Googling for answers costs time (2020)
prashanth.world
prashanth.world
When I ask ChatGPT it'll always give me the answer in a few paragraphs, and it's always written in the exact same way. What a boon!
> How is this the number one link on Google?
It's the number one link on Google because one proxy Google has to know if you got a result is that you lingered on the page for a long time. This is also, tangentially, why cooking recipe websites are so long. If you bounce immediately off a website, Google thinks that means that you didn't find it to be useful. ChatGPT does not have this problem.
ChatGPT’s strength is that it’s new on the market, needs to build an audience, and therefore their UI and UX are lean, devoid of gaming.
Same as Google, up until 2008.
A lot more than I've ever paid to use the Google search engine.
Paid for service over ads is the new way forward.
Yesterday we asked it what “RF” inscribed in a crest on a government building meant (“République Française”). A few days ago we asked why an area was called Gobelins, and if that had any relationship to the modern goblins in fantasy. I asked how to ask for more bread at the table and it gave me a pretty long, fancy French sentence. I asked it to simplify the sentence and it did beautifully. You can ask Google translate to translate something, but chatgpt can give you multiple translations, and tell you the cultural nuances of all of them and teach you French grammar while it’s at it. “How do I say this? Why does it use suis in that sentence? Is that a weird thing to say to the wait staff at a restaurant? What is the protocol there?” Etc.
We’ve been using it constantly.
For example, if I ask what the best vacuum cleaner on the market is, I'd rather it not regurgitate bestvacuumreviews.net which is affiliate (not quality) driven
There’s three reasons…(1) ad revenue, (2) there’s a belief that the extended life stories confer copyright to recipes, which otherwise cannot be copyrighted and (3) everyone does it.
Source: I made SimplifyRecipe.com, which requires diving into the minutiae of recipe authors.
Otherwise, I agree that getting consistent results from GPT is super useful.
Or they work for a selection of cases but not others. Or they seem to work in a limited number of tests but aren’t truly doing the right thing. Or they do work but have a security flaw that makes your service vulnerable.
Coding solutions have tradeoffs and subtleties which go beyond “it did the thing right a couple of times when I ran it”. If that’s all it took to verify code is good, there wouldn’t be any bugs; you can test code a person writes too.
But contrast this with other types of information. If I ask ChatGPT "how hot is the surface of the sun?" and it answers "15 million degrees Fahrenheit", I have zero ability to verify that information. As a result, I don't find ChatGPT useful for most informational searches, except the ones related to coding.
(The example answer is wrong according to https://www.space.com/17137-how-hot-is-the-sun.html, albeit in a way that I could see potentially confusing an LLM.)
But the nice thing about ChatGPT is that you can give it more context than it is possible in a Google query, and you can ask follow-up questions.
They are still useful, you just have to make sure you understand what you are doing and read the relevant docs before you paste the code.
Things like "How do I delete a value from a dict" probably either works or it doesn't.
Then again, for that level of simple stuff, one can probably search a PDF cheat sheet just with their eyes faster than they can type the question for ChatGPT
On the surface it makes no sense to do this since it already gave you an answer. But that's not how LLMs work - they can simulate a variety of different "humans" depending on the prompt and presenting a different prompt persona (security researcher, code reviewer that focuses on edge cases, etc) yields a different quality of response.
And in general I find it’s more accurate and far more concise than anything that’s going to show up in a Google search.
I mean, there’s no way I’m taking what it says at face value or copy / pasting code from its response, but that goes for stack overflow, or any other source too. (maybe sans official docs)
I think the internet is going to start to stratify into paid and "free" tiers going forward, and honestly it's about goddamned time.
I have no way to verify that.
What I care about is a search engine that's trying its hardest to serve me answers to my searches, rather than feeding me ads.
AI tools today are great for the same reason Google was great in its first few years of existence: they want to attract users and sell this awesome new product. But this will quickly change.
The ad and investor-fueled enshittification will inevitably also come to popular AI tools, but in even more sinister ways than web search results. Forget banners and traditional ad campaigns. Users will start seeing subtle product placements in the responses, and certain brands placed higher than others, all embedded in the friendly conversational tone they're used to, to make them even more susceptible to manipulation. Thanks to the sea of private data users are willing to share with their "personal" AI assistant, the responses will be perfectly fine-tuned so that ads are targeted at them _personally_. Advertisers will no longer need to target broad cohorts who may or may not be interested in their brand. They will be able to reach _exactly_ people who are most likely to make a purchase, especially if the ad is delivered at _exactly_ the right moment when the person is most vulnerable.
The only way to avoid this is to use open source and self-hosted AI tools, but the vast majority of people will use whatever service is popular and "free", same as today.
For basic stuff it is much faster than Google. But if it requires minimum thinking it still faster doing myself.
