Hacker News new | past | comments | ask | show | jobs | submit | MichaelZuo's comments login

How does that follow?

That would mean leaving some performance on the table the rest of the time.

It doesn't seem clear at all whether one outweighs the other.


Overwhelming an enemy involves getting inside their OODA loop. I can't see a real life-or-death scenario, outside of training, where you'd want your enemy to successfully get inside your OODA loop and disrupt your flow and rhythm, even for 0.1% of the time.

You of course don't want to become comfortable and complacent, risking losing focus, but there must be better ways of avoiding that other than being occasionally overwhelmed.


It doesn’t matter how many scenarios you enumerate, because the possibility space is infinite.

I don’t see how that could lead to a credible proof, even on the balance of probabilities.


You’re suggesting there are real deadly combat scenarios where it is beneficial to have your OODA loop compromised. Ok maybe you’re right, given the infinite possibilities.

But until you can present at least one such example scenario, no individual would be willing to take such a risk when their own life is at stake. Real combatants might value the motivating threat of being overwhelmed, but do not actually wish to be overwhelmed (i.e. have their OODA loop compromised).

In deadly combat, no one is looking to theorize. No one quibbles about their inability to prove the negative. They just want to live to see the next day.


And it's still missing certain things Cocoa had well before 2015...

That’s ok because it’s simple enough to escape SwiftUI. Use SwiftUI for the simple bits and UIKit for the harder parts. It’s not a nice solution, but it works

I consider use of UIViewRepresentable to be a kludge, and basically removing the advantages of SwiftUI.

I know that SwiftUI has some native wrappers, like maps, but the way that SwiftUI works, is so radically different from UIKit, that I think mixing them is problematic.


This doesn’t make sense, considering more than half of the population are shareholders (at least in the US).

By definition, if a large enough group wields the majority of political power then they will always have “primacy”.

It’s like complaining about rivers flowing downhill instead of uphill.


It’s a clear reference to shareholder primacy and it’s not wielded democratically nor is it something most even have a chance to participate in even though most have stock market exposure through retail buys, 401Ks, IRAs etc.

How does this affect the prior comment?

Half of the us population being shareholders doesn’t translate to a seat the table economically, not to mention is a red herring to the topic

The real issue is that there is legal doctrine that makes it hard for businesses to not be dominated by their largest investors / shareholders in such a way that extracting short term profits every quarter has taken precedence over building healthy sustainable businesses. Everyone is chasing the absolute most % of profit to the detriment to even the business


Who do you think supports and reinforces the “legal doctrine” on a daily basis?

Some mysterious beings outside of society?


Those percentages don’t seem right. I’ve never heard of anyone in the tech industry even lasting through 20 rounds of mutual defections with a peer opponent, in any real world conflict.

Let alone 100 rounds, or to continue as a live player.

Because in reality there’s almost always escalation to higher and higher stakes.


I'm not sure what you mean, those orgs last for decades and countries last for hundreds of years. But I'm sure you can tweak the percentages depending on the situation and your own judgement. It's a rule of thumb too and can be exploited if you're not careful, eg someone can save defections for when it really matters (getting a supreme court seat).

> Because in reality there’s almost always escalation to higher and higher stakes.

Well clearly no, because the world is not constantly at war (the highest escalation).


Can you provide an example of one “org” in the tech industry that has defected twenty times in a row against a peer?

> Well clearly no, because the world is not constantly at war (the highest escalation).

The world is not constantly at war because it’s almost always settled way before the count hits 20.


This seems like a tautology. By definition someone with superior political power can grind down someone else with less, if they were maniacal enough about it, or at least negate their efforts.

If they couldn’t, then they wouldn’t be considered to have superior political power.


It's more than that.

With enough power differential that won't even be a consideration.


Even perfect equals in an ideal spherical cow world can still engage in a 1 v 1 drag down fight.

Spherical beef is better than Wagyu!

If you wanted to be ironic via literally meaningless comments, then that is pretty clever.

That adds up to 110%…

He's a surgeon, not a mathematician.

Lol literal first four words of the comment said he's not a surgeon

He’s a brick layer, not a linguist.

Which must make it the most MD comment I've seen in a long time. It's hilarious!

That's just the normal level of performance people expect of their doctors, 24 hours a day, 8 days a week.

I really doubt LLM benchmarks are reflective of real world user experience ever since they claimed GPT-4o hallucinated less than the original GPT-4.

I don't have an accurate benchmark, but in my personal experience, gpt4o hallucinates substantially less than gpt4. We solved a ton of hallucination issues just by upgrading to it...

How much did you use the original GPT-4-0314?

(And even that was a downgrade compared to the more uncensored pre-release versions, which were comparable to GPT-4.5, at least judging by the unicorn test)


I don't recall the original version we used unfortunately :(

in our case, the bump was actually from gpt-4-vision to gpt-4o (the use case required image interpretation)

It got measurably better at both image cases and text-only cases


I begin to believe LLM benchmarks are like european car mileage specs. They say its 4 Liter / 100km but everyone knows it's at least 30% off (same with WLTP for EVs).

Those numbers are not off. They are tested on tracks.

You need to remove your shoe and drive with like two toes to get the speed just right, though.

Test drivers I have done this with takes off their shoes or use ballerina shoes.


Cruise control?

No you want to control the shape of the speed curve to not overshoot and not accelerate too much, when you follow the speed profile.

And keeping steady state speed is not that hard.


Hrm it is a bit funny that modern cars are drive-by-wire (at least for throttle) and yet they still require a skilled driver to follow a speed profile during testing, when theoretically the same thing could be done more precisely by a device plugged in through the OBD2 port.

Well in some ways it does, some insect species use various strategies to fool other insect species, such as a special type of caterpillar that does so against ant colonies, to live at their expense.

How does it compare with o1 and o3 preview?

o3 is okay for text checking but has issues following the prompt correctly, same as o1 and DeepSeek R1, I feel that I need to prompt smaller snippets with them.

Here is the o3 vs a new run of the same text in GPT 4.5

https://www.diffchecker.com/ZEUQ92u7/


Thanks, though it says o1 on the page, is that a typo?

It was not just totally private forums but many big forums had private sections and subforums where the real meaty discussions occurred.

Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: