Can somebody please share what your "summarize this page for me" use cases are?
I really struggle to understand how many people seem to have "summarize this" as their main LLM use case.
And I use LLMs every day, for so many tasks - just never this one. Can you please share a couple of real-world times where you couldn't bother to read the "thing" and truly got value from having a LLM read it for you?
jacobedawson · 2d ago
As a heavy LLM user, professionally and personally, I use "summarize this" a lot - I find that most content in the world is low signal-to-noise and a lot of the time the salient / useful information is hidden within unnecessary layers.
The only time I didn't think summaries are useful is with creative fiction or pure entertainment content.
geor9e · 2d ago
Imagine a horrific world where most articles are 5 pages of LLM generated fluff from a 10 bullet point sentences prompt. I know, truly a nightmare world. But in this world, wouldn't it be great to just collapse all those articles down to their original prompt? A sort of reverse-fluffing. So maybe I can't convince you of how great it is in our world, but hopefully at least in this hypothetical world it makes sense.
behnamoh · 2d ago
I summarize HN threads, blog posts, YT videos, etc. to see if the content is worth my time. So far I've used Perplexity extension.
berkes · 2d ago
Recipes.
"Summarize it. Keep only instructions and ingredients (more prompt about how to format instructions, what units to use, how to order instructions) and start with one sentence that describes the dish and it's origin"
Which is a terrible waste, really. Many recipes are blown up to make them copyrightable. Often done with AI. And then I use AI to remove all that again.
rimunroe · 2d ago
> Many recipes are blown up to make them copyrightable.
To attempt to make them copyrightable at least. My understanding is that fundamental recipes themselves as sequences of steps and ingredients are never copyrightable in the US. From having sat around a lot of recipe bloggers though it sounds like the main reason they do the long essays is to try and improve search ranking.
clearleaf · 2d ago
Recipe sites always have a print button which gives you just the recipe. If I end up on one of those sites I instantly hit Ctrl+F "print" to get it.
TiredOfLife · 2d ago
Try to look at the recipe page without adblock. You will see ads every couple sentences. Longer article more ads.
morkalork · 2d ago
I've been using it on my own written docs to see what my co-workers inevitably see
sebastiennight · 2d ago
Wow, that is both hilarious, depressing, and something that I think I'll start doing now. Thanks
eestrada · 2d ago
I use it on my own docs to remove extraneous details. I often write too many words in early drafts and LLMs summarize my writing faster than I can (although I don't know if they do it better than I would/could).
Then the next version of my doc becomes the summarization, and I only flesh out details where the summarization went too far and removed critical details.
easyascake · 2d ago
I’ve got one. When estimating tasks, there’s a particular blog article I like to reference [0] and I used an LLM to summarize it into a set of project instructions.
Did it read the summary? Nope, I already know the material and have been using it for years. But it was a great way to communicate the key points to the model as part of project instructions.
So in this case, this was not you trying to ingest a condensed version of the info, but rather transforming it to build a prompt (under the assumption that shorter instructions would perform better than the full article).
Makes sense
dvfjsdhgfv · 2d ago
New York Times (and other magazines) articles. "On a quiet Sunday morning..." - No, I don't care about it, just tell me the damn thing (which usually can be done in 4-5 sentences).
gigatree · 2d ago
The text summary for group chats provides a lot more information at a glance than “21 new messages”
aethelingas · 2d ago
I don't use it to summarize, but I like to ask about the key takeaways for long-form content. E.g., "what did he say that was noteworthy or new about X [in this long interview I don't time to read right now]"
lossolo · 2d ago
Long articles that I don't have time to read fully and that contain a lot of words serving only as placeholders, without changing the main message or information.
antithesizer · 2d ago
Yeah we already have a perfectly serviceable way to extract the meat from an article and avoid the filler (i.e. don't click just read the headline)
sebastiennight · 2d ago
> (i.e. don't click just read the headline)
And the HN comments!
kyriakos · 2d ago
Any recipe page on the Web is unreadable without summarization cause they are made for seo.
smallnix · 2d ago
Corporate org level announcement mails
sebastiennight · 2d ago
So Management went to ChatGPT and asked, "can you write a launch email about $event_x " and you then go there to ask, "what did management want to say in this email about $event_x"?
