Google's True Moonshot
stratechery.comIs it possible that google will build the next great AI? Of course.
But right now they really really seem to be failing at it. Maybe the core tech in gemini is great. But it's nowhere.
Just as an anecdote, I tried to actually pay google for an AI product they claimed to have launched - the image generation, Imagen 2. And apparently, I can't. Even after tens of e-mails and a call with an account manager, the response is "uh, follow us on twitter and explain why you are good at building AI tools". Jeez, buying a service is not supposed to be like a job interview. It's supposed to be like buying the same service from OpenAI - enter credit card details and go.
So, the issue with google is that they took the wrong approach - build it in-house at a big company. What a big company has, are lawyers. Very good ones. The job of lawyers is to avoid risk. And they are great at it. However, building these sort of cutting edge services requires taking risk. And you can't really do that at a large company.
This is why Microsoft is winning - they realized that investing into a startup that has no lawyers and is willing to take risk is the right path to quickly getting to the result. This is also why dalle3 and chatgpt4 are available for everyone today. And Geimini ultra isn't.
Does anyone remember Google Video? It was a horrible experience. The upload process was a pain, and I think some approval was needed at some point, or you had to be sponsored by someone already on the service.
Then Google bought Youtube.
In the AI space it's unclear what/who they could buy -- OpenAI being obviously out of reach -- but it's possible they could find a good match.
> In the AI space it's unclear what/who they could buy
They bought the most promising AI startup in the world for $650m and failed to integrate it or make use of it until ChatGPT was released.
That was DeepMind, nine years ago.
everyone in this thread agrees google problem is all management and product and that buying a company is just a palliative way of acquihiring a still ungoogled management team. which don't last long, obviously.
you're all just telling each other exactly this in hundreds of different ways.
>In the AI space it's unclear what/who they could buy
I think Google invested like $2B+ in Anthropic
Anthropic feels very much like google though. It took me 6 months to get access, once I finally did it was only a testing type account that then requires me to talk to sales before they take my credit card.
On the other hand OpenAI has been quick to take money. Mistral too, got an account immediately and able to add billing details and they don’t even have a usage page so I have no idea how much I have burned through.
mistral
I just tried their API mistral-medium model for some coding prompts that failed miserably on anything other than gpt-4 and it handled them well. Just anecdotal, didn't so any proper benchmarking.
IIUC their open Mixtral 8x7 maps to the mistral-small model (which while impressive is still behind gpt-4).
I guess they want to keep their best model behind the API for now.
Anyway, I second that Mistral does stand a good chance to compete with openai
Microsoft isn't any better. Competition in this space is dead. Who ever is capable of posing a threat to both these obese rentier parasites will be targeted and obliterated given the cash mountain ranges at their disposal. Quality is not possible when competition is an illusion.
The only way competition in this space can be described as "dead" with all the free-license models bouncing about is that there's only one model (or company with models) that's actually worth paying for.
That would be the one that MS invested in.
Well, for now at least. At some point one of those free models will best it.
> there's only one model (or company with models) that's actually worth paying for.
Yes. Perhaps. That said "the model currently worth paying for" isn't the only possible business model or approach to "winning AI." It was a pretty incidental business model that unexpectedly got big.
I'm not convinced this is the final state.
AI models need users. There's having the most users. But, there's also having the right users. This isn't a linear race. Image generation. Language generation. Those work now. They'll get better... but it's still not clear what better enables. Best way to find out is by finding out what current models enable.
IMO, there's a lot more room here for bravery and creativity than beating OpenAI at whatever the race appear to be at this particular moment. The budget necessary to build applications is nothing compared to the scale of strategic investment being made.
Make the LLM chat/email client. Make the LLM project management software. CRM. tech support. All the cheap, obvious stuff.
Once upon a time, Google had a let 1000 flowers bloom approach. They did succeed in launching lots of pretty good products. They did not succeed in turning that tactical success into a strategic business success. That said, the tactic worked.
Why take the risk/effort/etc of winning the race to cloud services? Google have already lost one cloud race to google and amazon. Just build the apps yourselves. Find a framework. Do it. Don't want to run them forever? Then sell them off. Shut them down and leave the space open for startups.
Why futz around?
> AI models need users
Not when AI agents start doing their own thing, synthesizing their own data to go beyond what humans have conversed. The main thing keeping this from happening is data corruption. I don't know how that will be solved.
The only real competition seems to be Meta, which is another behemoth.
I agree, there are some minions whose valuations have skyrocketed because they happen to also try and compete in this very bubble-y market, but other than that I haven't heard any "those guys' LLM product is great!" comments related to anything other than whatever OpenAI or Meta have made public.
>right now they really really seem to be failing at it. Maybe the core tech in gemini is great. But it's nowhere
I think it's still an open field. Ability to make AIs into products and businesses... that's a lot of the game.
I agree about housing it all inside Google. Not necessarily because lawyers, or even because risk. There are advantages to both in house and outhouse approaches but... Google just isn't that good an in-house.
We do not know, at this point, what are the big businesses opportunities presented by recent AI advances. There's a lot of focus on near-future breakthroughs. IMO, many of the breakthrough products may already be possible to build now.
There's also an innovator's dilemma with the Google search mothership. One obvious area for LLM's is search replacement. "How does adwords benefit?" is not necessarily a question you want to have to answer, if your goal is search-replacement.
