Hi, hier gibt’s ja ne Biografiefunktion.
I’m not sure making the tool you’re selling less useful is a smart idea. Hmm, yes liability is an issue. But other models are just gonna allow it anyway, so people will just stop using Claude for that usecase. Reverse engineering won’t stop. Claude will just not be used for it anymore
Moment mal! Wir sind hier doch in einem deutschen Unter!
I am, on my main machine. Even at a DNS level. Can’t quite do the same for my phone. It usually blocks everything but this website wasn’t for some reason.
I should really create a custom env for my TV.
What is this terrible ad-riddled website? Second time I’m seeing it today.
A lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we’re going to be able to run them on our computers, and so we’re not going to need all the stuff in the data center.
We’re going to be able to? That’s wrong. We already can. Right now.
https://github.com/Niko1221/Strata https://github.com/ggml-org/llama.cpp https://github.com/earendil-works/pi
There’s going to be use cases for data centers but I don’t think it’s going to be consumer LLM inference.
What can be run locally on mid hardware now couldn’t be run half a year ago. Imagine one year from now.
Newsom, in his veto letter to the legislature, said he was supportive of “protecting sensitive personal information of Californians,” but that a “categorical ban on sharing that information goes a step too far.”
The governor pointed to existing state laws that require businesses to give consumers the option to limit sharing their information and budget constraints that would hinder implementing and enforcing the law.
Newsom also signed a handful of other bills strengthening the state’s privacy rules:
- SB 923 expands the scope of the California Consumer Privacy Act to require businesses to delete all personal information they have on a consumer upon request, not just data collected directly from the consumer. Online-only businesses will be required to provide a webform or portal for deletion requests.
- AB 883 shortens the time with which data brokers must comply with a data deletion request from 45 to 30 days. It requires elected officials and judges receive information on how to access the deletion mechanism, known as DROP, to protect their information.
Other data privacy laws still awaiting Newsom’s signature or veto include:
- SB 690 would remove the private right to action for alleged violations under a wiretapping law, allowing just the state attorney general to pursue such legal challenges.
- AB 2561 would prohibit an operating system or application from re-adjusting a user’s privacy setting without consent.
They say they are about law and order but oftentimes aren’t.
Shoop da whoop, my fren.
So we’re getting open source Windows?
Actually, it could prove fatal to assume that the Republicans are going to lose the house(s). We should assume that they won’t and do everything so that they actually do.
But I do think that Putin is seeing Trump’s usefulness for him fading, even if we’re just looking at a potential defeat of his party.
He’s doing everything in his might to lose the midterms even harder. That’s the last Putin is gonna get out of him and he knows it.
Spring this year is like ages ago in terms of LLM development. I suspect that you would need to do considerably less checking now or none (it checks itself) with an up-to-date agent, if you use your harness correctly.
Let me put your point into perspective:
“Try calculating without a calculator and you’ll quickly realize how dependent on it you’ve really become. I haven’t used it at all since spring this year and it’s been great. I actually get more work done without it since I don’t have to double check the calculations anymore.”
I think overdependence on LLMs is a valid concern, not restricted to coding. Studies would indicate as much. It’s good to be able to write code yourself. But I wonder if that is really a skill that is going to continue to be necessary in the future. Being able to do big calculations in your head or on paper is a nice skill to have, even now, but it’s honestly not absolutely necessary in today’s world. It’s good cognitive training, so I would still recommend it and basic arithmetic skills are necessary to understand higher maths and be able to use a calculator.
As I’ve pointed out in another comment, LLM companies are terrible for a variety of reasons and I despise their business models, but saying that the models are not useful is a notion not rooted in reality, imho.
That’s a valid point. I was just getting at OP’s point that “AI code is terrible”, which it can be, but doesn’t have to be.
Something has to be done about data centers, especially in the US. But that’s also why I believe promoting local inference is important. My local model runs on the sun on my roof, for the most part. It’s not causing environmental havoc. The training of new frontier models is. The training of new proprietary frontier models is disincentivized if you can’t make money with them anymore. Open-weight models are free and hurt Anthropic and the likes. But it’s also a matter of energy infrastructure and how the centers are built. That needs more regulation. It’s good that local backlash seems to have an effect.
LLM agent code is good if you don’t try to one-shot things. Have it devise deterministic tests to test the code and give it the ability to check its own results via screenshots. Have it plan ahead, ask questions, work step by step in separate sessions. Give it enough context, always have it create a README.md explaining the structure and purpose of the project for other sessions. Fix some stuff yourself. It’s not either or, it can be both.
American LLM companies are terrible for their copyright infringement, the environmental effects of their data centers and for several other valid reasons. I hated that Microsoft tried to push Copilot on every system because it sucked (mandatory “I use Arch, btw.”).
So yes, LLM companies are terrible, people losing their jobs because of questionable business decisions is terrible. But LLMs are good tools. Like any tool, you can produce utter shit with it. You can use a screw driver to destroy the drive of an old screw and make it unusable. You can stick it in a wall and have a gaping hole. But you can also use it to remove and fasten screws regularly. The same applies to LLM agents. If you have got no idea what you’re doing, you may produce shit. The thing with shitty AI code is, you only notice it when it is shitty. Because the output that’s good is essentially invisible. The problem is not the tool, it’s who’s using it.
Use open-weight models locally and use a harness like Pi and expand it or modify it as you wish. Don’t support the per token pricing/subscription companies.
I think yawning, mold and the movement of sand dunes are pretty interesting.
Sounds good
They want to get local open-weight models banned so people continue buying their subscriptions.
“Anthropic is scared that local MoE models are going destroy their subscription-based LLM business model so they try to scare governments into regulating local inference”
You’re out-of-date, I fear.