How likely is it that the output summary is mostly made of stuff that the LLM made up in the first expansion process? (eg, you're getting summarized noise if the original signal - the prompt - was much shorter than the email)
rcMgD2BwE72F · 2d ago
Unnecessary long-form news articles that bury the lede.
leoff · 2d ago
How can OpenAI, Anthropic, or any other AI player compete against this? Google can just slap their products to everybody in Android, Google Chrome, Google Search, which creates an unfair advantage.
cheriot · 2d ago
Apple and Microsoft will do the same. The OS war became the browser war becomes the AI assistant war. I suspect there's enough complexity in AI use cases that there will be independent winners for things like coding assist, enterprise agents, etc.
Fun to watch.
rvz · 2d ago
> How can OpenAI, Anthropic, or any other AI player compete against this?
This is what Perplexity is already doing with their own browser. But it tells you that a website is not enough.
OpenAI or Anthropic either need to go and build their own browser or even buy an existing one.
I don't know - for a trillion dollar company, Google is remarkably inept at just about everything that isn't advertising.
They had multiple messaging technologies and never could figure out how to embed one into Chrome to take over messaging (a la iMessage). Google Home has somehow been getting worse with every passing year. I have no doubt they will totally fuck this up and just end up boosting Chrome's already insidious tendency to chew up every spare CPU cycle.
soulofmischief · 2d ago
What's unfair about that? Google is definitely a monopoly, but as long as other models can eventually enjoy similar integrations, through extensions or otherwise, I don't see how it's unfair for Google to add their model to their browser and search page, especially when their chief business model of search (OK, we know it's really just ads, but still) is currently being directly threatened by other model and search providers.
ramoz · 2d ago
Yea i dont think googles distribution has any real threat other than the enterprise monopoly equivalent.
And googles path to evolved ad revenue is clear in Veo3. Effective ads will be personalized videos - Google owns the infrastructure and data to make this real, others do not.
leoff · 2d ago
>as long as other models can eventually enjoy similar integrations, through extensions
They can't, Gemini is being baked in the browser. Other players can't do this.
TiredOfLife · 2d ago
If this integration is like the Googles other ones then it will use much cheaper to run models for it. And not their Gemini ones.
lanthissa · 2d ago
by designing ai first products that can operate at far lower margins. Google has to extract hundreds of billions a year, perplexity, anthropic, openai all dont.
Idk what the future of the browser is but i know if i was in the lab at any of these companies i'd be laughing at the competition putting a out a product that was just a text summary in a window.
ivape · 2d ago
It's very simple. Make a new browser. Chrome was a new browser and they worked hard to get market share. AI-first browsers with completely new workflows are what's in order, and no one is going to just hand that to any company. You'll have to innovate. Chrome literally saved web development, so we're not going to sit around and act like it was a mega company just pushing something down people's throats.
synergy20 · 2d ago
for those who can, DO. for those who can't, SUE (or cry).
anyone can fork chromium, or work with firefox to do similar things, or even write a chrome extension. Google worked hard to make chrome better and leave the rest in the dust, please don't make it like Google is at fault.
The cruelty of reality is that those who are better than you are also working harder/smarter than you, but it's not their fault.
lol768 · 2d ago
Hasn't this already existed in Firefox for the best part of the last year? I see a "Ask ChatGPT" context menu option with various tasks ("Proofread", "Summarise" etc) when I right click. It's easy to remove or point at a different provider too (browser.ml.chat.provider) if you prefer a different model, and I think the prompts are customisable.
sebastiennight · 2d ago
Sounds like you installed some kind of extension maybe? This definitely does not seem like any Firefox default behavior.
lol768 · 4h ago
No, it's part of the browser. Still experimental, though.
Molitor5901 · 2d ago
I wish companies would stop pushing LLMs on us. If I want to use an LLM I will use it, but I would prefer to remove the button for it. It feels .. like an encroachment I didn't ask for.
Maybe I'm just getting old enough to be called a Luddite..
aethelingas · 2d ago
You can right click the button to remove it.
nullbyte · 2d ago
Its a great idea, Ive been looking for an extension that can do this. Pretty nice its baked in now
andybak · 2d ago
Isn't it worse having it baked in? Usually extensions move faster, compete for best feature set etc etc.
"Baked in" usually means "that's the last update we're gonna see".
Plus I might want to - you know - switch browsers one day.
SoKamil · 2d ago
I wonder how The Browser Company wants to compete with this given infinite Google resources.
rvz · 2d ago
They don't. They just get themselves acquired.
But the question is by "who"?
6 months later after this [0] it now looks almost very obvious.
