svieira

> the quality

Measured by?

We also need universal good (otherwise eternal life is hell).

Summum bonum, for those who want more.

Yes, but what's interesting here is that everything didn't get better. The original "interesting" for transformers was that a translation AI turned out to also be good at generating novel prose. But if we now need to focus to train new capabilities into the models and those don't yield better behavior everywhere then we're no longer in the "free lunch" period of LLMs.

> As an imaginary example, I wouldn't be shocked if some company like Jane Street is secretly investing a billion dollars to create the highly specialized synthetic RLVR data needed to train an AI that gives them an average .08% edge on a certain kind of commodity, equity or asset trade.

The crazy thing is ... it's not imaginary, nor is secret, it's literally a video on YouTube from five months ago - https://www.youtube.com/watch?v=8J-GUnfSqeE (though that facility is only showing of ~$200 million worth of compute).

There are perhaps more dozens of us than we expect, eh?

:raises-hand: I still don't know this. But my definition of "coding" may be different from yours, might it not?

Now the real question is "is the revert rate the same because you still have the same process, or do you fix-forward more often now, so your revert-rate is because of higher-severity incidents?"

> a good choice

For who?

"SOCRATES: Yes, Callicles, if he have that defence, which as you have often acknowledged he should have—if he be his own defence, and have never said or done anything wrong, either in respect of gods or men; and this has been repeatedly acknowledged by us to be the best sort of defence. And if any one could convict me of inability to defend myself or others after this sort, I should blush for shame, whether I was convicted before many, or before a few, or by myself alone; and if I died from want of ability to do so, that would indeed grieve me. But if I died because I have no powers of flattery or rhetoric, I am very sure that you would not find me repining at death. For no man who is not an utter fool and coward is afraid of death itself, but he is afraid of doing wrong. For to go to the world below having one’s soul full of injustice is the last and worst of all evils. "

- Socrates (via Plato), Gorgias

https://www.gutenberg.org/files/1672/1672-h/1672-h.htm#:~:te...

> Along with desperately loving entertainment, they desperately want companionship.

And I am presuming that the companionship that people need is the companionship of other human beings. Otherwise, tamagotchi, fish, cats, dogs, parrots, and horses are already perfectly good companions. AI can help fill in for some of those other ones (tamagotchi especially), but it cannot replace human companionship.

> People will have more artificial friends than human friends within ~20-30 years.

That would be awful, presuming that the artificial friends are not people.

Separately, you might like Count to the Eschaton by John Wright with its ultimate project being "the sophistication of all matter".

I recently re-read Venus, Inc. and the number of times that the satire didn't hit because it read too close to our current reality was quite depressing.

> don't allow

Is this "allow" or "encourage"? If "allow", how do you do that deterministically or at least fail-closed?

In the vein of these 2 minutes of Guy Steele's talk on Fortress (How to think about parallel programming: Not! https://youtu.be/dPK6t7echuA?si=huyI3LR8LVaB6Yen&t=1589) I think that the real thing here is "do not hide what is important nor obscure that which ought to belong to the user". The question is (as always) "what is important"?

And I think that the "tool" vs. "automation" delineation is one of the clearest and most succinct takes on _how to talk about this problem_ that I have seen in the past few years.

Is a query planner an agent? (It might be, I'm not sure semantically the "agent" vs. "automation" distinction will aid as much in the discussion as "tool" vs. "automation", but I'm open to being completely off base here).

> Give me the quick automated decision every time

Don't we already have this experience and isn't it called "the Google support experience"? And I don't know about you, but I don't like it.

Via LinkedIn (ai-tool-rich-hickey-kjl8e) but I suspect that we're rate limiting slop from there. Mods, feel free to swap out the original link if I was just-wrong about that.

> These systems are inherently immoral (it’s just a lot of matrix math)

(nit) "immoral" is "anti-moral". I suspect you want "amoral", that is "without" or "separated from" morals.

But, if you wanted "immoral" ... what about matrix math derived from gradient descent is inherently "immoral" in your view?

> If you wanted to go absolutely wild, some hypervisors can checkpoint the memory of a running VM and revert back to a prior version memory and all. I can’t imagine a way to make money off that

You may want to check out what Antithesis is doing in this space - https://antithesis.com/ Specifically, https://antithesis.com/docs/resources/deterministic_simulati...

The Boys On The Boat was good, working through Magnifica Humanitas (https://www.vatican.va/content/leo-xiv/en/encyclicals/docume..., but in dead tree form) for my mind stretching.

Where did you jump to?

> I described the problem to the agent and it figured out WAY faster than I would have been able to that the customer's clock was set incorrectly.

:blinks: The architecture of your service relied on customer clocks being correct? That is, you built a distributed service without a clock synchronization primitive that relied on the clocks being sychronized?

"I prefer" was the bit that the parent was pointing to that wasn't true. There isn't a person-to-person connection being made here. Though there is the fascamile of one.

I wouldn't say it's all rubbish - it's very easy for someone new to theorem proving to presume the equivalent of "strong typing will eliminate the possibility of errors", but the more correct understanding is "strong typing will reduce the number of things you have to keep in your head at any one time thus reducing the possibility of errors".

Yes, it could, but then you would have re-invented React Native. But in WASM.

That one person is (according to the previous comment) using words to suggest "this is a thing that will happen" though they are planning to make it happen? And that those words are attributed to the God of Abraham, Isaac, and Jacob? And that the selfsame God also has the other words attributed to Him?

Or are you suggesting a particular ontological presumption about the declared speaker of those words, which would then reflect a different shade of meaning onto my declaration?

> A lover needs someone free to reject him

As G. K. Chesterton said,

> Marriage is a duel to the death, which no man of honour should decline.

https://www.gutenberg.org/cache/epub/1718/pg1718-images.html...

Actually, everything is infused with purpose (not agency, that's different). Including rain drops and hurricanes and black holes.

Then they are in for a very rude awakening, given that the same God says "You must not wrong the widow and the orphan; wronged, they will cry out to me for redress, and their cry will be heard. My anger will blaze out against you, and I will smite you with the sword, making widows of your own wives, orphans of your own children." (Exodus 22:22-24).

> And nature doesn't come with purposes, it's all something we impose on it.

That is the presumption others are calling you out for. Things very clearly act appear to act with purpose, they just clearly don't provide it for themselves. You are presuming one solution to the problem (that appearance does not hold in reality), but there is another one.