The other day I was lazy to find a generic scraper I have so I could save a reference manual. I know it is just like 50 lines of code and if you are smart you could do it async. I tried with ChatGPT and it didn't work at all. Many silly mistakes hidden between comments that looked reasonable
Spending time on the page should actually be a great proxy for usefulness on recipe websites. If it's a long page of introduction and anecdotes, maybe it took me three minutes instead of one to reject it as unpromising. If it's the recipe I actually used, however, that tab is going to be open for the next thirty minutes or more.
I’ve seen “they’re long so there’s more space for ads” and “they’re long so you spend more time there so it doesn’t get penalised by google”, and to me those are pretty closely related.
It's a situation where Google doesn't benefit much more from it being the actual answer, and the site doesn't either.
In the long run it could have an impact on the user perception, that could be nullified by the page still staying ranked higher than the competiting pages and the user finding nowhere else better.
Actually you're probably getting someone's list of 10 best films that got copied to 1000 more sites and ended up training ChatGPT...
1. Build a skill set to navigate millions of pages of fluff SEO garbage and gain intuition of which sites are more likely to be credibility based on something akin to a internet dosing rod.
2. Build skill set to evaluate if a statement or claim is factually true. i.e. ChatGPT just told me something, how likely is it to be true.
I know which skill set I would rather develop, obviously #2, as it's transferable to nearly every intellectual venture a human would consider, while #1 is not.
Also, I feel the people that play the FUD campaign on GPT hallucinations just haven't used it much. The average quality of GPT answers, at least in programming, is far above the average hit on google, and the fact that I can interact with the answer and shape it to go in any direction I want or mix it match it with nearly any other concept on the spot, it's light years more valuable than the average google search. At this point, I would guess I only resort to google 10% of what I used to, and it's normally to attempt to find a specific website that has specific information I remember accessing before that I want to look up again.
Probably because there would be barely anything left in the results once ads and spam are removed.
I also think Google's ranking is fair in this case, assuming you want to know most high quality information about "how to iterate through a dictionary". I certainly would prefer this kind of a RealPython article with a table of contents so that I can find exactly what I am looking for.
I do think Google (and the web in general) has started to suck because of SEO spam, and generally agree with the authors premise. Searching for recipes suck, searching for reviews suck, searching for places and events also suck.
I just think this particular instance is a bad example. RealPython values your time, and I wish there were a lot more websites that valued my time too.
I guess they wanted an answer to their specific question and absolutely nothing else? That might work for something as simple as “iterate through a dictionary” but in many other cases the full context is useful so you can figure out for yourself whether the immediate answer actually solves your problem.
Was the first 20 paragraphs on what a dictionary is really useful? No, it's unnecessary verbiage that anyone clicking on an article with the title knows.
I am declining your promise of erotic imagery but thanks for thinking of me.
But even if you click the link in the index, then you still need to scroll down two to three pages to get to the actual answer:
for key in a_dict:
print(key, '->', a_dict[key])
- "What dictionaries are, as well as some of their main features and implementation details"
- "How to iterate through a dictionary in Python by using the basic tools the language offers"
- "What kind of real-world tasks you can perform by iterating through a dictionary in Python"
- "How to use some more advanced techniques and strategies to iterate through a dictionary in Python"
Yes, if you spend 30 minutes reading the article, you will know how to iterate over dictionaries afterwards. But if you asked a human being, they would just say "for key in dict:".
Edit: I tried sending the query "How to Iterate Through a Dictionary in Python" to ChatGPT, and it immediately shows relevant code samples that exactly answer the question, without pages of pointless filler: https://imgur.com/a/CenZFl0
There's no arguing that the ChatGPT answer is a million times better.
Absolutely agree with this. I strongly prefer reading the official language API documentation directly, which makes me very sensitive to the quality of the documentation.
I have also replaced my google searches (which usually are more along the lines of “what’s the most idiomatic way to do x?” or “what’s the most correct way to do x in $LANG?”) with ChatGPT. It’s super fast, generally pretty good for the types of style questions you’d usually have learning a new language, and I can directly query it about what tradeoffs different techniques have in various common situations.
ChatGPT has finally gotten me off Google’s Wild Ride.
Phind gets an honorable mention though - if that’s the style you prefer I’d strongly recommend it.
ChatGPT does hallucinate all the time, but for “learning a language from scratch” it’s very very very minimal. I can’t recall a single time it gave me a response to “idiomatic way to do x in $LANG” which included hallucinated information. Just can’t back it into a corner trying to get it to customize something that needs rigorous optimization for your use case.
Overall, the large innate knowledge base of ChatGPT-4 seems to contain a great model of the modern features of most programming languages and their packages (popular PyPI packages, Spring, Boost, Qt, etc, etc). Once I let it browse the internet, quality goes down and roughly matches Bing and Bard.
