Meta is commoditizing their products' complement.[1]
OpenAI can't - these models (or rather, the APIs) are their product.
[1] https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/
OpenAI can't - these models (or rather, the APIs) are their product.
[1] https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/
The problem is that they have a for-profit arm.
If investors didn't get any benefits ("richer" in your parlance) why should they invest?
If they didn't invest where would the money come from? tax dollar? Can you envision tax dollar be spent on AI research? or would they bring ChatGPT as quickly or as capable?
Joel's reasoning requires that the complementing product be seen by the consumer as a requirement for consuming your product. Babysitters complement nice restaurants because parents need a babysitter to go. Gas complements cars because cars won't run without gas and the driver must buy gas periodically in order to use the car.
LLMs don't occupy the same space with relation to social media, it's more that they're quickly becoming an essential internal component for any social media site. Joel's reasoning doesn't apply to internal components that are invisible to the end user. A restaurant may benefit from finding a source of cheap lobster, but they don't benefit from publicizing that source to the whole world. Lobster is not a complement to restaurants, it's a component. LLMs occupy the same kind of space with relation to social media—they are something that every social media company would benefit from having cheaply, but not something their users need in order to consume their service.
Content creation is complement to social media, because content (videos, etc.) is shared on social media and in order to get it in front of people, you have to pay.
Platforms and Social media has replaced most of the world's ad surfaces, it's become THE way to get in front of people. Social media is a giant attention market but ultimately functions the same way as amazon: You pay to get your product in front of people.
LLMs commoditize content creation. Fewer people (after layoffs) can create more content. The money you save on laying off people then will be pumped pumped into boosting to get the content in front of people as the auction prices to get the right eyeballs go up due to increased competition.
Also on the Dwarkesh podcast, Zuck indicated one thing they’re afraid of is walled garden ecosystems they have to go through to reach users like with Apple and Google, and releasing open models is a way of preventing that happening with LLMs.
None of that changes is AI is commodified.
Ergo, OpenAI wins if they have better models. Meta wins by default if everyone has equivalent models: existing business unimpacted, more access to users.
For example Mobile Gaming: Mobile games are demand generation bound - you literally run fake ads on facebook, and only the games that convert well are made. Yes, that's why the fake game videos exist. Making games is no longer hard, you pay a bunch of chinese and get the game. The majority of a mobile game's budget, with few exceptions, is spent on user acqusition.
Now picture AI making it easier to make content. Games. Movies. Etc. It invariably results in more content (and less quality, but as we've seen with News, that's not Mark's concern and people who think quality is something consumers choose over commoditized volume haven't paid attention for the last 2 decades). More content means more demand for eyeballs on Meta's platform, higher ad auction prices. Higher user acquisition spent.
Lucky for you, making more content with fewer people is a good effect of AI. So you save on talent, you lay off people and ... then discover that because Meta made AI available to everyone, you're just going to spend the additional money you made to pay for ads.
They have masses of content generated for free by users and journalists and influencers and so on - if anything, a bunch of LLM spam is a threat to that.
However, Open-weights LLMs are a much smaller threat to Facebook than they are to Google (where it could replace a lot of search usage) or Open AI (whose business is selling LLM access)
Perhaps for Facebook the benefits of the open weights approach - where you give away the model and get back a load of somewhat improved models, a faster way of running it, and a load of experienced potential hires - pays off because it doesn't threaten their core business.
This is an overly narrow view of what an LLM can do. Generating text is the really neat parlor trick that people are trying to cram in to every possible startup, but if you take a broader view then what LLMs really are is the single largest breakthrough in natural language understanding.
Facebook doesn't need text generators, but they do need language understanding, especially for recommendation and moderation.
I'm not convinced that it's a complement—Joel's explanation is that you make a product that users consume alongside yours very cheap in order to keep people coming to you— but they definitely need LLMs.
The GPU cluster that they trained their Llama models on was actually built to train Reelz (their TikTok competitor) to recognize video content for recommendation purposes, which is the thing that TikTok does so well - figuring out users' preferences.
> A complement is a product that you usually buy together with another product. Gas and cars are complements. Computer hardware is a classic complement of computer operating systems. And babysitters are a complement of dinner at fine restaurants.
LLMs aren't really a complement like gas to cars because the end user doesn't need to consume the LLM in order to use the social media site. It's more like LLMs are becoming an essential component of a social media site—not like gas to cars but like an engine control unit, a part that ideally the user will never see or interact with. Joel's reasoning doesn't apply to that kind of product because users don't see the price of LLMs as a barrier to consumption of social media.
For instance, an Instagram account that shows cool AI generated photos generates ad revenue for Meta.