4 ms·

There is no moat. https://newsletter.semianalysis.com/p/google-we-have-no-moat-and-neither https://newsletter.semianalysis.com/p/google-we-have-no-moat...

livinglist 5mo ago
The article is from 2023, I’m wondering if things mentioned still stand true today, can someone pls let me know.
- raincole 5mo ago
  It's much truer today. You can say that article is extremely insightful, as it predicted today's open weighted models scenario 2 years earlier.
  - energy123 5mo ago
    It remains an unproven hypothesis. The revenue of the top 2-3 labs is still growing nearly exponentially, which is the ultimate piece of data that settles the question empirically for now. Benchmark scores aren't really proof. Benchmaxxing is possible, for example. Only revenue numbers (and gross margins) count.
    - quikoa 5mo ago
      The ultimate piece is not revenue but profit. At some point these enormous investments will have to be earned back. Good luck with that when open weight models are also continuously improving, have cheap providers and for many are already very usable.
      - DrScientist 5mo ago
        The other point to make is that companies are starting to worry about the risks of externally hosted models. This is at multiple levels if you have a remote API call as a key part of your workflow/software system. 1. Price risk - might be affordable today - but what about tomorrow? 2. Geopolitical risk - your access might be a victim of geopolitics ( seems much more likely that it used to be ). 3. Model stability/change management - you've got something working at the API get's 'upgraded' and your thing no longer works. If you are running on open weight models - you are potentially fully in control - ( even if you pay somebody to host - you'd expected there to be multiple hosting options - with the ultimate fallback of being able to host yourself ).
Gigachad 5mo ago
It’ll be priced slightly higher than the cost to actually run. But it’s still not clear what the real cost of the big models is. They seem very subsidised, but by how much?

livinglist 5mo ago

The article is from 2023, I’m wondering if things mentioned still stand true today, can someone pls let me know.

raincole 5mo ago
It's much truer today. You can say that article is extremely insightful, as it predicted today's open weighted models scenario 2 years earlier.
- energy123 5mo ago
  It remains an unproven hypothesis. The revenue of the top 2-3 labs is still growing nearly exponentially, which is the ultimate piece of data that settles the question empirically for now. Benchmark scores aren't really proof. Benchmaxxing is possible, for example. Only revenue numbers (and gross margins) count.
  - quikoa 5mo ago
    The ultimate piece is not revenue but profit. At some point these enormous investments will have to be earned back. Good luck with that when open weight models are also continuously improving, have cheap providers and for many are already very usable.
    - DrScientist 5mo ago
      The other point to make is that companies are starting to worry about the risks of externally hosted models. This is at multiple levels if you have a remote API call as a key part of your workflow/software system. 1. Price risk - might be affordable today - but what about tomorrow? 2. Geopolitical risk - your access might be a victim of geopolitics ( seems much more likely that it used to be ). 3. Model stability/change management - you've got something working at the API get's 'upgraded' and your thing no longer works. If you are running on open weight models - you are potentially fully in control - ( even if you pay somebody to host - you'd expected there to be multiple hosting options - with the ultimate fallback of being able to host yourself ).

Gigachad 5mo ago

It’ll be priced slightly higher than the cost to actually run. But it’s still not clear what the real cost of the big models is. They seem very subsidised, but by how much?