Hacker News Reader

Comment by simonw | Hacker News Reader

I worried about this a lot more at the tail end of 2003, when OpenAI's GPT-4 (since March) was still very clearly ahead of every other model. It briefly looked like control of the most useful model would stay with a single organization, giving them outsized influence in how LLMs shape human society.

I don't worry about that any more because there's so much competition: dozens of organizations now produce usable LLMs and the "best" is no longer static. We have frontier models from the USA, France (Mistral) and China now.

The risk of a model monopoly centralizing cultural power feels a lot lower now then it did a couple of years ago.

View on HN

Model competition does nothing to address monopoly consolidation of compute. If you have control over compute, you can exert control over the masses. It doesn't matter how good my open source model is if I can't acquire the resources to run it. And I have no doubt that the big players will happily buy legislation to both entrench their compute monopoly/cartel and control what can be done using their compute (e.g. making it a criminal offence to build a competitor).

Model competition means that users have multiple options to chose from, so if it turns out one of the models has biases baked in they can switch to another.

Which incentivizes the model vendors not to mess with the models in ways that might lose them customers.

I don't think anyone considers biases more important than, say, convenience. The model that only suggests Coca–Cola brands will win over the one that's ten times slower because it runs on your computer.

I don't think model competition is necessarily the fix to this issue. We're not even sure if the setup as it exists today will be the norm. It could be that other entities license out the models for their own projects which then become the primary contact point for users and LLMs. They are obviously going to want to fine-tune the models to their use-case and this could result in intentional commercial or ideological biases.

And commercial biases wouldn't necessarily be affected by competition in the way that you're describing. For example, if it becomes profitable for one of these companies to offer to insert links to buy ingredients at WalMart (or wherever) for the goulash recipe you asked for that's going to become the thing that companies go after.

And all of this assumes that these biases will be obvious rather than subtle.

Reply on news.ycombinator.com