Re: Haaaah-hahhhh-ha-ha-ha!!
Most surveys and studies are paid for on the sly by the AI vendors themselves; you can't trust the likes of Gartner & co. any more than the AI generated code they're trying to shill.
880 publicly visible posts • joined 14 Dec 2024
It's completely useless for remote access to systems, and to be honest, Vulkan drivers are the only thing aside from good old stable X11 and OpenGL that interests me, and that is only because of gaming.
I saw no difference in application availability under X11 and Wayland other than a lot of display issues with Wayland.
We never did need Wayland, so with this move, I can now add Gnome, Ubuntu, and Redhat to the list of things I neither need nor want.
Just what IS the Gnome team smoking anyhow?!?!?
Altman is the most egregious self-wanking specimen I've ever seen in existence. All he ever does is spew utter and complete nonsense about what LLMs are actually capable of.
Him and Musk's garbage about FSD both need to result in stock manipulation and fraud charges from the authorities against those individuals personally. Jail time is definitely in order for this magnitude of stock pumping.
I usually accuse someone who spouts the kind of nonsense that Altman and Musk spew of smoking CIA supplied crack, but Altman has taken the leap into "experimental" chemicals that mess up your brain far beyond what anything sourced from nature could ever be.
And I expect his models to be every bit as bad for society at large as those chemicals are for him.
This is a man who will screw over anyone, anywhere, anytime to pad his own wallet, same as any other hard drug dealers. He flat out doesn't care if you live or die once he has your money.
Really? You don't think the EU's internal market is comparable to the Chinese internal market that Baidu courted to build a business equal to Amazon? There is no reason the EU couldn't support an Amazon of their own if they chose to.
You buy the American bullshit because they advertise the shit out of it, promise you pie-in-the-sky-dreams before they tank your income stream and redirect it to theirs, and keep playing the fool for the tech bros and politician's bald-faced lies.
Another LLM-tard prophesizing the "benefits" of a dead-end futile hunt for artificial intelligence that instead delivers only artificially ignorant statistical summaries... WTF do you think the LLMs do except invoke system applications and utilities under the hood using the same syntax and shell commands as the command line?
It is no different than any other "new technology" that companies bought into because slick sales reps conned management into signing up with promises of "big savings."
I'm retired now but after almost 30 years in the trenches I've lived through at least three such "big things" that fizzled and disappeared over the decades.
The problem is there is no truly portable solution to the need for "the database" to handle the database storage of the metadata for DuckDB's proposal. I agree with the Apache committees - that is a non-starter. Metadata must be available in a text-only format that can be versioned along with the configuration files of the system so you have a proper recovery point. You can't easily do that with an active RDBMS. Nor do I know of any widely-accepted RDBMS that can be guaranteed to handle the sheer volume of requests that a large data lake system can be presumed to be dealing with.
And the closest most people get to mercury is either sucking on it or stuffing it up their posterior in the form of a thermometer. But the thermometer is only symbolic of what happens when a billionaire or purchased politician is around and you're not pulling in a 7-digit income...
The problem is I have an RTX 4070 Ti. It is far more capable than the so-called "AI PC" LLM capabilities are, at 12GB of memory and many more gigaflops (fractional teraflops?) of processing power as well. Yet still the only LLMs I can run are brain damaged 9GB models that are useless for anything except prototyping code because they hallucinate like psychopathic daydreamers with delusions of canned solutions for the "hard parts" the LLM can't "solve".
Just use this non-existent library/jar/maven package and all thine problems shalt go away... in LLM fantasy land.
Just a little out of touch with the realities of modern pricing for both consumers, hobbyists, and enterprise-level superscalars... nobody earns those kinds of margins except NVidia, and that is only because their "competition" hasn't proven themselves capable of going toe-to-toe with NVidia's hardware, especially at scale.
Intel has AMD as a competitor that has been eating Intel's lunch in the server space for some time now...
You want to get in a scrap with Windows 11 on install with the 24H2 builds? Just try to use the same email account for logging in and as an email provider for that non-Microsoft account...
I eventually was able to check my mail with their mail client, but it sure wasn't because their setup and configuration was smooth and easy to set up... So much for "user experience" nonsense.
