=OpenAI Exec Laments That China Is Giving Away Models So Good That For-Profit Companies Won’t Be Able to Compete
“Last week, a Chinese-made large language model called Kimi K3, developed by Beijing-based firm Moonshot AI, burst onto the scene. The open-weight model impressed with early benchmark results, trading blows with some of the most advanced closed-weight ones being developed by the likes of OpenAI and Anthropic — and at a fraction of the cost.”
Lest we forget, there is BIG money in oil, pharmaceuticals, AI, etc, and those executives do not like their profits being undercut by free community projects. Big money usually locks users in by glossy adverts and other enticements, and if those fail they can always use scare tactics about hidden Chinese backdoors.
It's true that cloud services running AI do require big investments, but when better things come along which are cheaper, the market moves as this is what capitalism is about. And in fact, this is what innovation is about. You innovate, or you become irrelevant. This is often what post offices have found out about as courier companies started overtaking them.
The US approach seems to have been about throwing more money and resources at AI, whilst the Chinese appear to have been innovating around doing more with fewer resources. The same goes for open source projects where many local AI models are running on smaller home labs and PCs. This how quantisation, GPU offloading, flash attention, context length tuning, etc all came about.
No proprietary AI business can either lay claim to any particular copyright of content as all or most of that content was all sucked up from across the Internet by them in the first place.
Open source is about collaborating to solve challenges, and resharing and building on what others have done. Yes not everyone can afford to run a mainframe computer at home, so the business model for AI as a service may well be a valid one going forward. There will be lots of room in the market for “selling AI services” because most people are not going to buy massive GPUs at home and do not want to tweak lots of settings, and the one that Microsoft scores on, is too many users can't decide what Linux distro to install (read: can't choose what AI models to use).
OpenAI though would do well by competing better instead of screaming foul at their competitors. Maybe they do need to lighten up on the resource requirements. No city wants to see their dwindling drinking water, or power grid, swallowed by ever more and bigger data centres.
I do remember about two to three decades back, when I was a PC programmer, that we had a very expensive IBM mainframe that needed upgrading because of resource constraints with the systems we were running. Then someone had the bright idea to try optimising the source code of the biggest systems. Turns out we saved over 30% on resource usage and no longer had to upgrade the mainframe! Sometimes working smarter is better.
Today with digital sovereignty becoming more and more strategic to many countries, I think they will be looking more at what AI models they can run locally themselves. AI is already being weaponised to breach vulnerabilities, and to be weaponised through poisoning sources etc. It is all the more reason why you want to run our own and set your own guardrails in place. Running your own AI also means your company or country data stays local.
Is it possible? Yes, of course it is. Before we had Linux, we all thought we had to run Windows to get anything done, but we've realised that was a steel box that was built around us by Microsoft's PR (to protect us from the cancer called Linux). Turned out though it was not cancer after all, it was the dream of having freedom and being in control of our own OS without their telemetry.
Hopefully AI never becomes another Microsoft, with only one player dominating the whole market. Right now it is looking good with many AI cloud providers (OK mostly in the USA and China) but at least it is not in just a single country, and we have local AI models doing very well at more specialised tasks. It needs to stay open with the providers competing to provide better services for less. If one or two players fall out along the way because they are just too expensive on resources, then so be it.
See
GadgeteerZA
OpenAI Exec Laments That China Is Giving Away Models So Good That For-Profit Companies Won’t Be Able To Compete
“Last week, a Chinese-made large language model called Kimi K3, developed by Beijing-based firm Moonshot AI, burst onto the scene. The open-weigh...
#ai #opensource #technology
#ai #bots #technology
