I wish the folks who clearly do not like Google would just not use their products instead of spamming every thread about how they will kill the product, true or not.
——
Anyway,
It’s not clear which model they’re using for this. I assume whatever Bard is using, but who knows. This is relevant because depending on the intended experience the latency will matter.
Overall it’s not a bad idea, but I do wonder what the monetization path will be for Google. I imagine this will be part of workspace. Perhaps they will add more tiers to include these offerings.
I wish they shared a bit about how this will be differentiated from Bard. Is this simply a new front end to Bard? It’s really an open question. I haven’t seen many products that use LLMs that are better than the prompt response UX.
The most interesting thing about this blog post is the “source grounding.” I’m curious if there’s actual engineering behind it, or is it prompt tweaking contextualized behind the scenes on a given doc.
People are wary about Google product launches because of a well documented[1]
pattern of corporate behaviour[2] when it comes to either the project's planning and execution, resourcing, or long-term support.
You don't have to agree with those comments, but they are legitimate concerns being expressed by people who have been burned enough times by the company that it has developed a degree[3] of notoriety[4] for it as a result.[5]
To be fair, there are a lot of low-effort comments here to this effect. I agree with you and will happily upvote the first well-written comment pointing out Google's self-inflicted reputational damage, but the myriad one-line jabs aren't adding anything.
I'd normally feel the same as you about the "one-line jabs" but in this case, IMHO they help to underline the fact that Google really did screw a lot of people over.
Like I mentioned, don’t use google products then. What’s the value of spamming this all of the time? The validity of the concern isn’t the problem, it’s the fact that it’s irrelevant to the content being posted.
How is it irrelevant? If a new project/product is being hyped by a company with a well documented history of hyping then abandoning projects and products, it's objectively a relevant comment.
For example, if there's a post about the new Threads app, and someone comments about Meta's record on, say... security, data privacy, moderation or support of Instagram, WhatsApp, and Facebook, or something about their failed projects like Beacon, Sponsored Stories, Creative Labs, and Facebook Credits, those are likely going to be relevant comments because the corporate culture and past record on projects and products reveals something about what we might be able to expect from Threads.
It's not irrelevant, it's just incredibly tiring reading the same comment over and over again on every thread that remotely mentions Google launching a new product.
> For example, if there's a post about the new Threads app, and someone comments about Meta's record on, say... security, data privacy
- Ctrl+F "security". Exactly 0 comment hits (on the first page).
- Ctrl+F "privacy". There are three hits, and two of them just mention "privacy policy" without making a comment about Meta's previous privacy violations.
- Ctrl+F "moderation". Exactly one hit about how Meta has historically been bad about moderation.
Let's look at the current post about Google launching this product (NotebookLM). Ctrl+F "kill". There are at LEAST ten comments about how Google will kill this product off soon. It's not comparable to the discussion re. Meta and its history.
Google killing off products is table stakes at this point. Imagine every time you clicked on an HN discussion about Python, you had to read ten comments about how Python is slow and if you want a faster language you should use Rust. "Python slow, Rust fast." Yes, we get that already. Can we just skip the memetic comments about "Google bad cuz they kill stuff" and read discussion with more meat?
I meant the Threads point as a hypothetical example not a literal one. I see annoying and repetitive comments on HN all the time. I don't have to upvote them or engage with them at all. I can merrily skip on by. Unless they're being abusive or threatening, I don't see what the problem is?
Clearly enough people think Google's record of hypeware-to-abandonware cycle is worth raising again and again because it's a frustrating aspect of the company's 'innovation' process and HN is a fairly influential place to keep that issue alive in the hopes that it might lead to some change or at the very least inform people who are considering investing a lot of time, energy, and capital into depending on such projects. It's also not exclusive to Google[1], though perhaps the one most notorious for it.
I've also found those "Python slow, Rust fast" threads genuinely useful because they prompt people to report on progress in Python on that front, or indeed even how the two languages can work together to play to each other's strengths.[2]
Eh, as a ride or die google reader fan, when there's six "DAE google reader" comments in a row, you all just read like lazy comments looking for some karma and detailing any (potential) interesting discussion before it starts. That's my main draw here, so definitely a reflexive down vote from me.
To warn users so they're aware that the lifespan of the product they're using may be very short, and therefore they may lose something they put a lot of investment in. It's empathy.
Nobody reading HN needs to be made aware of it, and those are the only people you'll be able to reach with a HN comment.
The thing is, these repetitive comments get spammed dozens of times on every Google-related post. They could be posted without reading the submission. Many of them probably are in fact probably posted just based on the title (seeing how many people posted basically exactly the same comments into a submission today that wasn't even about a product). These comments completely crowd out any other discussion just due to sheer volume.
There's 90 comments on this thread at the moment. Maybe like 5 are discussing anything about the article itself. The rest are just totally predictable and generic comments, which is something the comment guidelines specifically ask to avoid. (And yes, I'm now also participating in that.)
Now, fair enough. Maybe there really is nothing to be said about the concept of the product. There's definitely not much to be said yet about the execution given nobody here will have been able to use it, so that definitely limits the scope of meaningful discussion to be had. But since the spam won out, we'll never find out.
If there's nothing of substance to talk about, we could just talk about nothing and let the obviously boring and uninteresting submission die in silence.
And if you need to warn your family and friends not to use this product because you think it will be cancelled too soon, you'll probably reach them more effectively via methods other than HN.
The HN Guidelines[1] also say that commenters should:
"Respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith."
Just because people have posted similar comments about Google's penchant for killing projects, doesn't mean that they are "spamming". Consider, instead, that it may just reflect a groundswell of opinion borne out of bitter experience and disappointment about the company's record and their genuine concerns that this is yet one more tombstone in the expanding Google graveyard.[2]
Sure, let's take it as a given that everyone who posts wholly original messages about "lol this product will be canceled" are doing so purely because of the harm they have personally suffered due to a product cancellation, and they just missed the previous messages on that subject. So what? It's still turned this thread into a wasteland and prevented any productive discussion.
Posting in good faith isn't a free pass to post off-topic, repetitive ir low-effort stuff. Just think about what the outcome of that kind of policy would be if applied across the board.
For example, a lot of people have very strong opinions on divisive political issues. Should they be picking submissions that are tangentially related to those issues and hijacking the comments section for discussion about that subject? No, of course not. And I'm not saying that because I disagree with their hypothetical political opinions. I'm saying it because if that becomes the norm, we can't have a discussion about the actual thing that 200 people thought was interesting enough to upvote, because all the discussion will be a retread of that political issue.
It's as if Rust evangelists posted comments about "this bug would never happen in Rust" in response to a technical deep dive to a security bug. Yes, I'm sure they'd be right. Yes, I'm sure they'd be posting that in good faith. There's even a small chance they'd be posting it because they think there's still HN readers who haven't heard the good word of Rust. But it'd also be a shallow comment, variations of which would be posted hundreds of times, and totally generic in that the exact same comment could have been posted to any of a hundred other submissions about a security issue.
> So what? It's still turned this thread into a wasteland and prevented any productive discussion.
How exactly has it "prevented any productive discussion"? I've seen plenty of comments about the merits and limitations of its Bard foundation, the design choices, the potential business model, etc.
The example you give of political opinions doesn't feel like a fair comparison. We're talking about the company's record on supporting its projects in the context of a new project announcement not a divisive socio-political controversy.
A better example might be: If Red Hat announced a new open source initiative, would it not be relevant to bring up concerns about its commitment to the new FOSS project in light of the impacts its recent announcement about CentOS is having on other companies and projects that depended on it?
My favorite were comments on a post about Google Zanzibar, which is an internal Google infrastructure thing that they published a research paper about. Other companies started adopting that method, and there were HN comments warning about Google killing products. I don't think Google is going to cancel a research paper!
If a company behaves badly then I think it’s warranted to bring up their bad behavior whenever that company is the topic of discussion. In the U.S. we mostly have regulatory capture and so people need to fight back or otherwise hold companies accountable.
