[erdevs]

This is yet another clickbait-y title that overpromises and underdelivers. Little value, no specificity, and hardly any real description of a "formula" whatsoever. Here, though, is a tl;dr summary for one of the only concrete pieces of the "formula" actually described:

Bloomberg's early-stage VC fund will only invest in Series A or earlier companies. Then when looking at AI companies they ask two questions: "1) Do you have access to your own user? 2) Do you have access to a data set that's yours? If they have both those things, then they can create a virtuous cycle where the user contributes the data, the data gets better, and it makes for a better user experience."

Asking about access and usage rights to data and means of applying learnings from the data is a very obvious qualifier, so there is little valuable insight here.

[tlb]

Many AI startups don't have (1) because they provide a service to someone else who has the relationship with the user. For example, startups that provide APIs for image classification, e-commerce fraud risk estimation, ad targeting, or product recommendation. I believe there will be big winners in the API space too, but they'll look different from consumer-facing companies.

[cocktailpeanuts]

It's hard to become extremely successful (I'm talking Google/Facebook scale) if you don't have access to your users. It's much easier to build a consumer product that provides an experience 10 times better than what's on the market, than building a technology that is 10 times better than what's on the market. And even when you do come up with something like that, unless you actually capture your own customer (which brings us back to "having access to your users"), others will end up copying you or commoditize you.

[nostrademons]

They still have (2), right? An image classification startup that provides services to other companies will see more images than competing firms, which lets them train their models on more data, which gives them better quality models than competitors. Ditto fraud detection, ad targeting, or product recommendation (within a specific vertical, at least).

It seems like it'd be hard to build a durable competitive advantage without either owning the user relationships or owning the data, correct? Such a company would be just an algorithm and some tools around it, which are relatively easy for a competitor to copy.

[erdevs]

Yes, I only meant it's an obvious question to ask to qualify the sort of AI startup. Agree there can be winners without this qualification but it's an obvious distinction to be noted.

[ganeshkrishnan]

Some AI cannot have access to data set that's "ours" because it makes little sense.

We are trying to cluster user personalities and match it with our text corpus cluster in order to find the best documents matching the user's personality. For us Wikipedia raw data suffices since it's counterproductive to use proprietary searchable text corpus.

While the VCs viewpoint is totally understandable from their perspective, always chasing money & valuation gets aggravating for us startups.