Everyone is in AI now and we are just finishing 2023. The next innovation(s) (dilemma(s)) could happen everywhere. Rephrasing Andreessen's quote: "AI is eating the world". This does not mean that we are close (or not) to AGI, it means that capital is flowing towards AI and AI has REAL applications. Just see what happened in Web3 with not so real applications (but uses). In the cybersecurity space I now see cybersecurity + AI, in the Web3 space I now see AI + Web3. It perfectly could be that Google is the next Altavista for AI.
What, in your opinion / experience happened with web3?
Web3 saw an incredible gold rush in multiple cycles and gigant amounts of money flew to the sector. The Web3 also experienced a gold rush because many incumbents went "magically" rich.
They're behind in selling for now but ahead in research - who is to say that selling access to your AI is really where value and growth will be in the future?
I would love to know if this really is Conways Law exposing poor reaction and innovation reactions.
> This is why Microsoft is winning
Spoken like someone who has never used bing chat. Its so terrible its funny.
MS is winning clearly, if the Azure $1bn revenue beat is mainly from AI/Copilot they are on track to recoup their OpenAI investment in 10 quarters, likely less.
Google has the ability to build something better than OpenAI but that's not even a certainty
Huh, not sure. I've found Bing chat to be actually useful compared to Google search. For some recent projects, Bing introduced me to a few very useful packages that I wouldn't have known even existed, if I had stuck to Google search. I've only seen Bing fail for my use cases when I drafted queries that I wanted it to fail: for example, asking some esoteric question on how to beat the Burmese 1st mission of Age of Empires in under 10 minutes. In most cases, Bing is actually a good enough replacement for a market research bullshitter, for cases where I'm looking for non-exact data.
They just don’t understand selling to businesses in a way that Microsoft or Oracle does.
The lengthy post boils down to this quote " Google could build the AI to win it all" but it is not guaranteed. I appreciate the context as someone who hasn't kept a close eye on Google's AI efforts, but Ben doesn't cover why Google has a right to win outside of search data. Interestingly, there is no mention of AWS and Amazon's wider efforts to create AI tools. There is hype around chat bots but what is the likelihood chat functionality is the premier AI gateway in the near future?
Edit: I thought about the topic more while wrapping up my work work. I am on the periphery of AI at one of the large US tech companies and we've placed AI bets along many of our existing products. Every day I run into this – "I manage xyz product, we plan to add AI to help with XYZ in 2024" or we added this chat functionality for "manual job to be done." I don't claim to have insight into the future on which of these solutions will be successful for their intended clients. The pattern I see is that AI can be quickly (relative) integrated or coupled onto existing software services. Is that the secret to AI? It will permeate through our digital lives either in small or big ways – but critically it isn't one AI to rule them all. AI micro services will act like intermediaries between humans and some end service.
It is like chocolate. Why not pair chocolate with [enter any food stuff]? You could hit a home run like chocolate with peanut butter or chocolate chip cookies. Now we have chocolate everywhere including drinks, but chocolate isn't required for a tasty result. And importantly chocolate isn't always a standalone dish – it can be though.
The argument is more than that -- namely that in addition to having sophisticated AI, Google also controls the OS (Android) and hardware (Pixel). Being able to integrate best-in-class AI at every level of the stack is a tremendous advantage. OpenAI can't do this because they don't control the OS, and will always need to go through an app. Apple can play since they control OS and hardware, but at the moment they appear pretty far behind in the AI aspect.
Is anyone really that far behind on AI given that the newest open-source models are closing in on GPT4?
I agree, Apple needs to wait and see what team offers a good alternative to OpenAI and buy them, I hope they do not buy and close an open source friendly team.
I'm expecting:
* Microsoft to do it for business users
* Google pixie does it first for personal use. Android phones might even momentarily be more desirable than iphone
* Apple eventually get there, for their walled garden
* Amazon etc fall totally behind
* And, years later, users of google docs are still waiting for basic AI help writing docs whilst google completely doesn't try in that space
I know that I'm part of a minority, but I'll avoid touching an "AI" enabled smartphone as long as possible. All privacy issues and tech issues aside, so far I have simply not found a single use case in which AI-tools would make my life easier.
As for privacy - Apple does most, if not all AI processing on-device. E.g. Siri understands voice commands when offline (although may work worse then I think).
A good on-device assistant would be super useful: - instead of searching for a setting, you just tell it to change a device setting - decent appointment/reminder setting ("remind me to invite joe to my next birthday party" -> and it knows who Joe is, and how to set up such a reminder) - all kinds of search ("what was that book about startups that someone recommended to me a year ago?", "open up a tracking page for that thing I ordered last week", "show me a photo of XX I took last summer") - managing e-mails the way old-school secretaries did - "any important messages?", or "reply to everyone I'm out of office, unless it's related to X"
As I said, while this is usefull for others, I have zero need for any of this. Settings? I know where to find those. Setting alerts? Use the standard calendar and timer apps. Internet search? Type it.
I make a point to not share my search history with anyobe, or any other data as much as possible. And I wont start doing so to enable some AI gadget I am more than happy to live without.
