition to conversational search doesn’t kill the advertising golden goose. It’s not an impossible problem to solve but there’s bound to be some revenue disruption.
Will there be display ads—hello banner ads!—within the answers? What is the economic model if suddenly there is no space for ads? Would users pay a penny to ask a question?
So many questions that Google will need to answer. If they don’t answer them, someone else will. In twenty years time, Google will be sold off so a private equity company for pennies on the dollar if it doesn’t crack the code on conversational search before OpenAI does.
Google broke my brain. Chat will fix it.
“dead presidents mountains where at” — A search that makes sense to Google.
One of the things that folks ask is “Can ChatGPT be as good as Google?”.
Good? Yes, Google can find answers. But it’s awful from the perspective of a user experience. Even if you set aside the riot of commerce modules and ads and info boxes that litter a Google search results page, you need to speak a foreign language just to ask Google a question.
We’ve had to learn to speak Googlish. It’s a language that is comprised of strings of words that we’ve learned could lead to a search result that makes sense. And we’ve learned to be blind to the crap on the results page and zero in on what makes sense to us. Nobody would intentionally design such a thing if they were starting from scratch.
Could ChatGPT-powered search be better than that? Yes. I’ve already seen some systems architecture charts that show how a “truth” layer could be put on top of ChatGPT to eliminate all the “howlers” that ChatGPT comes up with when it tries to relate facts.
People are going to love a search engine that talks to them in normal conversational language. They’re also going to benefit from one that can determine truth from lies. There’s so much promise in this new user experience. This is Google’s game to lose.
Google has the dominant position. They have a boatload of cash. They have technology that is comparable to ChatGPT. But just because they can save themselves, doesn’t mean they will.
As Forbes described in their December 12th article about ChatGPT:
“Soon, you will be able to have helpful assistants that talk to you, answer questions and give advice,” Altman tweeted in reference to the future of AI chatbots. “Eventually, you can have something that goes off and discovers new knowledge for you.”
That’s what people want. An assistant. That speaks their language. That helps them understand their place in the world. That doesn’t require them to know a new language.
Get going Google. This is your most desperate hour.
I’m off next week!
Next week I’m taking some time off. I hope you’re all getting some well-earned rest during the holidays. Next week’s newsletter will be a reading/watching list that has helped inform my professional journey this year as well as in the past. I’ll see you in the New Year!
The ChatGPT Search
The Google Search
Ads. Ads. Ads. No answers.