While the misuse of the term monopoly is annoying, that's quite the misleading comment. You're allowed to disagree with what policy is a good idea, but laws like the linked do exist, and they were seen as pro-competitive in their time.
Today you can consider that "just business", and therefore "part of competing", but the were laws with the intention of allowing/disallowing types of tactics. E.g., you can't compete by leveraging your volume (see attached link), so you have to focus on making your service good.
The assumption made by a these people is that if there are so majy companies getting that big, there must be general disrespect of these laws. And correlating with the undisputed fact that antitrust enforcement did change.
It sounds like you're focusing on the problems of running local models, or running models yourself, but I don't think that many people seriously expect near term improvement on that, it's definitely more just hopeful thinking there. That's not what I meant to address, and I also am in more of a "wait and see" mode.
But at this point we do expect that open weights _hosted_ options become feasible for the tasks they're using the frontier models for. And because of the lack of "legal monopoly" (intellectual property of whatever kind), they're way cheaper, not mention more flexible.
The launch of the tinker platform from Thinking Machines is an example of the "more flexibility" part that people want (and they chose to make their model open weights, maybe because this is the angle they want to push).
At this point I think it's realistic enough that the ball is in OpenAI / Anthropic's court to figure out how to respond to this threat to their business model.
That said, I think it's concerning that there are apparently only a couple of providers of hosted open weights inference, due to the complexities of doing so (per Dax from OpenCode's tweets).
While preserving problems is undoubtedly a natural incentive, I think Hanlon's razor applies here. Just today I was reading Competent Bureaucracy - Rebuilding State Capacity (<https://cdn.sanity.io/files/d8lrla4f/staging/cf7eedaf5d21d27...>) on the topic of agency structure promoting success (the author has done a nice amount of work in the past - e.g. https://www.statecapacitance.pub into this history of this topic).
> Take a task, any medium-sized task, decently scoped that you'd trust to give to Sonnet to finish without a hitch. Now give it to ANY open-source frontier model and watch them struggle and go in circles while failing tool calls and randomly assuming things.
Claude used to be much worse than it is now, just as bad the open weights models are. And the open weights were worse. The labs will also try to keep the lead, but at some point people start seeing real value from open models. Maybe you say they're not ready yet for medium tasks, but everyone sees the writing on the wall.
I recently read Economic Policy and Law in America (by William Letwin) and he discusses a bit (in the context of the early slow enforcement of the Sherman act) the fact that ultimately a prosecutor has limited resources, and any allocation comes at the expense of something else, so priorities absolutely do matter. And the prosecutor's office will coordinate with other law enforcement of course...
This reminds of this article which contends that this mistake has carried across time, in every era people described the workings of the brain similar to the machines they knew how to build, which is why we have "ridiculous" descriptions that changed over time.
People with lots of power and clout contribute to making predictions reality by all making predictions that make similar assumptions. The predictions become more likely because people already behave as if they're likely.
Marketing people like the features they're getting, and Google and Meta are dominant, so big that they're the default, in the same way that we talk about github being the default option, and "no one ever got fired for choosing IBM / (big tech company of your choosing)". I wouldn't dream of saying they should choose something else, without researching and guaranteeing that nothing they'd ever want from GA (and they may not know everything they'll want in the future right now) is missing in the alternative. In a role (marketing) that's completely out of my wheelhouse. So I don't even bother.
Depends on exactly how the project is managed. Older python tooling (`pip` module) doesn't have a native mechanism to differentiate between the spec (direct dependencies) and freeze (all dependencies, including transitive).
The important part was the following paragraph(s) that explained why this coupling is a compelling problem. It's not the same as just having a platform API.
This reminds of the conversation the other day about the deleted production database at railway. "this person obviously didn't follow best practice of being hyper distrusting of LLM agents", and the response "yeah but every company is marketing it as safe. someone is gonna fall for it".
Or not, because telling the agent is misbehaving may predispose it to misbehaving behavior, even though you point told it so to tell it to not behave that way.
I remember this discussed when a similar issue went viral with someone building a product using replit's AI and it deleted his prod database.
Socials: - github.com/mosheduminer
Interests: Books, Hiking, Outdoor Activities, Philosophy, Robotics, Web Development
---