Google should have a huge advantage in rolling out computer use agents because of their ownership of Chrome. Other agents running in the cloud don't have my passwords and login cookies and history, but Chrome does.
andybak · 2d ago
And that is a terrifying sentence to read.
charcircuit · 2d ago
So does Microsoft with Windows. Or Apple. The platforms these companies have built are very useful to consumers do there is a lot of potential value in supercharging them with AI.
modeless · 2d ago
I use Chrome a whole lot more than I use any native Windows or macOS app. And when I'm not using Chrome I'm more likely than not still using Chromium (VSCode, Slack, etc). The OS native UI bits are mostly an annoyance at this point.
_pdp_ · 2d ago
From the marketing page it does not feel any different than a browser extension. I don't know. Besides, a browser extension is actually more useful given it could theoretically provide access to other models too.
tpm · 2d ago
"Google AI Pro and Ultra subscribers can now get early access to Gemini in Chrome. Learn more"
And that link goes to a page that does not know I already have AI Pro subscription and does not say anything about "early access to Gemini in Chrome". Great.
jadbox · 2d ago
Note that the current version does not support Google workspace actions, like checking email, creating calendar items, or looking up notes in Keep. Essentially Gemini in Chrome is just the LLM with the current page context. (Also does not support multiple tabs yet)
I really struggle to understand how many people seem to have "summarize this" as their main LLM use case.
And I use LLMs every day, for so many tasks - just never this one. Can you please share a couple of real-world times where you couldn't bother to read the "thing" and truly got value from having a LLM read it for you?
The only time I didn't think summaries are useful is with creative fiction or pure entertainment content.
"Summarize it. Keep only instructions and ingredients (more prompt about how to format instructions, what units to use, how to order instructions) and start with one sentence that describes the dish and it's origin"
Which is a terrible waste, really. Many recipes are blown up to make them copyrightable. Often done with AI. And then I use AI to remove all that again.
To attempt to make them copyrightable at least. My understanding is that fundamental recipes themselves as sequences of steps and ingredients are never copyrightable in the US. From having sat around a lot of recipe bloggers though it sounds like the main reason they do the long essays is to try and improve search ranking.
Then the next version of my doc becomes the summarization, and I only flesh out details where the summarization went too far and removed critical details.
Did it read the summary? Nope, I already know the material and have been using it for years. But it was a great way to communicate the key points to the model as part of project instructions.
[0]: https://jacobian.org/2021/may/25/my-estimation-technique/
So in this case, this was not you trying to ingest a condensed version of the info, but rather transforming it to build a prompt (under the assumption that shorter instructions would perform better than the full article).
Makes sense
And the HN comments!
How likely is it that the output summary is mostly made of stuff that the LLM made up in the first expansion process? (eg, you're getting summarized noise if the original signal - the prompt - was much shorter than the email)
Fun to watch.
This is what Perplexity is already doing with their own browser. But it tells you that a website is not enough.
OpenAI or Anthropic either need to go and build their own browser or even buy an existing one.
Does anyone have an "Idea" who it could be? [0]
[0] https://news.ycombinator.com/item?id=42213288
They had multiple messaging technologies and never could figure out how to embed one into Chrome to take over messaging (a la iMessage). Google Home has somehow been getting worse with every passing year. I have no doubt they will totally fuck this up and just end up boosting Chrome's already insidious tendency to chew up every spare CPU cycle.
And googles path to evolved ad revenue is clear in Veo3. Effective ads will be personalized videos - Google owns the infrastructure and data to make this real, others do not.
They can't, Gemini is being baked in the browser. Other players can't do this.
Idk what the future of the browser is but i know if i was in the lab at any of these companies i'd be laughing at the competition putting a out a product that was just a text summary in a window.
anyone can fork chromium, or work with firefox to do similar things, or even write a chrome extension. Google worked hard to make chrome better and leave the rest in the dust, please don't make it like Google is at fault.
The cruelty of reality is that those who are better than you are also working harder/smarter than you, but it's not their fault.
Maybe I'm just getting old enough to be called a Luddite..
"Baked in" usually means "that's the last update we're gonna see".
Plus I might want to - you know - switch browsers one day.
But the question is by "who"?
6 months later after this [0] it now looks almost very obvious.
[0] https://news.ycombinator.com/item?id=42213288
And that link goes to a page that does not know I already have AI Pro subscription and does not say anything about "early access to Gemini in Chrome". Great.
Thought it was important enough but somewhat hidden on their website behind a scrolling faq.
https://stocktwits.com/Johnnyalgo/message/616611886
TLDR: Judge on the case appears to be more sympathetic toward Google than reported.
—-
Imho, Google has been the most relentless on pushing AI than any other company atm. Sundar is not getting enough credit for this.
No comments yet