Tuesday, May 28, 2024

As a Human, I’m Excited About AI. As an Investor, I’m Nonetheless Wrestling With a Few Questions

Riffing Off VC Charles Hudson’s Weblog Put up, Right here’s What I’m Making an attempt to Reply

a bunch of various robots working a race, digital artwork [DALL-E]

If startup founders typically ‘Construct in Public,’ is the analogou sventure capitalist motto to ‘Assume in Public?’ Anyway, there’s little question that the story of the trailing months has been Synthetic Intelligence. Over Homebrew’s first decade we’ve all the time been all for what we’ve referred to as ‘Utilized AI’ (together with Utilized CV, Utilized ML)— alternatives the place the know-how itself was being prolonged and commercialized for a particular function (contrasted with core R&D or base mannequin improvement). Firms comparable to Protect.aiKettle, and MasterfulAI, amongst many others, have been Homebrew investments which match this definition. Nevertheless it’s additionally clear we’re at a brand new inflection level the place our earlier hypotheses wanted to be up to date. So like a stone in a sharpening tumbler, ‘what are our ideas right here’ had been tossing round my head for a handful of quarters. After which I learn Charles Hudson’s put up, which prompted me [AI PUN] to only write this down.

In “Trustworthy and Naive Questions from a Generalist Seed VC Grappling with the Generative AI Revolution,” Charles (whom I like) touches on similarish matters to what Satya and I’ve been chatting about.

I. Base Fashions

  • Given workforce, knowledge, and compute prices, will the ‘value of entry’ and ‘value of innovation’ on base fashions improve or lower over time
  • Will completely different knowledge sorts produce/require their very own base fashions, and below what circumstances are these base fashions prone to be produced by completely different corporations/sources vs below a single company umbrella
  • How does one measure ‘high quality’ and what traits will base mannequin house owners compete on apart from ‘high quality’ [price, latency, privacy, etc]

II. AI ‘Middleware’

  • In a multi-base mannequin world, received’t there be some worth created by dynamically switching between fashions relying on the use case? Received’t most utility house owners who search to combine ‘AI’ be all for “greatest outcomes” extra so than having to decide on a mannequin upfront
  • Will this middleware layer have entry to sufficient mannequin attributes to even know when/the best way to handle between fashions
  • Can these corporations defend their margins or will they be topic to both (a) intense competitors pushing margins all the way down to ‘base mannequin question value + just a few foundation factors or (b) the bottom mannequin corporations behaving just like the document labels and mainly being very deliberate about taking nearly all of income created by a service constructed on high of their IP
  • Will middleware have the ability to increase the bottom fashions with new proprietary knowledge with a purpose to create a differentiated product
  • Will middleware corporations search to combination proprietary knowledge sources with a purpose to enhance base fashions in distinctive methods

III. AI ‘Native’ Functions

  • What are the circumstances by which the addition of AI catalyzes new product choices constructed round this know-how versus ‘AI’ being a function that the market main purposes can construct into their platforms. Will Zendesk get replaced by an AI Buyer Assist startup or does Zendesk combine AI. Repeat this query for all the pieces B2B.
  • OpenAI is a for-profit, working a enterprise fund, and so on — what sorts of ‘partnership danger’ is there in backing alternate options who’re competing with OpenAI funded startups. Are all the bottom fashions doing to make use of their money to attempt to develop their very own ecosystems and implicitly/explicitly attempt to decide profitable apps?
  • What is going to the engineering groups at ‘non-native’ adopters must appear like with a purpose to efficiently combine, handle and compete with the native apps
  • Will companies who consider they’ve proprietary knowledge to might assist enhance base fashions have the ability to promote that knowledge and/or ‘pay it in’ to the mannequin in return for discounted utilization? Will they search to create improved layers atop the bottom fashions

When you’ve got POVs right here I’m all the time completely satisfied to listen to from you [hunter at homebrew dot co]! Keep in mind, we make investments our private capital (usually a $100k-$500k preliminary funding, though capacity to go bigger when applicable) in your corporations after which get to work supporting you.


Related Articles


Please enter your comment!
Please enter your name here

Latest Articles