HNHacker News
TopNewBestAskShowJobs

gregw134

1,247 karma · joined December 5, 2012

submissionscomments
gregw134··on Mojeek – The alternative search engine that puts the people who use it first
All we need is to show less SEO and more sites made with passion, that have communities involved in them, etc.
gregw134··on From grep to SPLADE: a journey through semantic search
It's various measures of recall rate. Recall@500 means what percentage of the time does the target document show up in the top 500 results from the retrieval system.
gregw134··on Perplexity AI is lying about their user agent
Pretty sure 99% of what Perplexity does is Google your request using a headless browser and send it to Claude with a custom prompt.
gregw134··on Hate Chatbots? You Aren't the Only One
My only experience with chatbots outside of Chatgpt/Google was with comcast yesterday, in which it tried to convince me for hours that it was a human fixing my internet connection, and I just had to wait patiently. Comcast mixed together real humans with bots, so you never knew if you're just talking to a human helping you or a bot, just that you're in the seventh ring of hell for customer support.

I'm souring quickly on the concept

gregw134··on Hate Chatbots? You Aren't the Only One
Nice sentiment, but are you using chatbots now and having a positive experience with them?
gregw134··on Hate Chatbots? You Aren't the Only One
Technical support with Comcast yesterday: https://imgur.com/a/NJMEIkm

Their actual human support couldn't figure out my problem, but instead of admitting it they just turn on AI-chatbot mode, effectively hanging up on the customer. There's no UI indication you're not talking to a human anymore. Meanwhile the AI tries to engage in small talk and tells you to patiently wait while they're fixing your problem. Eventually an actual human came back and admitted they couldn't fix the problem and was escalating me to "tier-2 support", which turned out to be another chatbot. Incredibly dystopian experience.

Meanwhile their engineers probably got a promotion for this and wrote an arxiv post bragging about their work: https://arxiv.org/html/2405.00801v2

gregw134··on DuckDuckGo was down
Pretty sure kagi is a bing wrapper too, although they blend in some other datasets.
gregw134··on More info about synonyms at Google (2010)
Please no
gregw134··on Google made me ruin a perfectly good website (2023)
Doesn't Microsoft have an ad network?
gregw134··on How ruthless is Amazon, really?
How so? I'm building a startup now and am using aws, curious if you'd still recommend.
gregw134··on The man who killed Google Search?
Ex-Google search engineer here (2019-2023). I know a lot of the veteran engineers were upset when Ben Gomes got shunted off. Probably the bigger change, from what I've heard, was losing Amit Singhal who led Search until 2016. Amit fought against creeping complexity. There is a semi-famous internal document he wrote where he argued against the other search leads that Google should use less machine-learning, or at least contain it as much as possible, so that ranking stays debuggable and understandable by human search engineers. My impression is that since he left complexity exploded, with every team launching as many deep learning projects as they can (just like every other large tech company has).

The problem though, is the older systems had obvious problems, while the newer systems have hidden bugs and conceptual issues which often don't show up in the metrics, and which compound over time as more complexity is layered on. For example: I found an off by 1 error deep in a formula from an old launch that has been reordering top results for 15% of queries since 2015. I handed it off when I left but have no idea whether anyone actually fixed it or not.

I wrote up all of the search bugs I was aware of in an internal document called "second page navboost", so if anyone working on search at Google reads this and needs a launch go check it out.

gregw134··on Don't bet on the joys of pokies (2011)
I found a local poker club, fortunately. $40 buy in tournament, no rebuys, every week, with the same 30 or so people that show up. Way more fun than playing at the casino with sad addicts and sad poker pros grinding $40k annual salaries.
gregw134··on SentenceTransformers: Python framework for sentence, text and image embeddings
I'm sure it would lose a ton of meaning, but for me it's easier to fit into a traditional search pipeline.
gregw134··on SentenceTransformers: Python framework for sentence, text and image embeddings
> [recommend against] feeding very large pieces of text to the embedder

Sounds right, I've heard this from multiple sources. That's why I'm leaning towards just embedding the keywords.

gregw134··on SentenceTransformers: Python framework for sentence, text and image embeddings
Any idea which model has the best performance across languages? I'm checking out model performance on the Huggingface leaderboard and the top models for English aren't even in the top 20 for Polish, French and Chinese
gregw134··on SentenceTransformers: Python framework for sentence, text and image embeddings
How are you guys deciding what parts of a document to turn into embeddings? I've heard paragraph embeddings aren't that reliable, so I'm planning on using tf-idf first to extract keywords from a document, and then just create embeddings from those keywords.
gregw134··on 'Lavender': The AI machine directing Israel's bombing in Gaza
Risk any American soldiers? Definitely not. Support with drone strikes, sanctions, intelligence? Already doing that.
gregw134··on DeWitt and Stonebraker's "MapReduce: A major step backwards" (2009)
Short answer: yes
gregw134··on DeWitt and Stonebraker's "MapReduce: A major step backwards" (2009)
Isn't the Hadoop story also playing out with spark--the open source version is kinda buggy, while the spark committers at Databricks have retained all the fixes for themselves.
gregw134··on Julius Caesar's Year of Confusion
Any other history channels you enjoy? I finished every historia civilis video and need a new history channel to watch.
gregw134··on Julius Caesar's Year of Confusion
Historia Civilis is pure gold for history enjoyers.
gregw134··on Google to pause Gemini image generation of people after issues
Interesting, it says 36% of software developers are Asian but only 9% of web developers.
gregw134··on Nvidia is now more valuable than Amazon and Google
Bell labs, xerox parc, and sun microsystems. Quite the technical resume.

>Early in his career, Schmidt held a series of technical positions with IT companies including Byzromotti Design, Bell Labs (in research and development),[19] Zilog, and Palo Alto Research Center (PARC).

During his summers at Bell Labs, he and Mike Lesk wrote Lex,[23][7] a program used in compiler construction that generates lexical-analyzers from regular-expression descriptions.

Sun Microsystems In 1983, Schmidt joined Sun Microsystems as its first software manager.[19]

gregw134··on Stract: Open-souce, non-profit search engine
I'm just using common crawl for now
gregw134··on Stract: Open-souce, non-profit search engine
Wanted to say congrats on launching! I'm building a search engine myself, I can tell a lot of work went into this.

I think the biggest thing you overlooked are page titles. When you issue a query it's a bit hard to quickly scan and judge what a site is about because the page titles are missing.

gregw134··on New embedding models and API updates
Probably better than hallucinating a wrong answer if it's stumped
gregw134··on How Euler Did It, by Ed Sandifer
Most people can't name any mathematicians, but amongst those that can I think he's top 5, behind more famous names like Newton, Archimedes and Euclid
gregw134··on Salt and salary: Were Roman soldiers paid in salt? (2017)
Yeah that's what I figured. Julius Caesar marched on Rome with a hired army, then after his death Marc Antony and Octavian kept having to bribe their soldiers to fight and not switch side. Sounds like the army and leaders established an unhealthy precedent.
gregw134··on Salt and salary: Were Roman soldiers paid in salt? (2017)
The centurions were only paid double what soldiers were during the Republic era. Interesting their pay jumped to up to 30x during the empire.
gregw134··on Intel's CEO says Moore's Law is slowing to a three-year cadence but not dead yet
Have any examples of what's being fixed?
← PreviousPage 2 of 9Next →