Hacker News new | past | comments | ask | show | jobs | submit login
Neural Search Frameworks: A Head-to-Head Comparison (dmitry-kan.medium.com)
52 points by bobvanluijt on Dec 23, 2022 | hide | past | favorite | 10 comments



Google's golden goose may be cooked.

The two things they've got keeping people on their ad product are search and (hypothetically) anti-competitive practices that may be hit by antitrust judgments.

If Google looses search ad revenue, the company is going to be put in a very precarious position where a lot of their scale and largesse becomes difficult, if not impossible, to maintain.

Amazon, Apple, and Microsoft have a healthy mix of revenue streams. Google does not. For Google, the entire story revolves around search ad revenue.

All Apple needs to do is switch to using Bing. An antitrust judgment against Chrome by default on Android opens the door for alternative browsers and search engines. And a judgment against Google search by default on Chrome sinks the ship altogether. Both the EU and the US are looking into this, and I wouldn't be surprised if one or more of these things happens in 2023.

Google search quite frankly sucks. A lot of this is spam at scale, but there are also increasingly perverse incentives that keep Google from delivering an S-tier product. They make their money diverting your attention to the highest bidder.

There are so many search startups getting off the ground now. The new breed of search is going to leverage AI/ML to do better than anything that came before. Paul Graham identified this as a promising attack surface, and he's totally right.

This is incredibly good opportunity for startups. Not just search, but anything else that might chip away at a Google product that they don't really care about. That they might not be able to afford when their core revenue stream begins to falter.


I think you're quite possibly correct. ChatGPT has already replaced at least 25% of my search volume despite being a tech preview.

Nothing has been able to disrupt Google so far but there's no reason it can't.


Ironically, you might be increasing Google's profitability if the searches you're taking off their platform aren't really monetizable.

Ex: "what is the tallest mountain" isn't really monetizable whereas "best bbq restaurants in Nashville" is very monetizable


> whereas "best bbq restaurants in Nashville" is very monetizable

And that's precisely why I'd ask ChatGPT instead. Until they become too obvious in monetizing that.


I think this is an incredibly farfetched position. Google has more hardware, several orders of magnitude more data, highly-developed experimentation capability, and at least equivalent ML expertise compared to anyone. If something chat-GPT-like ever actually starts to take off as a product, Google will clone it and crush it.


Also, the core algorithm behind chatGpt, Transformer Neural Nets, was actually built by Google in 2017. They'll be fine


Inventing something doesn't mean you can't get killed by it.

https://en.wikipedia.org/wiki/List_of_inventors_killed_by_th...


I wouldn't put Google out of the AI/ML search race. They are on the forefront of LLM exploration with players like OpenAI.

And I would not underestimate the power of incentives for injecting advertising into LLM style search.


Dmitry is one of my favorite writers/speakers in this space, definitely also check out his Vector podcast if this area interests you.


thank you, Doug! It is mutual. Had a fantastic discussion with you on the podcast: https://www.youtube.com/watch?v=Kpua1Euc-B8




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: