Around the time I graduated from the research institute in 2020, it seems my lab already had a similar infrastructure, just without LLMs and agents.
Back then, we had data repositories, databases, Jupyter Notebooks, Slurm batches, open computing platforms, and so on. It could do similar things ---- just by hand.
While adding an LLM agent can indeed drastically improve usability, it must be a massive headache for system administrators. It honestly sounds like introducing a huge, uncontrollable wildcard into the system.
I'm a Firefox user for about 20yrs (since Firefox 3);
but too often I have to use Chrome, as so many sites only work properly on it; Firefox is really buggy or laggy on those websites;
For a time, all those AI chat web pages were just very slow on Firefox even with very little context, whereas Chrome only gets laggy when there is a lot of context.
Are you really sure it’s not because of an add-on?
If I remember correctly, Mozilla has said that about 95% of all pages that don’t work aren’t due to Firefox, but to an add-on.
I use Firefox exclusively and don’t usually notice that pages don’t work. When that happens, as I said, it’s almost always an add-on that’s to blame. And I dont notice its buggy or laggy. So could be good check your addons next time.
Here are some cases where Firefox really sucks: some of them are specific CSS styles, some are downgraded features, and some of them I just don't know why. As I mentioned here, the ChatGPT web and Gemini web used to be very laggy for no reason—or maybe it was just a bug for me?
I don't think any of this is caused by add-ons, though.
But it's getting better, and most of those problems are just gone;
Same here, but when a site completely fails in Firefox I either A) use my phone because mobile Firefox occasionally works or B) use Ungoogled Chromium.
You can report websites that don't work, or block Firefox, to Mozilla at https://webcompat.com/. Mozilla engineers try to reach out to the website developers or ship site-specific workarounds in Firefox.
Not using many extensions on my case, but Google meet remains unusable for a long time, sound is horrible during meetings. Chrome on the other hand works fine
So, have you ever been to China and could hadely found anything familay?
- Oh, they must have been blocked from entering the Chinese market!
But none of that is true. You could see global brands everywhere here — Tesla, Unilever, KFC, Apple, and so on.
---
Or have you ever actually done cross-border trade? Or any international business collaboration? If you had, you’d definitely realize that what’s really stopping you is U.S. legislation.
At least, that was the case with our former U.S. partner
Basic science is often cheap, I don't know where you're getting it's generally expensive. I've yet to meet someone whose equipment costs as much as any of the stuff my friends design for defense contractors. Even the head of the lab I'm in is making less than my friends are making as engineers and the lab equipment is pretty cheap compared to the stuff my friends are designing (we have a radar that cost maybe a couple hundred thousand but that's the majority of the equipment cost for the past decade).
Idk what your idea of budgets are for these sorts of labs but I think most engineers would be shocked at the shoestring budgets they run on (at least the ones I know are a fraction of the cost of a single engineering team).
The “whole thing” is never cheap. Running something like a university — providing the environment, infrastructure, administration, facilities, compliance systems, equipment, libraries, grants offices, laboratories, and institutional platform that allow professors and graduate students to do research — is itself extremely expensive.
After all, if you look at the fiscal budgets of major economies, public spending on education and research is often much larger than military spending.
so called AI Art is OK for those pic themself.
but it's far too often abused.
All you get are these pieces of glossy junk, yet they expect you to believe it’s some form of creative work.
"People with minor cases of major brain damage", indeed
Several years ago, one of my Gmail accounts (mainly used for non-serious purposes, such as registering on gaming forums) was stolen due to a password leak. I received an login alert via a forwarded email, but since I hadn’t set up a recovery email address, I lost control of the account. I couldn’t even find any way to reach out to someone to take action and recover my email account.
All you can do it post a thread on the support forums, and nothing happen anymore;
I think for ordinary users (rather than developers or merchants), this is even worse.
Over a decade ago, my father would fix washing machine controllers, replacing mechanical timers, buttons, panels, or other parts;
Now, for the same problem, we just need to replace a control circuit board; the circuit board itself is sealed with adhesive for waterproofing, which also means the circuit board is not repairable.
Maintainability is actually not a mandatory standard, but a design trade-off; the biggest problem with the MacBook is not this, but rather that Apple does not allow other means of repairing the MacBook, such as various certification chips, etc.;
I think this topic is not about safety, but about profit and responsibilities.
The reality is that users should take responsibility but are not allowed to, so Google takes over and makes a profit.
You don't need a CS degree to use a phone, but you can be a power user by time....but not anymore, the company needs you to stay fool and pay for "help" (not directly sometime).
This is a marketing tactic, similar to a side-load.
The profit marginsof these industries are ridiculously high, to the point that if you’re willing, you can manufacture many useful, high‑quality products.
only when China could build them, there are real "free" market
I think wayland is OK as a user. But Wayland is just not really that UNIX.
As ordinary user, I actually don't care about any of this. However, from another perspective, I think this is a bad thing—open source projects have become product-centered, defaulting to the assumption that users are ignorant fools. This isn't how community projects should behave, but those projects is not that community-driven anyway.
After all, for a long time, so-called security has only been a misused justification—never letting users make mistakes is just a pretty excuse, meant to keep users from being able to easily access something, and eventually from ever accessing it at all.
Back then, we had data repositories, databases, Jupyter Notebooks, Slurm batches, open computing platforms, and so on. It could do similar things ---- just by hand.
While adding an LLM agent can indeed drastically improve usability, it must be a massive headache for system administrators. It honestly sounds like introducing a huge, uncontrollable wildcard into the system.