I don’t know if Google qualifies in my mind for this sort of action but companies like BP, Exxon, and others do.
If that's true, then you're only exacerbating it by attacking "said thread".
Furthermore, threads raising the issue of googles seemingly erratic depredation of their products and threads discussing the product itself aren't mutually exclusive. You can do this neat thing called scroll down, and not read these threads if they bother you so much.
That's the thing I don't get. Why don't people just ignore the comments they don't like and don't want to boost? Feels like some kind of Stack Overflow-esque over-persnickety moderation syndrome.
The value is that it provides context around googles substantial issues with long-term support for their products. So anybody who's new to hacker news or even new to the tech world can see that this is still an ongoing issue, and can avoid getting burned like so many of us before.
And given the forum, I like to think that there's probably a fair number of Googlers who see these comments, so it's always on the forefront of their minds when they may bring it up to their managers and their managers managers.
It’s not spam to some when they are reminded that google never was, a “do no evil” corporation.
Google has tarnished itself. Yes it’s the biggest. But this is because of a well documented cycle - and when evidence of this horseshit pops up “yet again” it’s important to remind everyone to not get too attached to anything google.
Bard isn't a model, it's a product. Bard was launched with PaLM originally, the same model that the one employee claimed was sentient. At Google I/O they announced PaLM2 and switched over to Bard being backed by it, while also making Bard open to the general public on the same day.
Google is currently in the process of training Gemini, which is positioned to be their GPT-4 killer, as it's multi-modal, and presumably quite a bit more powerful than GPT-4 is. PaLM2 is more analogous to GPT-3.
NotebookLM is a very different product than Bard. Bard has a vague understanding of the information on the internet, and NotebookLM will, presumably, have some level of understanding of your personal documents. It will almost certainly be using PaLM2 until Gemini is ready. That will be neat because Gemini will be able to look at images and read visual charts in your documents. PaLM2 is only able to comprehend text.
Right - to clarify I meant the model used by bard. I was too lazy to see what bard currently uses since it’s transparent to the user, to your point. Multi modal models being used would be killer.
But the monetization thoughts remain, this stuff isn’t cheap.
> I wish the folks who clearly do not like Google would just not use their products instead of spamming every thread about how they will kill the product, true or not.
Me, upon reading something about this: "Oh, this sounds interesting, I would like to have a go at something nea--"
observes it is released by Google
"Well, never mind, I'm not going to invest the energy into it until it reaches its third birthday."
Google could simply stop taking a scythe to products at a rapid clip and shed their reputation if they want to benefit from not being slagged for it everytime they release something new.
While I understant the feelings about Google wrt product longevity, I fail to see why that would prevent you from using this specific product as the 'investment' seems to be you uploading a Google doc and the payoff is instant.
I feel like most social media would be much more usable if there is some kind of browser extension that automatically clusters the comments and hides the same opinion/sentiments that have been repeated a dozen times in the same thread(you can label the clusters with comment counts of the same topic).
edit: Actually before someone submits a comment, they should be informed that how many people have said the exact same thing, like a repost warning.
> I wish the folks who clearly do not like Google would just not use their products instead of spamming every thread about how they will kill the product, true or not.
It seems weird that you think these people do not like google. Then why would they care? The uncertainty of the "google graveyard" only affects you if you do like and use their stuff. And fans in particular can be very gloom and doom about anything, e.g. see any esports game update and how it will completely break everything.
I want to like Google’s products, and used to be quite the Google evangelist, but the lack of commitment is one of several reasons I won’t invest time in them now.
Google Docs has been going strong for almost 2 decades, longer than the iPhone. NotebookLM is another way to read your Google Docs, with LLM assistance. If NotebookLM gets killed off, you still have your Google Docs and can read them with your eyeballs or some other LLM.
HN loves to shit on Google for their loose trail-and-error approach to product releases, but Google Docs is evidently not one of those loose products.
I think everybody used to love Google. Around 2012, they were just making one hit after another. Google Docs, Google Maps, Gmail, they revolutionized search!
What have they done in the last 10 years? Jack shit, except, launch a bunch of things that might have been cool, but then just killed it.
How many messenger apps have they had? Google Stadia? I mean, come on, the list is incredible.
The idea of "stanning/hating" a megacorporation, one of the most powerful entities on the planet, is a priori absurd in the vein of Bryan Cantrill's "don't anthropomorphize Oracle".
Apple spends billions of dollars promoting its global brand as a culture and way of life; this does not mean we need to apply this framework to other brands. Spewing negativity (or positivity) about Google in this context is a self-own, because it signals that you have fallen for the "brand loyalty" cantrip.
Why are you replying to a comment about the lack of recent innovation at Google and product positioning missteps with a comment talking about 'stanning/hating. That's not what the commmenter is doing.
Besides, this discussion isn't about "innovation", it's about Google killing products. Google is possibly the most innovative tech megacorporation precisely because of their experimental approach to product releases.
> Google Docs has been going strong for almost 2 decades
I have to disagree here. There have been long stretches of time where it’s appeared to be a mothballed project with virtually no user-visible changes happening to it.
Barely anything happened mid 2019–2021. It looked like a product in maintenance mode for years.
Take July 2019 to July 2020, for instance. In over a year, the only changes of note were: editing headers and footers, an accessibility fix, word count, page numbers, and dark mode on Android. Does that sound like a flagship product from a trillion dollar multinational “going strong”? Or does it sound like a paralysed project with a few interns keeping things ticking over?
And it was worse in the times before that changelog started. That’s hardly “going strong”, that’s surviving on life support. Sure, they never actually killed it and they restarted development once alternatives like Notion started becoming popular, but Google have completely lost interest in Google Docs before and they could do it again – they don’t have the two decade track record you are making them out to have.
For whatever its worth, I worked near docs during that time, and that support note elides a lot of the launches that happened. The big focus during that time was on Office compatibility, and so there were a lot of investment in things that were not end-user visible, but allowed for better coexistence in mixed environments.
A massive launch in terms of impact and usage was the ability to directly edit word documents in docs instead of converting. This also extended to sheets and slides, and so was a major focus.
So - totally agree that there were not a lot of shiny user-facing features during that time, but there was a huge engineering investment going on based on a particular strategy that just didn't show itself sirectly in the UI.
Fair enough. But from an end-user perspective, I’ve been in multiple conversations where a topic was “can anybody remember the last time Google Docs launched a new feature?” and everybody was completely stumped and nobody could think of a single thing launched in years. When the company in question is Google that’s a very strong signal not to rely on that product being around much longer, which is why the state of Google Docs has been so confusing – clearly lost all momentum yet surprisingly not killed.
One major difference being that Google does not read or incorporate the contents of your docs (or at least, we hope not). Once your docs are hooked up to an LLM, consider them both training data and readable to whatever minions are training the AI.
I can't sign up (because I'm outside the US). I think I could build a local-only version with a local LLM possibly local datasources that would do a pretty good job.
Local LLMs are sufficiently powerful for query expansion techniques so the capability gap there to GPT4/Bard isn't a problem.
They are definitely behind on generative capabilities, but since I'm unable to use the product it is unclear how important that is in this use-case. If there is a video I'd appreciate a link!
I wonder how popular a local version of something like that would be?
Not to plug too hard, but I am building a version of this to be local and friendly for your workflows.
The aim is to help people organize, discovery, and pull insights out of their knowledge bases using LLM's as cognitive functions for Q/A as well as pre-thought generated points and summarizes.
“ A key difference between NotebookLM and traditional AI chatbots is that NotebookLM lets you “ground” the language model in your notes and sources. Source-grounding effectively creates a personalized AI that’s versed in the information relevant to you. Starting today, you can ground NotebookLM in specific Google Docs that you choose, and we’ll be adding additional formats soon.”
While it sounds promising, the blog post should have explained the obvious thing: how is source-grounding different from giving a document to a model and generate insights?
