Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI assisted search-based research works now

simonwillison.net

Post reply on HN
Print view View original

71–80 of 156 posts

Re: AI assisted search-based research works now

#71

Post by dghlsakjg »

neom wrote: post #64
CSMastermind wrote: post #57

The various deep research products don't work well for me. For example I asked these tools yesterday, "How many unique NFL players were on the roster for at least one regular season game during the 2024 season? I'd like the specific number not a general estimate." I as a human know how to find this information. The game day rosters for many NFL teams are available on many sites. It would be tedious but possible for m…

Is it accurate that there are 544 rosters? If so, even at 2 minutes a roster isn't that days of work, even if you coded something? How would you go about completing this task in 1 hour as a human? (also chatgpt 4.1 gave me 2,503 and it said it used the NFL 2024 fact book)

If the rosters are in some sort of pretty easily parsed or scrapable format from the nfl, as sports stats typically are, this is just a matter of finding every unique name. This is something that I imagine would take less than an hour or two for a very beginner coder, and maybe a second or two for the code to actually run

Re: AI assisted search-based research works now

#72

Post by jeffbee »

The Deep Research stuff is crazy good. It solves the issue that I can often no longer find articles that I know are out there. Example: yesterday I was holding forth on the socials about how 25 years ago my local government did such and such thing to screw up an apartment development at the site of an old movie theater, but I couldn't think of the names of any of the principals. After Googling for a bit I used a Deep Research bot to chase it down for me, and while it was doing that I made a sandwich. When I came back it had compiled a bunch of contemporaneous news articles from really obscure bloggers, plus allusions to public records it couldn't access but was confident existed, that I later found using the URLs and suggested search texts.

Re: AI assisted search-based research works now

#73

Post by levocardia »

TrackerFF wrote: post #48
jsemrau wrote: post #3

My main observation here is 1. Technically it might be possible to search the Internet, but it might not surface correct and/or useful information. 2. High-value information that would make a research report valuable is rarely public nor free. This holds especially true in capital-intensive or regulated industries.

I'm not a researcher, but don't most researchers these days also upload their work to arXiv? Sure, it's not a journal - but in some fields (Machine Learning, Math) it seems like everyone also uploads their stuff there. So if the models can crawls sites like arXiv, at least there's some decent stuff to be found.

Not outside of ML, physics, and math. Preprints are extremely rare in many (dare I say most) scientific fields, and of course many times you are interested in not the cutting edge work, but the foundational work in a field from the 60s, 70s, or 80s, all of which is locked behind a paywall. Or at least it's supposed to be, and corporate LLMs are not "allowed" to go poking around on sketchy Russian website for non-paywalled versions.

Re: AI assisted search-based research works now

#74

Post by krainboltgreene »

dghlsakjg wrote: post #71
neom wrote: post #64

Earlier quoted context omitted.

Is it accurate that there are 544 rosters? If so, even at 2 minutes a roster isn't that days of work, even if you coded something? How would you go about completing this task in 1 hour as a human? (also chatgpt 4.1 gave me 2,503 and it said it used the NFL 2024 fact book)

If the rosters are in some sort of pretty easily parsed or scrapable format from the nfl, as sports stats typically are, this is just a matter of finding every unique name. This is something that I imagine would take less than an hour or two for a very beginner coder, and maybe a second or two for the code to actually run

FYI for readers: All the major leagues have a stats API, most are public, some are public and "undocumented" with tons of documentation by the community. It's quite a feat!

Re: AI assisted search-based research works now

#76

Post by btbuildem »

fzzzy wrote: post #8
btbuildem wrote: post #4

It's a relevant question about the economic model for the web. On one hand, the replacement of search with a LLM-based approach threatens the existing, advertising-based model. On the other hand, the advertising model has produced so much harm: literally irreparable damage to attention spans, outrage-driven "engagement", and the general enshittification of the internet to mention just a few. I find it a bit hard to i…

You need a copy of r1 and enough ram to run it, and a web searching tool, or a rag database with your personal data store.

R1 would be the reasoning model - as in, the initial part of the output being the "train of thought" revealed before the "final answer" is provided. I was able to deploy a heavily quantized version of that locally, and run it with RAG (Open Webui in this instance) -- with web search enabled, sure, but it's still a far cry from an actual "research" model that know when and how to seek extra data / information.

Re: AI assisted search-based research works now

#77

Post by ac29 »

sshine wrote: post #32
simonw wrote: post #7

Earlier quoted context omitted.

I fully expect one of the AI-related business models going forward to be charging subscriptions for LLM search tool access to those kinds of archives. ChatGPT plus an extra $30/month for search access to a specific archive would make sense to me.

Kagi is $10/mo. for search and +$15/mo. for premium LLMs with agentic access to search.

The $10 plan includes the LLM assistant a now as well (with a more limited selection of models than the $25 plan).

Re: AI assisted search-based research works now

#78

Post by sandspar »

oulipo wrote: post #20

The main "real-world" use cases for AI use for now have been: - shooting buildings in Gaza https://apnews.com/article/israel-palestinians-ai-weapons-43... - compiling a list of information on Government workers in US https://www.msn.com/en-us/news/politics/elon-musk-s-doge-usi... - creating a few losy music videos I'd argue we'd be better off SLOWING DOWN with that shit

You seem ideology motivated instead of truth motivated which makes you untrustworthy.

Re: AI assisted search-based research works now

#79

Post by CSMastermind »

neom wrote: post #64
CSMastermind wrote: post #57

The various deep research products don't work well for me. For example I asked these tools yesterday, "How many unique NFL players were on the roster for at least one regular season game during the 2024 season? I'd like the specific number not a general estimate." I as a human know how to find this information. The game day rosters for many NFL teams are available on many sites. It would be tedious but possible for m…

Is it accurate that there are 544 rosters? If so, even at 2 minutes a roster isn't that days of work, even if you coded something? How would you go about completing this task in 1 hour as a human? (also chatgpt 4.1 gave me 2,503 and it said it used the NFL 2024 fact book)

544 rosters but half as many games (because the teams play each other).

Technically I can probably do it in about 10 minutes because I've worked with these kind of stats before and know about packages that will get you this basically instantly (https://pypi.org/project/nfl-data-py/).

It's exactly 4 lines of code to find the correct answer, which is 2,227.

Assuming I didn't know about that package though I'd open a site like pro football reference up, middle click on each game to open the page in a new tab, click through the tabs, copy paste the rosters into sublime text, do some regex to get the names one per line, drop the new one per line list into sortmylist or a similar utility, dedupe it, and then paste it back into sublime text to get the line count.

That would probably take me about an hour.

Re: AI assisted search-based research works now

#80

Post by neom »

CSMastermind wrote: post #79
neom wrote: post #64

Earlier quoted context omitted.

Is it accurate that there are 544 rosters? If so, even at 2 minutes a roster isn't that days of work, even if you coded something? How would you go about completing this task in 1 hour as a human? (also chatgpt 4.1 gave me 2,503 and it said it used the NFL 2024 fact book)

544 rosters but half as many games (because the teams play each other). Technically I can probably do it in about 10 minutes because I've worked with these kind of stats before and know about packages that will get you this basically instantly ( https://pypi.org/project/nfl-data-py/ ). It's exactly 4 lines of code to find the correct answer, which is 2,227. Assuming I didn't know about that package though I'd open a…

I see. When you said "game day rosters for many NFL teams are available on many sites" - I thought "that sounds like a lot of hours!!" heh. - I didn't realize it was packaged well, I also know sweet fa about football. Thanks for explaining it more. :)
Post reply on HN