I think it will be more like:
"Pixie, reply to that email for me"
"Pixie, book that car service for me and remind them they promised a free brake fluid replacement"
"Pixie order all the ingredients I need to make my wife her favourite dish"
"Pixel, buy everyone in my family birthday presents"
I will 100% pay for that capability, and as someone that uses GPT4 to basically do my job for me, it really does not seem that far off. And I do think Google has an advantage here over Apple or anyone else, not just in Hardware, but in the enormous amount of data and information they have about both people in general, and specifically me.
I won't let any AI read my e-mails, let alone reply to them. I discuss car services when I drop of the car at the workshop, and the workshop confirms the details. I don't want an AI to know anyones favorite dish, and delivery services suck for fresh ingredients anyway. And I think about presents before buying them, that is a crucial bit of gifting things to people, especially if you care about them.
By the way, if I would use ChatGPT for my job, let alone letting it do for me, I'd be fired, worst case go to jail. And even if not, what would prevent my employer from firing me anyway if a basically free to use web service can do it at the same level I can?
I think you'll find yourself in the minority soon.
My grandma said, "I won't send any email through a computer, I prefer to write it by hand and have it delivered by a human being".
Although maybe it will be even weirder with AI assistants sending birthday greetings to other people's AI assistants and then your AI assistant summarizing who sent you birthday greetings.
The fact is the next generation moves on and uses new tools. I don't think AI will replace our jobs but people who are good at their jobs can, with the help of AI, outcompete people who are just good at their jobs but don't use AI.
The parent you were replying to may currently be able to automate what they are doing with the help of AI but I don't think that will last long as jobs end up requiring a mix of human and AI capabilities.
Replying to email is one thing - a pita to implement securely.
But reading? On-premise? You already allow basic algorithms to read your mail, for spam control and whenever you search through your messages. Having a better algorithm (as long as it's still on device) has no difference in terms of safety.
I can see it being wildly popular. Especially if it can edit photos.
Phone, remove the blemish from this picture.
Phone, make my hair slightly more red, and my waist thinner by one inch.
Phone, tell me who that guy in the background of my pictures is.
Phone, what time do I need to leave to get coffee on my way to work.
Those sorts of things.
AI would be much better at code _review_ than code generation.
AI would be much better at auto-completing whole sentences and paragraphs and suggesting rephrasing etc in docs and presentations, than at answering questions about the world.
But the race seems to be to answer questions about the world based on a 2021 dump of the internet, so ... :)
Oh, as a tool I see a ton of use: personally for image processing (sharpening, noise reduction) even if I can perfectly live without it, professionally I see a ton of potential use in optimizing planning (supply chain, production, scheduling, netwrok planning...) by proposing scenarios and supporting the people doing said planning. I don't need AI to search the internet for me, summarize a book (if it is worth or important enough for me, I just read it) or all the other stuff AI currently is doing.
Just to be that picky guy; > Android phones might even momentarily be more desirable than iphone
Most of the world is already there, and it's not momentary. iPhones only dominate the market in North America - Android has the majority basically everywhere else
As for docs, I can see them pushing for more Gemini integration depending on how M$ Copilot goes when it reaches saturation
> Just to be that picky guy; > Android phones might even momentarily be more desirable than iphone
> Most of the world is already there, and it's not momentary. iPhones only dominate the market in North America - Android has the majority basically everywhere else
Market share is not necessarily proportional to desirability. Cars make a useful analogy here. I think a Ferrari is more desirable than a Volkswagen, but Volkswagen’s market share is much higher.
Android is there because it’s what the cheapest devices use and most of the world does not want to pay even iPhone SE pricing. That’s relevant because the play described of making the Pixel lineup more compelling will hit the same problem if it’s tied to more capable hardware which is outside of the budget for many people.
OpenAI has close ties to Microsoft and we've already seen them integrate AI into Bing; MS may not be a player in phones but they make a major OS and some of the best hardware money can buy. I'm really not convinced vertical integration matters all that much, but if it does, they can do it.
Unsurprisingly, they already integrated OpenAI into Windows. You can ask it questions and pass it files.
>Apple can play since they control OS and hardware, but at the moment they appear pretty far behind in the AI aspect.
I disagree strongly with this, and it has everything to do with how we conceptualize AI.
I get in my car and plug in my iPhone. CarPlay immediately causes the Maps app to pop up and route me to my next meeting. I can say "hey siri, set a reminder to call Mr. Jones at 3:00" and she will gladly comply. If my buddy texts me while I'm driving to the meeting and asks if I'm free for golf tomorrow, she will automatically try to pin that on my calendar. I can throw out lots of examples here, but you get the idea.
Now granted, voice recognition in Siri has been pretty bad. She struggles with a lot of basic things, like putting on the music I request. But, there's no question in my mind that these augmented reality moments are where AI is actually making a difference in our lives and represent the actual business opportunity bridgeheads. In Apple's case, they not only already control the hardware (the phone, the watch, the earbuds, the tablet) but they've also figured out how to start bridging this into other hardware like a vehicle.
The impressive ML is ML that’s part of daily life. Even as I tap on this iPhone’s keyboard, button regions (tappable, not visible) slightly change in size depending on what it predicts the next letter would be. No one calls this “AI” yet it’s the same tech, and arguably more beneficial for the society as a whole than “AI” as a dedicated commercial service purpose-built to launder copyrighted creative works for profit (which is what AI is in the eyes of an ordinary person these days).
And yet the iPhone keyboard has gotten worse than it was a few years ago. I used an iPhone 8 years ago, and when I got an SE last year the keyboard was too smart, and so I disabled all the AI features.
Adjusting invisible tappable key regions, swipe typing, etc., these are based on ML even if you disable predictive typing and other settings though.
Conversely, I've recently had Apple software mess up in a whole bunch of different ways:
• Maps doesn't understand that I don't own a car, defaults to driving sometimes
• Autocorrupt rather than autocorrect
• Calendar suggestions only work for the simplest of dates, so it suggested an event for the wrong month
• One case where it seemed to think the only timezone in the world was California
(It's not all negatives: for me, Apple has the least wrong voice transcription AI, and I do like their computational photography and definitely the ability to select text in images and Safari's website translation — but even then I don't think they're way ahead of the rest with these things, and website translation was definitely behind).
I suggest you try the assistant features on a Pixel 8 Pro. They have all the features you mentioned (except the creepy eavesdropping golf one) and the interaction is miles ahead, especially the text to speech.
I do not have any Android products, I only use Google Assistant through my Sonos speakers. Do you know if that code is different from that on the Pixel? (Because my experience of Google Assistant is long from good/useful, it struggles with basic tasks, I have to overpronounce to be sure it can differentiate “lights on” from “lights off”, etc etc.)
I don't know, because I haven't used Sonos for this purpose. But it's easy to imagine that a mobile handset with an array of microphones that you tend to hold near yourself would be more suited at the hardware level.
AwS is just a reseller of 3p LLMs at this point (via Bedrock). If you don’t believe me try their 1p LLM (Titan)
That may be true but consider that AWS was “just” a reseller of Intel servers a decade and a half ago (remember people saying S3 was too expensive and hard to use with tons of competition, too?). Integration matters and if they do a good job of making it easy to build into applications I’d easily believe that the path for a lot of businesses will involve a partner they already have extensive relationships and experience with.
S3 was actual innovation that AWS led the market in. Apples to oranges comparison.
EC2 was innovation, but many people didn’t see it. S3 was more distinctly new although still not the first object store or even object store over HTTP as a service, but if you were there a LOT of people did not share that perception and said it cost too much, would never scale, or required too much software to be modified. I’m not saying that to diminish the importance of S3 but simply that we often wildly overstate the importance of being first to ship an idea versus easy to work with, low-drama for operations, etc.
Ok we can both agree tho that AWS isn’t the platform leading LLM innovation. That said I’m not disputing that AWS will make money with Bedrock and enterprise contracts
I understood the crux of the argument to be that the holy grail UX is having an always-on AI assistant you can have a spoken conversation with. He says that Google’s combination of hardware, Android, loads of data on the user, and good enough AI chops might be a combination that makes an assistant so helpful that it would get people to switch away from iOS to use it.
Intelligence makes almost everything better, so maybe salt is a better analogy than chocolate (especially once the hallucination rate of LLMs becomes more acceptable).
I like your chocolate argument and will steal it for future use.
The one part of this thesis that I pretty strongly disagree with is the idea that people would prefer to have long, meandering voice conversations with an AI, compared to text.
Just look at anyone under the age of 25 (35 maybe?) They can easily have long, meandering conversations with actual humans using voice, yet I see them go for text 9 times out of ten. As someone on the backside of middle age, I often find it pretty baffling. I like the succinctness of text when I need to send a quick update or ask a short question, but I normally always call someone for an in-depth conversation. But I'll see my nieces text back and forth with friends for literally hours, sometimes getting emotionally worked up, and I'm thinking "OMG, just pick up the phone to your ear and just talk to them."
But I think the reason people prefer texting is the same reason most people still prefer typing, despite tech that, these days, could easily transcribe with great accuracy. At least for me, typing frees up my brain to actually move faster. When typing, I can think about the next phrase or sentence. When talking, I find it much more difficult to "think ahead", so to speak.
So I'm really skeptical that voice interfaces will be "the wave of the future". Sure, I use OK Google a lot, but basically for the same sets of commands as everyone else (What's the weather? Set my alarm. What's next on my calendar? Etc. etc.) Occasionally I'll ask it "search-like" questions. Perhaps I suffer from a dearth of imagination, but I just have a hard time believing long voice conversations with a machine are something most folks would want.
Some potential "costs" of voice/video: it's (often) exclusive to one person, it's immediate, and it's easy to expose your own emotional state. In an emotionally fraught, developing situation, perhaps your nieces want to take it slow, check what their friends think, not let on that they're upset, etc.
This is all just speculation, I'm not really a texter, but I do find it interesting when limitations might turn out to be features.
Typing also has the advantage that several people doing it in the same room at the same time don’t interfere with each other. A train carriage full of people texting is considerably less annoying than the same carriage full of people chattering away with their voice assistants
> A train carriage full of people texting is considerably less annoying than the same carriage full of people chattering away with their voice assistants
Can't be worse than a bus full of people recording voice messages aloud and listening to them on loudspeaker which is daily life in say China. I would be lying if I said it's not maddening
> At least for me, typing frees up my brain to actually move faster. When typing, I can think about the next phrase or sentence. When talking, I find it much more difficult to "think ahead", so to speak.
I wonder if this would change as we got more familiar with interacting with a voice AI. I think a lot of the extra brain power that goes into talking versus typing is the assumed need to keep talking at a constant pace because that is what a human listener expects. If we were more comfortable with pausing while talking and not feeling like we need to always know what to say next as soon as we get to the end of a word, it might not take as much brain power to speak to a computer anymore.
> Just look at anyone under the age of 25 (35 maybe?) They can easily have long, meandering conversations with actual humans using voice, yet I see them go for text 9 times out of ten.
Where do you live that they prefer texting?
My acquaintances in that age range love sending voice (and fucking video, yes yes) messages. And I mean not intimate chats with close friends or relatives where you want to hear the voice and see the face. No this is just how they convey info. Including some people from UK.
Young people text because it’s considered bad etiquette to call unannounced.
No such issue with robots.
That's definitely not the only reason. Again, I've known folks to have long, drawn out, back-and-forth conversations over text, certainly not just a "Hey, can I call you?" text intro, which is what I do.
Talking also bothers the people around you, and can tire you out if you do it for hours.
Not all people experience thought the same way. Some people (like me) find it no problem to compose a sentence in advance while speaking some previous words. The speaking doesn't conflict with the composition. Some people can count, do math, or keep time silently in their head while reading. Other people can't do that because their thoughts and language processing are linked differently somehow.
For me, I prefer to dictate to my Android phone in contexts where most people choose to type, such as in a short message. It's faster than typing on-screen, and the way I think I am able to compose ahead in a way I can't do while typing. The dictation is so good these days that there are relatively few mistakes and ambiguities to correct, and the UIs for doing so have become easier.
Big companies drag themselves down – not because they can't innovate – but because they can't go "all in" into innovations which compete against their real cash cow.
An AI assistant fundamentally hits at Google's gut: Search ad revenue. Yeah, Google can make dozens of AI demos. Are they truly ready to put Gemini as the default interaction model, pitting it against their search? Can they give super accurate answers to questions, without the potential of 3 sponsored results on top, in such a way that it would make their search obsolete?
Shareholders "wish" to see competitive demos, to not be left behind, etc. But are shareholders & market ready for a 10% balance sheet revenue drop that comes with making an AI assistant a go-to product instead of search?
AI in Google could end up getting crippled, not by design, but by the environment and "thou shalt not touch search revenue" constraints under which it operates in.
It would be interesting if Alphabet (and Alphabet's moonshots) totally distances itself from Google instead of being an internal structure sharing the same tickers, and treat Gemini as an all-out cannibalistic competitor to Google.
Their ad revenue is a local optimum. They could reach new highs with alternate business models but only at the cost of short term reduction in revenue from ads. And this is of course risky. So, they keep on dancing around the topic without ever committing to anything.
An additional problem they have is that running AI at scale is super expensive. Even for Google. Add to that legal risks, pressure from regulators in different countries, privacy concerns, copyright issues, etc. and you get a whole lot of risk and friction. Add to that the usual organizational infighting and politics and you pretty much have a company incapable of doing anything.
> They could reach new highs with alternate business models but only at the cost of short term reduction in revenue from ads.
Yes. Long term, if you control peoples attention and their flow of information, you will almost certainly figure out a way to control their pocketbooks.
> An AI assistant fundamentally hits at Google's gut: Search ad revenue.
I understand this argument, having made it multiple times in the past - any development at google that threatens the cash cow will, ultimately, go nowhere. Why, after all, would they spend money developing something that reduces their income.
After a little bit of thinking about it, I offer an alternate future for google+AI: the AI improves their ad relevancy to such a degree that it makes google search more valuable, not less.
With all the data they have on each individual using their search, it is not inconceivable that that data + an inferior model beats out a superior model that has less context.
> the AI improves their ad relevancy to such a degree that it makes google search more valuable, not less.
More valuable to who? In this case it seems like more relevant ads increases the value of Google Search to advertisers but not necessarily users.
If, as a user, I see ads that are more directly targeted at me, that is a negative value in a search engine. I never want targeted ads. I only want to see what I searched for.
That said, for the small minority of users like me, Google may never be able to overcome their “advertising company” label. And that label is an instant black list.
> > the AI improves their ad relevancy to such a degree that it makes google search more valuable, not less.
> More valuable to who?
To the paying customers - advertisers.
> In this case it seems like more relevant ads increases the value of Google Search to advertisers but not necessarily users.
So? Google doesn't care about its users, and hasn't for a very long time now.
Google became profitable purely because they delivered the most value to advertisers, by way of having the most eyeballs. More relevant advertisements are not going to reduce the number of eyeballs that google sells.
> If, as a user, I see ads that are more directly targeted at me, that is a negative value in a search engine.
And those users will be lost in the noise, a statistical rounding error, if that. Losing unprofitable users helps Google become even more profitable.
"Are they truly ready to put Gemini as the default interaction model, pitting it against their search?"
Probably not, since they have not figured out, how to integrate the ads into the AI.
What's always frustrated me about Google is that I would pay a substantial amount of money on an ongoing basis for the capabilities they could provide, but I'm absolutely not willing to pay what they're asking and what they're asking is tanking their products.
Google as a products and services company - Google with Apple's business model - is something I would've been a happy customer of for the last decade or two easily. Google as an ads company is an entity I go to great efforts to remove from my life.
There's also the very, very big (and very hidden) opportunity cost of always incurring the danger of having your Google account getting nuked by Google itself, with almost no possibility of recourse, for some small payment inconsistency related to any of their products, no matter how big or small.
That's why I personally would never pay Google other than for very basic and needed stuff like GMail, I wouldn't want my email account nuked and my life momentarily turned upside down because of some payment misunderstandings related to their streaming service, let's say.
I think that’s the same problem: once they did the acquisition with DoubleClick, everything solidified into “we don’t have to do customer service, our customers are advertisers” at the same time they were trying to push everyone to turn more substantial parts of their lives than search over to them.
I was thinking something similar earlier.
I am huge fan of youtube and the incredible content and creators on the platform. If it was not owned by Google, I would probably happily pay 50 bucks a month for an ad-free and tracking-free experience.
But instead I've mostly left the platform. I don't want the company with eyes all over the internet to also know exactly what I watch and when. It's creepy to have a single company know exactly what you did/read/watched for most of your waking hours.
Aside: excruciating one hundred paragraphs about stuff he said previously, recently, or in another lifetime, before getting to the actual topic and title subject of the article. Geez. I don't need articles to be transcripts of podcast-esque rambling. Get on with it!
This is a very common pattern (and criticism) of stratechery.
This pattern of writing isn't necessarily a criticism; it can also be beneficial. Visit infrequently (thus saving time to use for other worthwhile activities), and you'll receive a summary of his thoughts over a decade, all in a well-organized and smoothly flowing text. Additionally, if you prefer to skip the historical recap and setting, you can simply start reading from the end.
I've allways found it valid and helpful background
> Yes, Android has its advantages to iOS, but they aren’t particularly meaningful to most people, and even for those that care — like me — they are not large enough to give up on iOS’s overall superior user experience.
That's not the real question though - the real question is whether people find iOS compelling enough to pay a premium for using it and put up with the ecosystem lock-in. And the user experience is only superior when you're used to it, as a longtime Android user I regularly get frustrated when having to use an iOS device...
> the real question is whether people find iOS compelling enough to pay a premium for using it and put up with the ecosystem lock-in
It’s only paying a premium at the lowest end of the market - if you’re buying a Pixel, you’re paying as much or even more than someone buying an equivalent iPhone. The ecosystem switching cost is real but given how most Android users I know complain about app developers favoring iOS, I’m not sure how substantial that is.
In summary. At some point an ai assistant will replace 80% of the functionality/apps offered by iOS & android - it will finally become the new OS (as we all knew from watching her).
This is bad for Google as google’s main feature is search. Ai assistants will replace search and kill the main revenue stream which is ad clicks.
Google’s secret plan is to eventually canabilize it’s pixel phone with an ‘agent first’ device to try to beat Apple, MS & OpenAI with a horizontal offering that is significantly better than the fragmented world of ChatGPT &iOS & azure.
But it’s worth remembering and I think the article fails to point this out. Agents will still ‘recommend’ things, you ask them to book you a flight - they still have to recommend a couple of options out of many - the agents need to decide and ultimately that decision is the same as choosing who to place at the top of the search results page - whoever wins the agent wars will win a significant proportion of Google search revenue as referral fees AND a significant proportion of IOS App Store revenue.
It really is winner takes all.
People act as if there are two companies (Google and Apple) and two business models (ads and selling devices). In fact Google is perfectly capable of charging a subscription in a context where ads don't make sense, like Drive, or even as an alternative, like Youtube Premium. Many people would pay a subscription for an assistant that's actually good, and of course having such an assistant integrated with email and calendar would be extremely valuable.
Google does have non-ad revenue but it doesn’t seem to be enough to affect the culture: Drive customers get the same level of support as ad-supported search users. I’d like that to change but having used GCP I’m not sure that can happen before they have a better CEO.
How does Drive customer support compare to Dropbox?
Imagine you lose access to your account, with what company do you think have better chances recovering it?
Now imagine you get some very bad data loss bug affecting just your company account. What chances are there you will ever have your problem looked at by Google? I totally believe it's 0% unless you get to HN frontpage.
It's not like Dropbox have best support ever, but I totally believe that getting a living human to look at your problem would be so much easier with them. With Google it's just impossible.
We should consider the fact that AI ain’t a company, it is a feature, the same as highly intelligent people per se don’t earn tons of money.
Paradoxically we are in a phase of technological stagnation. Cloud movement was/is maybe the latest paradigm shift for a couple of years to decades to come.
AI will accelerate features and won’t mark a product category itself, same as gifted people are capable of outperforming others in intellectual fields, but don’t necessarily need to.
The mundane stuff like generating text is what is AI paradoxically needed for. (Imagine that, that humanity’s greatest gift is now outsourced.)
This is what makes AI so hard to grasp. We know what it is, but it is hard to applying it in concrete business contexts. “OK Google analyze my company with 100.000 workers” won’t happen soon. Society needs to take care of possible side effects first.
Oh I am pretty certain someone™ will put the whole personal record into an LLM on someone elses computer (aka the cloud) and ask who needs to be fired and come up with a reason. Then they will fire them. Then the reason will turn out to be bullshit.
On the other side, there will be a ton of cowards who will say the computer told them to do X, where X is the thing they wanted to do all along. And the neat part is that you don't even need to ask a computer for that, you just pretend you did. And in the end if Xbturns out to be a shit choice, you can use the famous software error-excuse and be off the hook immediately, because apparently liability ends there automatically or something.
I dunno, man, I can’t be arsed to trust Google to not scrap their thing a few years in anymore.
They will make something remotely nice, half ass the final 20% of it, leave it to rot, complain that there have been only X billion dollars of profits and not Y, and kill it off.
How they are not sunsetting Search yet, I cannot fathom.
For anybody else expecting Google to start shooting stuff at the moon: don't bother reading, it's just about more AI.
True, but today the term "moonshot" can only be about AI.
> After all, if a user doesn’t have to choose from search results, said user also doesn’t have the opportunity to click an ad, thus choosing the winner of the competition Google created between its advertisers for user attention. Google Assistant has the exact same problem: where do the ads go?
This is Google / Alphabet's biggest problem. Unlike Microsoft that has Enterprise, Gaming, Cloud and multiple other 10B+ businesses that don't rely on ads, Google has only cloud.
And Google Cloud is a distant third where they actively sabotage their own success by not caring about customers. 90% of Google revenue comes from ads. From Search, youtube, Adsense e.t.c
If they build an AI that gives exactly what the user wants without ads, Google's stakeholders would fire the CEO the next day.
My perception is that Google is in AI race to be relevant and capture the talent, but they don't want to actively put AI anywhere near their products that would eat into their ad revenue.
Google Brain/Deepmind produces great research, they have the most number of published papers, by a big margin, but Google is not the place if you want to work on productionizing AI.
Top AI Researchers probably make the most money being at the big tech labs - DeepMind / Open AI / Meta FAIR. I've heard comp being into the 10s of millions per year.
> Google sells its own phones which could be configured to have a conversation UI by default (or with Google’s Pixel Buds). This removes the friction of opening an app and setting a mode. Google also has a fleet of home devices already designed for voice interaction. Google has massive amounts of infrastructure all over the globe, with the lowest latency and fastest response. This undergirds search today, but it could undergird a new generative AI assistant tomorrow. Google has access to gobs of data specifically tied to human vocal communication, thanks to YouTube in particular. In short, the Gemini demo may have been faked, but Google is by far the company best positioned to make it real.
How does he square any of that with the rest of the preceding article? Google having a technical advantage on paper and still fumbling the ball is their M.O. for a decade. Their ecosystem of devices, even just sticking to the Google-branded hardware, are inconsistent crap. Take a simple, first part app like Google Calendar and look at how sloppily and poorly it works across Pixel phones, Android Wear devices (including the Pixel Watch) and Home devices. Google sucks at building polished, consumer products.
Google is by far the best company positioned for AI assistants? Is this guy forgetting about Apple, the company that has a much stronger hardware ecosystem and that will demonstrate this in the coming months with a huge hardware launch that only they could do, leveraging a huge base of iOS apps, AirPod users, etc. Apple is building a network of personal hardware devices with a weak but well-integrated personal assistant. In 1-2 years, Siri will get a massive improvement based on recent advances in LLMs and it will be 80% as good as Google’s then probably thrice rebranded AI, but it will be 300% more polished and install automatically on hardware people actually like to use.
> Their ecosystem of devices, even just sticking to the Google-branded hardware, are inconsistent crap. Take a simple, first part app like Google Calendar and look at how sloppily and poorly it works across Pixel phones, Android Wear devices (including the Pixel Watch) and Home devices. Google sucks at building polished, consumer products.
Uh, I think you're living in another universe. I have all of those products and I have no issues with the Calendar experience. What use case do you think works poorly exactly?
Instead of turning Pixels with Assistants into iPhone killers and becoming Apple, author should have looked straight at Microsoft Copilot and how a paradigm shifting LLM in G Workspace could kill Office. The Pixel is still only sold in 21 countries, scaling to Apple level would be brutal.
Also that idea Google will build such a magically superior AI to every else is a bold claim. Can it make a competitive one? Yes. Can it make one so good that it makes every other AI obselete and everyone trade their iPhone for Pixel? I think that's unlikely given the level of competition
Undertaking that kind of focused, bold, visionary attack would first and foremost require to have a CEO with the same kind of attributes:bold, focused and visionary.
Unfortunately, ball-less wonder milquetoast Sundar has none of the above attributes.
He was hired 10 years ago not to rock the boat, but to stay the course, a strategy that requires a CEO with all the exact opposite attributes: meek, boring, visionless and all over the place.
I was under the impression that the data google has isn’t too valuable for ai training, as quality is so important.
If textbook quality data is needed, then we are basically limited by the current best LLM’s ability to create synthetic textbooks.
Or perhaps this is a path Microsoft is trying (and presumably openai) due to a lack of good non-synthetic data.
Google Books would probably be useful, although I don't know if they're able to take advantage of it.
Is the phrase "OK Google" very hard to pronounce for anyone else? Specifically the G after the Kay.
You can say "Hey Google" instead, like "Hey Google, what have you guys been doing this whole last decade?", to which it'll answer, "Okay, playing guy music from last decade on kitchen speaker".
Does anyone know if it is actually possible to completely disable the voice assistant on Android? I've changed every setting I can, and no matter what, if I plug in headphones, it'll prompt me if the connection goes loose.
Good read.
> Google’s collection of moonshots — from Waymo to Google Fiber to Nest to Project Wing to Verily to Project Loon (and the list goes on) — have mostly been science projects ....
...a car service rather far afield from Google’s mission statement “to organize the world’s information
...What if “I’m Feeling Lucky” were not a whimsical button on a spartan home page, but the default way of interacting with all of the world’s information? What if an AI Assistant were so good, and so natural
So... I think we should distinguish between "moonshot" and "silver bullet." One is a big, difficult goal that can be a approached with lots of determination, resources and such. The other is a future breakthrough that just fixes everything.
Google has always struggled making (great) technology and concepts into products, and products into businesses. The biggest miss, IMO, was cloud. Msft & Amzn relative successes highlight where google isn't strategically strong.
Waymo might be the biggest investment (probably >$100bn risked). Cloud is the large business category that actually exists. Google should have been here, considering where everyone was circa 2008. Google had the tech, the concepts, even the products. It wasn't effective at making that a great business.
The "OK Google" assistant story tell objectively, because no one else has done a great job with voice interfaces either. That said I think it demonstrate the difficulty of going "concept to products."
IMO, the problem with voice assistant has been a problem of imagination. Voice is a UI paradigm. What are the key use cases, where this new UI paradigm is powerful? They never invented it.
Anyway... I think the strategic logic is flawed... if that is indeed the strategic logic It's "singularity thinking." An expectation that version N+2 makes version N=1 obsolete. He who attains the GPTn, owns driving, personal computing, etc.
>The potential payoff, though, is astronomical: a world with Pixie everywhere means a world where Google makes real money from selling hardware, in addition to services for enterprises and schools, and cloud services
So this is what I mean. A "moonshot" would be defining these and going after them with real big intent. Not one that considers everything side effects of some big breakthrough that makes all linear approaches irrelevant. Voice UIs, even self driving, whatever wing's mission is.... these aren't impossible ideas even with current science & computing power. They're just hard. Requiring imagination. Risk. Vision. Strategy.
Drones are a method. Delivery is the task. Human-like drivers are a method. Transport is the task. Voice recognition. LLMs. These are methods. Not tasks. If you're doing silver bullet, it's nice to avoid defining the task. If you're doing moonshots, you want to be brutalist in defining the task.
Google no longer seems relevant to anything much apart from what it’s already doing well… YouTube, search, android, maps.
Can you provide some context about why this is relevant to the article, which posits that Google is uniquely positioned to win the AI space and hence the future of computing?
The potential payoff, though, is astronomical: a world with Pixie everywhere means a world where Google makes real money from selling hardware, in addition to services for enterprises and schools, and cloud services that leverage Google’s infrastructure to provide the same capabilities to businesses. Moreover, it’s a world where Google is truly integrated: the company already makes the chips, in both its phones and its data centers, it makes the models, and it does it all with the largest collection of data in the world.
I don't know enough about the AI space to comment on it specifically.
But this distinctly reminds me about articles around about the time of the first iPhone release which posited that Microsoft was in a unique position to ultimately win the smartphone market.
After all it was dominant on the desktop and Windows Mobile already had a decent (albeit not dominant) marketshare in smartphones at the time.
The theory wasn't wrong and still isn't wrong. The problem is in the execution.
The two things are related. That's the insight of The Innovator's Dilemma. They cannot do it.
> which posits that Google is uniquely positioned to win the AI space and hence the future of computing?
They have been "uniquely positioned" as you put it to win the AI race for the past 15 years now.
And they have done exactly zilch (other than produce an OpenAI-wannabe fake video with Gemini a couple of weeks back).
I see no reason why that will change in any way.
Astronomical payoffs is not something Google is culturally capable of anymore, because they require risk taking, and strictly no one has any incentive to take any kind of risks at Google anymore.
But it doesn't seem to be true - there's more of a moat around search than AI it seems, and the recent trend is people replacing the Google search app on their phones with Perplexity.ai.
You currently still need search too, for AI's to access data newer than their training set, but I wonder how that's going to evolve over time ... Could see AI's being updated increasingly frequently and tapping direct into news sources (not via search) for news/sports/etc.
Because Google has burned it’s good will by cancelling so many projects.
Yet another new Google project is simply met with “ok, but when will it be canceled”.
Chrome, gmail, google docs/sheets, google meet, google drive, cloud, pixel, chrome book, nest, fiber, waymo
Waymo?
Way too early IMO. Think about something like Google Fiber. It looked extremely promising, but the inability (or lack of desire) to expand it left it as mostly irrelevant. With Waymo they also have some major issues to work out as they're bleeding billions while charging as much, if not more, as human driven cars charge. It might end up getting worked out in the future and turned into something neat, or it might just eventually find itself on the ever-growing scrap-heap.
aghem, Stadia, aghem
> it’s already doing well
> search
many people disagree..
It needs to be at least many hundred million disagree to matter.
> hundred million
this is about how many people now going to ChatGPT while were going to Google year ago for similar searches
At least in my experience, ChatGPT uses Bing a whole lot. I wonder what the stats for Bing usage are now compared to Google since ChatGPT has been out for a while.