Having implemented my own "ask a PDF a question" GPT wrapper, there's a lot of design decisions / complexity that can make it better/worse and I'm not sure how much a "one-size fits all" google approach will work for google. Convenient that it plugs in directly to your gdrive though.
Google will be leveraging the millions of people using google docs that will never go out of their way to use an ask a PDF tool. I know I haven’t bothered even though I am technically skilled enough to do so.
Originally when teased at Google IO this product was called project tailwind but the URL was thoughtful.sandbox.google.com (it seems to now redirect to the notebooklm URL). "Thoughtful sandbox" feels like a much more fitting name.
> Google: "Look, we're capable, but just beta-test this sh&t so we can make a product that will definitely overshadow our little contribution, okay? We need the product, we need the money. You'll get a nice little badge on your profile to signal your interest with your peers. This is open and responsible AI."
Respectfully, nobody gives a f&ck. The actual way you do it is at least publish a paper(or butchered code) or make the product accessible to all.(Or to a number of people that scales proportionally with your claims of greatness). You cannot do either of those? Well that smells of trying to be opportunistic, doesn't it? If i recall, OpenAI in it's infancy with the GPT* family at least published some papers (which Google also used to do, by the way), then they built upon that to eventually release a product. Yes the product was never free at the beginning nor very publicized, because it wasn't advertised to be the holy grail of anything. To sum up my opinion on this matter: you can never truly scale and innovate with >only releasing a lobotomized product< or >only releasing a paper with no visible applications<. You need a little balance between sparking interest and satisfying the hype you give about the product(if any).*
This is exactly what it does, no? The product explicitly is about taking your notes that you write and being able to discuss those notes with the AI. It's in all of the screenshots and the description.
I mean I would take AI autocomplete in the style of copilot. That feels conversational enough already, it would be even better if it’s optimised for prose instead of code. The MS word autocomplete is getting pretty good and I would love more of it personally.
Woof. Talk about a bad reputation exemplified. I can't imagine what it's like working on stuff like this at Google, earnestly just trying to make a great product people will want to use, only to have to fight decades of user-hostile product destruction that the company has done.
Take a look at Instagram's Threads. They had 100M users in a few days. Do you think people around the world really care about data being used?
On a related note, I am sad about the state of HN comments, which resembles Reddit more and more.
The blog post specifically mentions the data is private and won't be used to train models, which can't be said about the ChatGPT based tools which do this, but yet, here we have people not bothering to read and just write a cliched comment about "collecting user data".
I don't think the audience here cares as much about the data. I think people here (myself included) are mostly just deeply cynical about how Google will handle this product long-term. Why invest time into using it when it's going to just get yoinked in a year or two?
Isn't the solution to it -- use Google product if it's superior and won't fuck you if its abandoned? Instead of not using it because it may be abandoned, if those products instead get more users, it is less likely to be abandoned.
I mean, purely considering this product, isn't it also likely for some ChatGPT extension which provides similar feature to also be abandoned, or to have privacy or quality issues?
Only us big nerds on Hacker News actually pay attention to this or care.
Most people never even hear about this stuff. From what I understand, in the Google world, you release a product, you get a promotion. So, as cynical as it is, and everyone in Google knows, their product will get cancelled eventually.
Screw it, I got a promotion, I got some money. It's good for my career.
Dear Google, you've burned me so many times after I've fallen in love with your products. I have experienced the cold sting of disappointment far too many times. When you abruptly discontinue services like Google Buzz and Google Reader, you leave me stranded with no reasonable alternatives. This pattern of abandonment has plagued our relationship, instilling in me a sense of apprehension each time you introduce a new product.
I think a Language Model being used to summarize text with specific direction is one of the primary use cases. I don't really see too many gimmicks here, besides the obvious Google issues
This was immediately my first thought when I saw the product. When is the expiry day? Why aren’t they making it clear?
Also I think this will inspire other companies and help those companies (cough microsoft, apple) create their own version which they will integrate to their own lineup.
It will be interesting seeing it in future.
Disclaimer: I am not AI and I typed this response in my tablet.
People for some reason have tendency to assume I answer like a bot.
According to GPT4, you had a 75% chance of being a bot before the disclaimer. With the disclaimer it gives me a 100% chance or OpenAI will give me my money back.
Haha. That was a funny response from the GPT. I attacked your bot unintentionally!
On serious note, do you think as society progresses with use of AI, our brain will be geared toward filtering everything or we will lose the trust system that is prevalent in the society?
Think about it, the more people are being shut for bot, they more likely they will stop interacting online and this might eventually lead to a lot of people discarding the interaction. For most part, life is pretty average. And if average people are out, what might be the implications?
I've long assumed I'm just a brain in a vat. Solves all those thorny theoretical problems by just tossing the concept of other conscious individuals out the window altogether.
My thoughts exactly! These days I don't take any new service/product from Google seriously because the incentive system in that company is just flawed.
Why would I use any google service with a death countdown on its forehead? I have migrated away from Google almost completely. I can’t imagine being a product team there and having to defend product’s existence for more than a year.
It’s not like some random unproven startup is any better though, unless the product really takes off and makes them money (and they don’t ruin it somehow).
This is amazing! I can't wait to post about this on Google Wave, Google Currents, Orkut, Jaiku, Google Friend Connect, and Google+, read all about it on Google Reader and Google FastFlip, chat about it with friends on Buzz, Duo, Google Talk, Hangouts, and also blog on my website made with Google Page Creator or Web Hosting on Google Drive or Google Sites. It's just so reassuring to know I can depend on the Google ecosystem of products long in to the future.
I get it. Google has a lot of killed/closed products. But that’s kind of their MO. They try a lot of things. Sometimes they work (Gmail was an experiment at one point), but they often don’t.
Making fun of their killed products is a favorite HN past time, but what would you have them do? Just sit back and not try to make new things? Or keep zombie projects alive, but unmaintained just because a few of us find it really helpful?
Those are the projects I find the saddest. The ones that would have been a good product for a smaller company (or was a smaller company that was acquired), but that isn’t profitable enough at “Google scale” to keep throwing money at.
I get it. Google kills a lot of products that early adopters like. Google Reader was painful to lose at the time. But how many RSS feeds do you now follow?
Let’s just evaluate this new project on its own merits. Do you find it helpful or not — without thinking too much about Googles track record. If this way of using AI is helpful (and I’m very hopeful), then either Google will keep it, or this is the POC a small startup will use to keep something like this around. If it fails, then at least we’ll all have learned something.
They can adjust their culture so they promote employees for maximizing the value of existing products, finding product market fit, etc instead of building new projects only to move on from them.
They can gate the creation of new product and require more robust market and user research to ensure they're not releasing something that doesn't have a chance of hitting whatever their targets are, or would require more budget than willing to invest.
They don't just kill products, they often leave them painfully unfinished and unsupported. Support forums are complete ghost towns.
Did you know the "new" Apps Script editor doesn't have a version history or way to revert? Too bad, the old one was sunset. Google would tell you to pound sand, but there's no one there to listen.
Old Google was great. What have they done in the last 10 years? I feel like they just collected engineers like Pokemon and had them spin their wheels on things they would just kill the next year. This is a thing because it's real.
They make so much money on ads they just don't care.
——
Anyway,
It’s not clear which model they’re using for this. I assume whatever Bard is using, but who knows. This is relevant because depending on the intended experience the latency will matter.
Overall it’s not a bad idea, but I do wonder what the monetization path will be for Google. I imagine this will be part of workspace. Perhaps they will add more tiers to include these offerings.
I wish they shared a bit about how this will be differentiated from Bard. Is this simply a new front end to Bard? It’s really an open question. I haven’t seen many products that use LLMs that are better than the prompt response UX.
The most interesting thing about this blog post is the “source grounding.” I’m curious if there’s actual engineering behind it, or is it prompt tweaking contextualized behind the scenes on a given doc.