Discussion about this post

User's avatar
Tetra's avatar

If Big G allowed sites to spam AI content they would destroy their business model.

The benefit of searching on Google for the user is getting unlimited perspectives on different topics. If everyone just spams ChatGPT then there's only one perspective.

If all the information on the internet is generated by chatGPT then there's no incentive for anyone to use Google instead of going directly to the source.

Google has the smartest people in the world working for them. They won't let that happen.

This will end terribly for everyone using chat gpt on their site

Expand full comment
@member's avatar

A bit of commentary from the trenches:

1) Google is not the "place where the smartest people in the world work". It's Silicon Valley tech company like any other - pay well, hire a lot of woke capital, run some really exciting projects, mishandle investments badly into bad ideas and pull the plug late.

Fundamentally their success in the last 10 years revolves around google.com, reasonably good play into the market with the GSuite for Work and half-hearted performance for GCP. All of these have some glaring holes which have allowed for other players to push them hard:

- Google is becoming less and less useful for practical searches, with several different information venues attracting interest in recent years (whether alternative search engines, the "tik tok as search" teen crowd or now chat gtp). The real dominance is serving web traffic via SEO, but thats only one angle of the story (if very $ valuable).

- GSuite for Work was one of the big winners of the pandemic, with a lot of adoption in the tech scene, together with other SaaS work tools like Zoom and Slack. That has mostly cooled off now due to the revamp of Teams, which is slowly going back to the "default" choice. MS's dominance on work tools is extremely difficult to overcome and they've invested smartly into improving the quality of the apps and rolling it out on any platform.

- GCP and the GCP Marketplace is an also-ran in the cloud space. AWS is the default choice, with Azure going hard at it as well. GCP is competitive and arguably has an interesting set of native tools on the platform but they continue to be very slow at integration of external apps. They try to cover it with rather approachable GTM initiatives with external partners but so do the rest of the players. To give you an idea of how silly some of their decision making can be - if you want to onboard new resellers for private offers between you and the end customer (i.e. big ticket business), if they were registered until 1/1/2022 they can't allocate margin to them since they stopped onboarding the new resellers. They stopped onboarding them since they were keeping the margin allowances per reseller and sub customer on a manual spreadsheet and the process completely grinded to a halt. They were supposed to fix it 6 months ago, it's still not fixed. They launched a full new private offer platform and STILL didn't fix it.

2) Most of you are looking at the topic from an end user perspective and e-Commerce/twitter creatives hustling for crap. The actual big money topic is related to the infrastructure required to run this and how do you scale it across applications. MS is making a very aggressive bet here of utilising Azure capacity (which atm is very valuable) on running this at an affordable scale and baking it in into native apps.

Bing getting Chat GPT integrated is one thing, adding it to Office 365 is an exponential power bet for both customer retention and new subs (both private and corporate), putting it as advanced ML tool you can implement within your apps natively and run them on Azure is another upside for going with a MAC committed spend (1M+ annual cloud commit to MS). Adding integrations further into LinkedIn, GitHub, etc further changes the info value of those products.

Ultimately the outcome will be that if you are a user and you want to find information, you can go on google and get ads and dubious results, or you can use a search bar integrated in most platforms that MS servers and get low ms usable result on the spot (powered by Azure infrastructure). One way or another this will divert significant traffic away from google.com as a platform and a lot of websites relying on that traffic.

As other players start to make their own plays into it (AWS will not sit on the sidelines), who is willing to reserve capacity and take a margin hit to gain market share will become a big topic of deploying this type of ML into applications with significant downstream effects. There is not enough hardware long term to cover all of the computational needs of the economy, and that was based on existing usage. Big play into AI powered search will eat a lot of cloud capacity, both due to the ML jobs as well as the expectation for fast search results (several miliseconds at best). That means hot machines, which means the most expensive hardware to keep the service running AND do relevant ML jobs on additional hardware.

3) Google is dominant in search - it's dominant in one subset of search which is indexing the web and moving traffic around for money. The majority of day-to-day applications doing search and running in the backend are powered by Apache Lucene open source forks and utilized in a variety of use cases. Google might be referring traffic to H&M webstore, but the actual search on the store, keeping the store running, storing the backend dataset of SKUs and tracking usage, fulfilment backend search etc it's all other platforms.

It might be contrarian, but this is more of a "the emperor has no clothes" moment rather than Google can just flex around and do whatever they want.

Expand full comment
45 more comments...

No posts