And it will never be, because they will be always worried about PR, about regulators, about cannibalizing legacy business, etc. A new player who isn't held back by this has a good chance of disrupting the market with inferior technology. It happened over and over in the history of tech.

I'm sure there were quite a few SGI, Sun, and IBM executives laughing at that amateurish thing called Linux...

This is a fair observation, and it’s certainly possible that they get disrupted like this. (I find the “cannibalizing legacy business” fear most plausible of these).

However I question your level of confidence. The idea that a company is incapable of avoiding being disrupted is pretty dubious; now that disruption theory is well understood by all executives, it’s possible to take steps to avoid it.

For example, DeepMind is an Alphabet company, and they could push them to make chatbots profitable completely ignoring Google’s ad market. They could even transfer tech/people over and to give them a boost in productionizing their efforts.

> And it will never be, because they will be always worried about PR, about regulators, about cannibalizing legacy business, etc.

They don't have to completely come up with a ChatGPT clone. They could do some of the following things:

- Enable some use cases on Google Search - for searches which are purely information based - above the search results. They already show such cards right now.

- Integrate it with Google Assistant. They already have excellent voice recognition devices. Assistant responding with generated answers, would be a game changer. You don't even have to type anywhere.

Except it’s been reported that it lead to a board meeting and since Sundar has said that Google will pivot in some regard internally because of it. So it doesn’t matter if they beat it or not— it’s existence is a tangible threat.

I would not be surprised if they pivoted to providing either demo access to their models, or otherwise lowered the safety bar for publishing them given the mostly-positive response to ChatGPT. Just because they see a need/opportunity to pivot doesn’t mean they are actually threatened. It can also mean the opposite.

Previously they were playing their cards close to their chest because they were terrified of getting crucified by the NYT for off-color quotes from their models.

> Google’s chat bots are way ahead of ChatGPT, at least from what we can see from the outside.

And there's the project. If the public (developers and/or end users) can't use it, it might as well not exist.

> ChatGPT is clearly nowhere near being ready for actual product use.

The only thing preventing ChatGPT from being used in production is that it's not exposed through a proper API. It's a demo preview.

There are lots of produts out there using GPT3 right now, and they will all benefit from switching to whatever next iteration (as result of feedback from ChatGPT) is.

> If the NYT really thinks Google is somehow caught unawares here, they clearly have no understanding of what Google’s research program looks like.

Journalists always exaggerate, but research is not production, and in the case of Google the difference is painfully obvious.

> Google’s chat bots are way ahead of ChatGPT, at least from what we can see from the outside. (Nobody has mistaken ChatGPT for a sentient being, but they did with LaMDA.)

Google’s search technology is way ahead of anyone on the market, yet their search has become garbage because of the marketing choices and UX decisions they are making.

Them having a technically superior competitors] to ChatGPT doesn’t automatically mean it would be a better product.

As Steve Jobs famously put, Google has great engineers and awful products. Or something like that. Anyways, even if they do have the better technology I can see infinite ways they fail to produce a convincing product. I can also see many ways where they succeed and become even more dominant than they are now.

Time will tell.

Models are inherently more portable than the Google index. Think of Stable Diffusion, anyone can generate images on their own computers. When we get to the SD moment for text, then what will Google do? They can have better models, but people can have decent models as well, loyal models, that maintain privacy.

> The reason Google hasn’t released this product yet is they (unlike the NYT) well understand that it’s not ready yet.

It is ready, and Google is losing. Google having a better product is meaningless if they refuse to release it. I am replacing Google with ChatGPT for about 20-30% of the things I used to use Google for. That percentage will only go up as OpenAI keeps improving their product and Google continues to drag its feet. Google has to move quickly here.