The article frames a GPU cluster as if it was rocket science. If you swap a GPU cluster with, say, a power plant, the question doesn't change a bit. Also high complexity, multiple variables, you don't know the current output power, scheduled maintenance, skilled labor needed etc.
In the end, they will either figure it out, or contract a company that can figure it out. Truth is, there will be many companies wanting to do the rental of those clusters for AI...
This is what I hate about AI. Not that people use it, it's great to accelerate specific workflows, make less mistakes etc. It's just blindly trusting it and just saying "Make a post about a CV library release, make no mistakes" and calling it a day.
Where is the human creativity in writing release notes gone?
This is how I treated LLMs from the beginning, maybe because of my impostor syndrome of not knowing if my understanding of _anything_ is correct, and going down the rabbit hole of the concepts that are presented there...
Now the question to the round: in your opinion, are LLMs ok to learn in this way? At least on the theoretical side of things?
I think there is worlds between definitely defining what consciousness is, and what are some of the scenarios and conditions under which consciousness cannot ever happen.
And on top of that, they put a sedative, just in case.
> But to me the challenge is that it’s hard to imagine everything that needs to be preemptively captured and prohibited. I have to imagine this stuff for living, and I literally did not think anyone would just move a mouse pointer like this.
In requirements engineering, you do only codify things in the positive sense. The reason behind is that the system shall only do things that are described. If there are no things described, the system shall do nothing.
You can forbid a lot of things, but then you would have pages upon pages of requirements telling you things you may not even wanted implemented in the first place.
Honestly, that doesn't sound too bad. It does not say you can't use LLMs, it just doesn't let LLMs be the author of a commit. Meaning, if you as a developer make yourself responsible for what the LLM wrote, go ahead. But be ready to answer the technical questions, be ready to get grilled in the code review, and be called if you get a CVE on that part of the code...
Jokes on you I got a 20kWh backup UPS on the entire house.
Now on a serious note: internet can go down anywhere, even for them. There is nowhere stated that, if your old hardware misbehaves, you can use others, and there are privacy concerns with running your stuff on their software stack… So I don’t see a huge advantage honestly.
What's the difference between this and 1. setting my laptop at home and 2. connect it through Tailscale?
I lose ownership of my laptop, you install whatever software you want on it (with the security risks that it conveys) and in turn "you let me connect to my computer"?
What? They should freaking think of sanctions, not about "how easy is to lose Google account". Both Google and Apple are American companies. If someone lands on a sanctions list, they close your account without further notice [1].
Let me get this straight: you can be a defender of human rights, aligned with the country you live in, but if you fall in disgrace with the American government, _you can't even do transactions with your own country_.
So this is fundamentally flawed, and violates the fundamental rights of German citizens in Germany.
Knowing the German, how much of a fiasco will this be? Many Germans despise having to go online with specific services due to "Datenschutz". Now you are telling them that they need an external (American) service in order to use this?
What I don't understand is: ELSTER (taxes) already uses electronic signatures, don't these signature already fulfil the requirements of eIDAS? Why do we even need Google/Apple?
In some industries it’s critical. Think about aerospace where code is almost always homegrown or done by specialized company, and are specific implementations for specific needs. You don’t have that many COTS due to the criticality etc.
The question is: if we keep the same context and model, and the same LLM configuration (quantization etc.), does it provide the same output at same prompt?
If the answer is no, then we cannot be sure to use it as a high-level language. The whole purpose of a language is providing useful, concise constructs to avoid something not being specified (undefined behavior).
If we can't guarantee that the behavior of the language is going to be the same, it is no better than prompting someone some requirements and not checking what they are doing until the date of delivery.
This is something that could be distilled from some industries like aviation, where specification of software (requirements, architecture documents, etc.) is even more important that the software itself.
The problem is that natural language is in itself ambiguous, and people don't really grasp the importance of clear specification (how many times I have repeated to put units and tolerances to any limits they specify by requirements).
Another problem is: natural language doesn't have "defaults": if you don't specify something, is open to interpretation. And people _will_ interpret something instead of saying "yep I don't know this".
I see the fancy methodologies and processes as the way of streamlining what you have to do in order to "sit down to think about the software", particularly in teams of more than one developer.
Most of it happens, as always, at the interface. So these methodologies help you manage these interfaces between people, machine and product.
In the end, they will either figure it out, or contract a company that can figure it out. Truth is, there will be many companies wanting to do the rental of those clusters for AI...