Phind I think gets around this by having a custom ranking of sources or something. Not sure what their secret sauce is but it mostly works pretty well.
I can't begin to describe how much it slows me down to have to keep typing out my question repeatedly to an AI, only to be robbed of the ability to judge the information in the context of where it came from, and oh yeah, any confidence that it's accurate.
This is a key point, and I don't think the ability to acquire knowledge online is a skill which only young people specifically have failed to acquire.
Said ability must be trained on, as with anything, a suitable dataset. This objective cannot be meaningfully achieved if all data sources someone has ever consumed and learned from are TikTok, ragebait, clickbait, affiliate link blogs, email chains, WhatsApp message chains, sensationalist news outlets, etc.
This is why people still fall for the Nigerian Prince, Hot MILFs in Your Area, job application with deposit and all the modern variants; not only old people but all types of people.
I think some of us nerds have a slight advantage in that we experience the internet early on in a slightly different way: as a parallel to the real world, with friends and foes, with truth and deceit and as a playground to write a game mod or a bot for shit and giggles. We are, due to that I'd pose, much more keen to learn how and be able to effectively navigate it.
Because it's rich in content, it's long, has a table of contents, the whole site is relevant to the search query, etc etc etc; from a search engine's point of view, it's more authoritative than Stack Overflow where the authoritartativity (?) is debatable at best, decided by consensus otherwise.
What should happen is that the community behind Python (in this case) works their ass off to make authoritative sources, like MDN did for web standards. And they had to work and fight for years to overtake w3schools.
I wonder how far Google could get with manual curation. Like if they said, these sites (eg, Stack Overflow, official documentation) have been around forever and have high-quality content, let's crank up their rankings for the appropriate classes of queries. That has its own problems, but it's gotta be better than the status quo.
In the case described in the linked post, Google potentially benefits twice, financially. Why would they want to reduce their profit by showing the more relevant posts (those that quickly get to the answer)?
I don't know though, maybe StackOverflow also uses Google for their ads, but a quick looks suggests they simply have less ads than the results above them in the article's example search.
It should be a required step when learning something--anything--that you learn some good sources for its documentation first, so that you're not blindly Googling or LLMing around for random bad answers.
I notice others downthread discussing recipes. The first thing I did when learning to cook was to find just a few reliable sources for cooking techniques and recipes. I collected a few YouTube channels and I realized that WikiHow is a really good, spam-free site that teaches technique and also provides basic recipes. And so I just stopped Googling "how to bake lamb rib chops" because that's a fool's errand.
Load up the "docsets" of your languages (lightly edited HTML docs for indexing purposes) and use a global keyboard shortcut (F8 for me) to pull up Python/Postgres/Terraform docs, searching for the right function without internet query.
This isn't straight up applicable to all questions of course, but "How do I search regular expressions in Python again?" is now as easy as <F8> "re" <ENTER>
Note that the docsets can be converted from normal HTML ones via doc2dash[3], especially useful to load up custom docs like private providers.
[1]: https://zealdocs.org/ [2]: https://kapeli.com/dash [3]: https://doc2dash.readthedocs.io/en/stable/usage/
The author searched for: "python iterate over dict".
The number 1 result was an article called: "How to Iterate Through a Dictionary in Python". The article doesn't have obnoxious ads, has a clear table of contents at the top so you can jump to specific sections if you don't need the full context, and is written in a way that novices can learn the topic while still providing all the technical details that an advanced user would need.
Also, I'm not sure what resolution screen the author is using (1280x768?), but on my 1080p screen I can see the top of the table of contents so could jump straight to any of the methods to iterate through a dictionary.
In my opinion, this is an example of Google providing a great result.
If you're a novice you start with "what is a dictionary" which is where the first half of the article belongs.
I'm well aware of ChatGPT's ability to confidently say wrong things, but the cost of reading SEO fluff is so high I'm more than willing for that tradeoff.
Interestingly I've found that GPT also tends to use SEO fluff, likely because its trained off SEO-fluffed content.
The difference is that I can just tell ChatGPT to be concise and the fluff is gone
I preface all prompts with "Be concise and ask follow-up questions if you need more information" -- Greatly improves the output
(not affiliated in any way)
I believe the whole discussion here is about the site being anything but concise.
If you want to write an article on Python dictionaries, sure spend 20 paragraphs on it. But don't title the article "How to iterate through a dictionary"
I'm surprised they didn't spend another twenty describing the origination of term iterate from the original Latin 'iterum'.
In contrast, I read all of K&R before writing a single C program, and when I had to learn PHP, I went to the official documentation (it was a downloadable .chm) and read through most of it first. It's a larger upfront investment, but I think taking the time to understand the language first instead of "JIT'ing" knowledge is worth it.
