Is ChatGPT a “Code Red” for Google Search? | Hacker News Reader

Is ChatGPT a “Code Red” for Google Search?

garymarcus.substack.com

21 pointsmirzap8 comments

The reality is that ChatGPT, while not yet advanced enough to surpass Google, is just the beginning... Prior to ChatGPT, it was difficult to imagine a world without Google's dominance. Now it is very clear that Google has a real threat.

It's not just about ChatGPT version 1. It's about the potential for chatbot technology to bypass Google altogether. While the answers provided by ChatGPT may not be perfect right now, what happens when they are close to perfect? Why would you use a traditional search engine when you can get the information you need in a far more efficient manner?

Google isn't at code red right now because of ChatGPT V1 but because of the next 10 years of chatbots.

For me, I have already changed my searching behavior for a quite a few (not all) queries .. A lot of my queries these days go ChatGPT -> Brave Search -> Google. Google being third down the list for me.

It's not yet. But it clearly shows the direction where search is going, and the ways in which we will likely ask for information.

Classical search enhanced with AI algorithms is probably not the future. And Google invested a lot in it. It's unlikely that Google will be fast at adapting (changing) their business model.

I think if anything we're at a crossroads. Natural language is great for some types of queries, but has its limits for others. Internet search engines have come to be used for a lot of different tasks, and a large part of Google's sticking power has been that it's at least decent at most of them. But that may also be a bit of a problem. Based on everything I've seen so far, it appears that the more conversational a search engine is, the worse it tends to become at the task search engines originally did (i.e. finding documents about a topic). We've also sort of started to perform all sorts of things through this medium of document-finding, that perhaps could be done differently. I think it does a disservice to both those other tasks and the usefulness of document-finding.

I heard Google leaders say a few years ago that conversational AI or search is not a priority. I was a bit amazed by their thinking and always thought Google is a bit disconnected from making great products and understanding what customers really want or need, even if they do have the best tech and talent usually in the field of AI. Conversational AI has been such a dream of mine to witness and popularized in sci-fi cultures and here Google was saying that it doesn't feel that important as part of AI innovation.

I think that kind of decision making led to working on language models but not for the goal of providing fundamentally better search or intelligent companionship to end users, but more as an addendum to their existing products like Search, Youtube, cloud offerings etc.

Now that there is a huge public hype around ChatGPT, they have been forced to revisit their priorities around conversational AI as a standalone customer facing direction.

The main challenge here is obviously how to ascertain what is true. I am not sure much work is done in AI research yet or how hard of a problem that is. And that's independent of AI chat as you could apply the "truth" model to any piece of text like fake news, tweets and so on.

Without that probabilistic truth model, it is going to be difficult to offer reliable conversational AI or search that one can rely on beyond entertainment or partial/subjective advice. I have no idea how far ahead Google and OpenAI is for "truth modeling".

I think a good metaphor for AI progress is 3D graphics.

We've definitely passed the Castle Wolfenstein era of AI, and are solidly in the Doom era: It's not actual real intelligence yet, but it looks and feels a lot like it. One could even say we're approaching Quake level AI, where you can actually look up and down. It's still just mostly smoke and mirrors though, just like early pseudo-3D, making you feel like the system is rendering everything in three dimensional space, were it's mostly just flat images created to look like there's more depth than there actually is. Once you clip through a wall, you realize the limitations pretty quickly.

There are some Mario 64 level AIs out there, showing what's possible, but they're very niche systems so far. And again, you quickly realize the scope of what's possible isn't as big as it seems at first.

But we're definitely going to get there, and it's going to happen fast. When we get to the AI equivalent of real time ray-tracing, ambient occlusion, etc. things are going to get really interesting.

Here's the other five threads:

https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

Not really. This is attempt to refute that NYT article that was posted 5 times.

No.

These chat bots are just "agreeable". They can elegantly create agreeable nonsense. They can't synthesize data that is not already inferable from their main knowledge. They're modern day Mentats.

In order to replace Google Search, they would need to put the answer in front of you (or some data from which you can infer the answer).

Reply on news.ycombinator.com