NVidia has no choice but to deal with the Drumpf Regime, and it costs them far more pride and loss-of-face to do so than to deal with China! At least China honours their deals instead of tearing them up the second time they're elected like Drumpf did the free trade agreements with Canada and Mexico that HE SIGNED on his last go round at office! :o
I didn't say professional juniors were common. But like a trained, ideal junior that doesn't exist, I've found that the models I've used (primarily free Claude 3.5 with the VSCode plugin) do a good job of citing standards, software release changes, and pertinent articles on coding practices and styles. I found OpenAI models far less knowledgeable about coding practices.
YMMV
Realistically, the best way to tackle using an "AI" is the same as working with skilled junior programmers who've read a lot of books but have no practical experience using what they've read about.
Be clear with your requests. Be precise with your corrections. Argue in the philosophical sense if need be to "convince" the AI to do things your way when it comes up with a "bright idea" that doesn't work.
Skip the epithets, the cajoling, the whining, and for crying out loud: remember this is not an actual intelligence capable of learning unless allowed to treat scrapable web information as "fact," when we all know that 80% of what is out on the net is absolute crap, and that sites as focused as The Register are rare. Most either are flooded by people extolling the virtues of their (non-functional) solutions and dated approaches to coding. Very little good content of any kind, especially in the internal corporate software repositories, exists. Most is boilerplate copy-paste-modified from something vaguely related to the problem at hand.
The original mainframe concept of a "CopyBook" never went away; it just went online.
Gee, 20 years later they're catching up to Linux - both RPM and apt and a host of other package managers have been available for a long time and open to membership by simply submitting software for consideration in the standardized catalogues, or providing instructions and signing keys for adding your distribution servers to the client's repositories.
Betcha Microsoft has a fee associated with doing so, though!
It is definitely an idea worth pursuing, but I have a bone to pick with their weight-based comparisons. Weight has to include any and all retention, warm-up, cool-down, or other ancilliary hardware required for the fuel to be safely transported by the vehicle in question, whether it be Lithium, Sodium, Hydrogen, or Hydrocarbon based. Hydrogen isn't all that light when you also include it's cartridge and all the metals in those cartridges, nor the prewarming circuits required in cold weather climates. Sure Hydrogen itself is the lightest of the elements, but it's transport is weighty.
GitHub copilot with Claude 3.5 and the free service can't even make simple changes straight up without doing a build that generates errors it then tries to correct while virtually ignoring the original request.
That's not "intelligence." That's a damned bass-ackwards way of producing maximum screw ups and mistakes.
If Microsoft is relying on it that much, no wonder the busted "patches" and "fixes" they've been releasing have been so bad lately.
This is just the first of many bankruptcies that "Artificial Ignorance" is going to cause...
I've tried out a couple of the VSCode AI plugins, and even have a locally hosted brain damaged 9GB Ollama LLM running for cline on my box right now. That one is brain damaged because it's so small, but it will provide me with a testbed for trying to write some MCP services.
The Claude online/free signup LLM for cline is much more useful, but too expensive. Deepseek-R1 with Claude training enhancements is much cheaper and good enough, but still costs money I don't have.
Microsoft's free introductory service is comparable to Claude, but can't be used for the MCP services I want to play with.
But none of them can be trusted at all. They hallucinate APIs and packages that don't exist in Java 21 and it's family of frameworks and jars. For every moment of brilliance they show in using the latest code styles and letting you know which JDK added that feature, they drop the ball somewhere else in the code they produce.
Are the paid versions better? Somewhat more recent in their code styles, but still no more trustworthy from what I've read.
Companies need senior staff that know when the AI is screwing up. The productivity comes at a serious code quality cost, especially if people let the AI do the writing for them.
The old adage of "You can have it fast; you can have it cheap; you can have it good. Pick two." applies as much as ever.
It's high time the EU and other regions of the world did their level-headed best to eliminate the Americans from their markets. Far too much money gets fed to the American cesspool for comfort, and most of it is at all-time high ripoff prices.
Take your cue from some of the cities and nations in the EU: DUMP MICROSOFT ENTIRELY!