And also, googling for something has a high hit rate; nine times out of ten, googling 'how to sort in python' gives a good result.
And also also, these questions are often one-off; you need to look up how to do X in $lang, then either you can copy / paste, or you never touch it again; why put in the time investment if you don't need to rely on your memory?
That says more about the horrible state of the industry than anything else. Big languages like Python won't be disappearing anytime soon, so even if I might not be doing more of it for now, learning the language more thoroughly will pay off again in the future.
Wrong! At least in the case of Python. The main reason Python search results are so bad is that the official documentation is really badly organised. People literally can't find the answers they want there because everything is on one mega-page with a bunch of unrelated concepts. So Google ranks it badly.
The author's example ("python iterate over dictionary") is a good one. I tried that search - doc.python.org is NOT IN THE RESULTS (at least for the first 3 pages).
It's not hard to see why - the closest answer I could find is here:
https://docs.python.org/3/tutorial/datastructures.html#loopi...
but it's on a page that also talks about lists, tuples, assignment expressions, nested list comprehensions, etc. No wonder people don't want to read all that!
And it doesn't even answer the question what `for x in somedict:` does - for that you have to go to this mega-page:
https://docs.python.org/3/library/stdtypes.html#typesmapping
and see that `iter()` is "a shortcut for iter(d.keys())".
Compare it to Rust "rust iterate over hashmap".
The 4th result is https://doc.rust-lang.org/std/collections/struct.HashMap.htm...
It's just about hashmaps. The example at the top has iteration over one.
[1] https://kagi.com
To get free training so you can solve your problem yourself and understand the overall situation seems a very good deal. The example to complain about seems very badly chosen.
I think that Google is getting worse and that advertisements in webpages should be regulated. But to complain that you are being taught to fish instead of given a fish seems too entitled.
It does not look like SEO spam to me. The site seems actually nice and well produced.
I may be missing something, thou. Why is that page so bad that the author does not want it in the list of results?
But somehow the author of this article thinks that this kind of unnecessary level of detail is appropriate.
I always end up forgetting how to do X if I haven't done X in awhile, so having it documented by an author I trust in a no-nonsense place I have ready access to is useful.
I'm amazed by how well very specific answers become top 3 results.
Last week, I wondered how to parse a BCC recipient using Cloudflare Email workers myself. Wrote a blog post about it, and 5 days later it is the first result when googling for: parse bcc recipient Cloudflare email workers
So, what is the alternative? You plaster the page with an eBooks-worth of content because that satisfies Googles EAT (something something topical authority) signals. It’s that simple.
In many ways you can steer Googles results in a direction but developers are too lazy to make a /link section on their blog and save a bookmark once in a while. It actually helps the people whose content you like and it costs you absolutely nothing.
I don't know if anyone asked and I don't know if it can be figured out but I was wonderin are the bots any better than figuring out them search operators?
To counter this, websites employ a strategy known as the "Bucket brigade" to ensure you spend at least 30 seconds on their site. It's a basic principle of website optimization and it's just the tip of the nasty SEO iceberg.
Or it takes one click on a relevant item in the table of contents
Basically, just go to cheat.sh/git and you get a list of useful examples for how to use the git command.
I have it mapped to a small bash function so that I can just type ‘cheat git’ and I get the content right in my terminal.
[1] - https://pluralistic.net/2023/01/21/potemkin-ai/#hey-guys
Far faster than wading through the verbose and insufferable output of some chatbot and your give the source of all that knowledge some recognition.
It works less well when you have a question that doesn't have an obvious answer.
For fun, try 'How can I write to a Parquet file in Java without Hadoop?'. It will vacillate between giving answers that pull in hadoop dependencies, and nonsense that doesn't compile at all.
(here's a useful proper answer: https://stackoverflow.com/questions/59939309/read-local-parq...)
Complex questions that involve things that aren’t well documented are obviously going to be harder for an LLM to answer, particularly when the LLM does not have a capability to test its own code.
I briefly checked if the mentioned dependencies exist (they do). Too late over here to actually try the code.
Main.java:27:33
java: org.apache.parquet.io.OutputFile is abstract; cannot be instantiated
Main.java:29:17
java: cannot access org.apache.hadoop.fs.Path
class file for org.apache.hadoop.fs.Path not found
Google should stop this. It's just going to make ChatGPT more attractive.
That's mostly because SO search is (or, has historically been) terrible.
> or documentation page
Ironically, documentation pages exist and many of them are terrible. But even the great ones aren't linked to in code snippets. Got a squiggly line? It would be handy to have a link to documentation!
It still is. There is a trick though - start typing a new question and it will suggest you possible duplicates. Works like magic, they really did a great job with this "search".