Doing that in a way that actually makes search better is hard (and ACS isn't that at all)

It's worse than Splade https://i.imgur.com/oGliEIg.png

We've tested various hybrid approaches as well, but that's too much to go into in once post.

what’s that output from and how can i understand it? i’m happy to rtfm - just want a pointer. thanks!

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.

It's almost certainly the RLHF, not the base model.

But the base model, when its trained on the whole internet, will have some extreme biases on topics where there's a large and vocal group on one side and the other side is very silent. So RLHF is the attempt to correct for the biases on the internet.

> So RLHF is the attempt to correct for the biases on the internet.

...or it can be used to reinforce a specific ideology. Completely dependent on who does the RLHF and what their motivations are.

I think a model could do some basic eval but there are too many hidden assumptions for it to do especially well.

OAI function calling can solve this more or less

Seeing a lot of folks criticize the claim about needing side projects/GitHub. I think if you already have experience it doesn't matter too much, but if you have a nontraditional background/lacking experience, it is totally the difference between an offer and never hearing back. I've experienced this from both sides of the table.

This is just not true.

Tons of developing countries have super aggressive financial/general regulations, for example India requires all data be stored in-country so you can't shard across geozones [https://www.lawfareblog.com/key-global-takeaways-indias-revi...]. In fact, the countries with the most complex financial compliance are almost all developing [https://www.bnamericas.com/en/news/five-of-10-most-complex-c...].

But the bigger thing is generally how advanced the existing finance infra is. Lots of countries have really slow bank transfers, processing a refund requires handing in physical paperwork, etc. Stripe relies on having underlying bank infra that's somewhat functional.

As far as the EU, the regulatory environment is manageble, but the payment landscape is very different from the US where Stripe originated. For example, in Germany only around 30% of users pay for stuff online via card [https://askwonder.com/research/german-market-what-s-breakdow...], that means for every country a ton of new payment methods have to be added for the service to be actually useful.

They already have. https://support.google.com/reserve/answer/7514650?hl=en

Coverage is spotty. It's crazy how much people here underestimate the lack of willingness of many business to use these systems. I live 45 minutes away from SF and recently got a haircut at a decently popular place and it was cash only. Adoption of new technology by small businesses is really slow.

I agree with the sentiment about slow adoption, but in this case, it might not make sense to switch.

If it's decently popular and is fully booked, going cash only would reduce their costs and either give them healthier profits or let them charge less. Most businesses accept cash and card, so they likely already have the daily task of taking money to the bank or having it picked up by a secure truck.

I'd suspect a cash-only business of tax evasion before technophobia.

They already do this actually, but the coverage is pretty spotty. There are still restaurants that don't have websites and you're asking them to set up a digital booking system through Google. It's just not going to happen. Also, I see this tech as more of a way of training up their AIs on tasks that provide a bit of value while they wait for people to connect to the official booking APIs and such.