Today RAM is cheap and fast enough to where some instruction compression is basically worthless.
Complete nonsense.
SRAM is not cheap at all, and DRAM is slow as fuck. That's why things like AMD's x3d processors kick ass.
Linus has said he doesn't know Verilog and knows little about FPGA and ASIC design.
Utterly irrelevant.
He's a computer scientist and knows ISAs and CPU architectures.
RISC-V is an upgraded MIPS, which was a pretty good architecture.
In your opinion.
In my opinion, anything with delay slots is stupid.
So Dollar Ton and gweihir are the same person are they? Ok.
Of course they're not. Nobody said they were.
For the sake of ironic humor, they may as well be for this conversation, however.
Both are fond of responding to everything they disagree by accusing it of being "fake news", essentially.
Tell me more about these things you think you're fit to evaluate.
I never told you anything about the things I felt I was fit to evaluate. You have trouble keeping your thoughts coherent, don't you?
You said:
You should explain to the GP how they're actually responding to the wrong person.
When of course I (the GP) did no such thing.
It occurs to me that you may have been being sarcastic, but the humorous part is, again, that it would have applied to either of them.
If a problem is unsolvable, declaring it so is the most efficient route.
Every LLM I tested declares that it's unsolvable in English.
Also, you really are just being stupid about this.
Imagine the LLM is a person.
You say, "give me 5 numbers that don't have 'e' in them".
Are you really about to argue that the percentage of people, recognizing this is impossible, and responding with another language or roman numberals is zero?
Are you surrounded by stupid people?
You just can't stop lying to yourself, can't you
Feel free to point out the lie.
Show your work.
You're a gaslighting shit-for-brains is what you are.
I guess you are, if you have to claim it in a Slashdot discussion
Have to?
This statement is too stupid to even point out the logical fallacies present in it.
When the fuck did they stop teaching basic debate logic in high school?
No. They can have results from somebody that had general intelligence in their training data for the specific question asked.
This is a pointless distinction.
Of course any intelligence they have comes from their training data. However, generalization is a quantified effect of pretraining.
If you show an LLM enough riddles, it will learn to generally solve that class of riddles.
Beyond that, if you show an LLM enough classes of riddles, its performance in unseen classes of riddles will improve.
This is just math. No magic required, though I assume you'll accuse me of magical thinking.
But you are just playing dumb again. Stop it.
I'm not playing anything.
I'm shining a light on your double standards.
Wow, what an innovative tactic, an implied strawman!
I don't think you know what a strawman is.
When was I "wrong" and about what? Usually people are specific, but you, you accuse me abstractly, which is a new low even for you.
Wrong about them (and every other entity that says anything positive about AI) lying... Christ, are you as stupid as you are crippled by your cognitive dissonance?
On the point, yeah, what a funny worldview, where instead of simply reading the disclaimer in the press release and admitting its truthiness isn't very truthy, you swallow everything that people tell you at face value.
I assume you're talking about the boilerplate in the earnings call (since there was no press release) that says, paraphrased, "these numbers are as truthful as our precisely stated methodology allowed them to be".
To misrepresent the numbers in an earnings call is a federal crime.
It's hard to imagine it's led you to much success.
I'm ridiculously successful. But I think you're probably stupid.
How many Bavarian Illuminati does it take to screw in a lightbulb? Three: one to screw it in, and one to confuse the issue.
