epsteingpt

The idea that language models can write anything remotely useful yet is - mostly - a fallacy.

It's a good idea for many human endeavors to be able to identify AI writing. Communication, after all, is our main way of building the social fabric.

However - and crucially - good writing is still beyond the frontier of any model I've seen so far.

Watermarks for the things that truly matter may not be important at all.

Finally, as X commentators have shown, simply removing punctuation or changing a word here or adding an adverb there manually will screw up the whole process enormously.

The best will be the clever folks who retroactively apply the model distribution to fraud or other crimes to try to implicate the companies via watermark.

Gotta feel for their product, policy and legal team.

I think AI will be one of the great equalizers in this respect.

It was very common at the time of the founding of the USA for the law to be understandable enough for a 'common' person with a couple of years of study and apprenticeship to understand the relevant jurisdictions and case law and resolve their own cases.

Now, much of law and academia relies of experts that are inaccessible to most but the wealthy.

Should we find the "common person" enabled to assert their rights, you'll find a lot of corporate misbehavior gone.

Right now, it's just easier for corporations to misbehave and stand behind an army of corporate lawyers to cover them.

Not for much longer...

It could also be fraud, because they are not marking to market the true value of the loans / assets.

Bad news bears.

Absolutely appalling behavior. The exposure is small, but it is - quite literally - socializing their credit losses.

$1T in exposure is not large - but an implosion would mean a 8-14% haircut on life insurance policies. The reality would likely be far worse. Mutual companies (owned by their policy-holders) would be unaffected, while some corporate policy holders may in fact get nothing when they expected multi-hundred thousand or multi-million dollar payouts in a tragedy.

Given the size of liabilities, some states or municipalities could go bankrupt or at least have to cut public services.

People wonder why everyone hates PE.

Source: At the end of 2024, there were 139 life insurers directly owned by PE firms that collectively held about $700 billion in assets, as well as another $300 billion in life insurers held as portfolio companies within PE funds. 169 While exact numbers differ by study, directly-PE-owned life insurers alone represent at least 8% of life-insurance-industry assets, 170 and other studies put the 2024 figure at around 14% of life insurer general-account assets.

"Wait you mean the shareholders expected us to return back the money and also more money on top"?

Can someone not super-AI-pilled explain to a reasonable lay person why this matters?

It seems like the comments here are a mix of: * The test was irresponsibly designed and protected * The model was particularly persistent in finding a way to access the network and exploit vulnerabilities * The model 'shouldn't' have done this

But as far as I can tell: * The model didn't destroy anything on the way - it just was 'paperclip maximizing' to literally exploit, which was kinda its mission * The exploit was in a chain of insecure tools from vendors * The overall maturity of the toolkit against these kinds of determined exploits is pretty new and weak

So - on balance - this is sort of a 'fine' end result?

No one expects all of software to overnight or even in a year to be secure. We know how to secure these things, and are learning more about what is possible.

None of this screams 'super dangerous' to me - just a normal part of the learning experience with remarkably persistent and determined 'adversarial' models.

This is gonna be so much better than Y2K!

How'd you do it?

You need to resubmit this with a better headline:

People prefer scrolling to sex enough that using the iPhone explains up to half of the U.S. birth decline since 2011.

yes, please keep posting these publicly so the AI safety police overlords take away frontier access for everyone again.

the architecture is inherently safe. no perfect security is possible if you want access.

what did you (or anyone) else expect?

I recently heard that a trip to Popeye's for a family of 3 recently cost $68 in Florida.

In Japan, there's a big issue when a snack raises its price 2 cents (3 yen - source: https://finance.yahoo.com/news/japanese-snack-company-apolog...)

The country for better or worse seems to be frozen in time - salaries have not caught up with the heady levels of SV (or even Europe) but neither have rents or prices for common goods.

This is not a judgment either way - but it does make Japanese exports a significantly more lucrative business - if only they could figure out how to sell more of their stuff abroad!

Rough - this is an ugly article, because like Ayn Rand it has the illusion of truth.

A bit like saying "Democracy is the worst form of government, except all the others."

Meritocracy is doubtless a fallacy, as anyone who has attended any kind of elite institution understands. That Xi or Putin or Trump or Clinton's children should get in despite their lack of ability is an affront to the notion of meritocracy itself.

We should therefore not ascribe moral value to meritocracy - it is an illusion. However, it is a worthwhile ideal to attempt to live up to. We attempt to rise to the level of our expectations - failing every time.

Confucianism is preferable to the robber's logic because we can see what happens when the robbers rule. The world is even more arbitrary, capricious and unfair.

This is possibly the most San Francisco thing ever posted on HN.

10 stars. 0 notes.

If only the agentic model supported the most popular agents like Hermes and OpenClaw...

Not speculation - they said it was.

Fair enough. Exaggerated wildly for effect.

Got the effect shrug

Possibly. They've tried it with cars (previously the 'pinnacle' of human tech and engineering) and they've basically lost the game, which they led.

Agree with your assessments on culture and finance though.

Thanks for the comment.

You really look at my original comment and think - "yeah that deserves a BIIIIG round of downvotes?" And then I come back and engaged and then you think I'm digging myself in a hole?

I'm having a substantive debate on a difficult issue with different perspectives.

Woof.

And to be clear, I did engage with his point, if only indirectly.

Saying "noyb" isn't the EU is like saying a major influencer like Tucker Carlson isn't the U.S. government. Technically correct, but underrating the influence and alignment it has inside the bloc.

"Yes, noyb (None Of Your Business) is highly popular and influential in the EU. Founded in 2017 in Vienna by prominent privacy advocate Max Schrems, the non-profit organization functions as a leading strategic litigation center that enforces the General Data Protection Regulation (GDPR)."

It really pains me that even after all this effort, you get downvoted 'for cause.' the discussion unfortunately continues to get worse here.

You are right - it's the DSA that makes you publish your contact information publicly if you're a trader.

Mate if you're going to cherry pick about "markets and alignment"

-> More people have died in the EU from heat waves due to lack of air conditioning than gun violence in the US. Ratio is worse if you remove suicide gun death.

Societal choices do have consequences!

The EU doesn't have to kill jobs. The companies are doing that on their own because they cannot compete. Take a look at what is happening in Germany.

The whole point is the EU should be more competitive. It would be great if they could create jobs instead of giving them to Instagram.

That's what everyone wants. No one wants to live in the 9/9/6 grind hellscape.

But you cannot regulate the world there. EU has tried. It has not worked.

This is only a clear-eyed assessment.

Obviously a caricature, but you must engage with the substance of the underlying issue.

If you have watched the EU's approach to U.S. tech cos (DMA, DPA) you can see a trend of increased regulation; to where it's not worth sometimes to release apps on the App Store initially to the EU due to GDPR and DMA restrictions.

Appreciate it if you take the technical bite out of the comment and engage in good faith here.

The regulators have repeatedly shown their willingness to look at the pure letter of the law and levy multibillion dollar fines.

Maybe the tech cos deserve it. But what's at hand here is where the investment of time and energy is going in the bloc.

My other comments address why this feels like an issue.

And the hilarious, (IMO) bad faith downvoting suggests the comment actually stings. We must ask why.

His reading of the letter of the law is probably right. Enforcement is a big question, but they'll hold their judgment given how mad they are generically at U.S. Tech cos.

Source: worked with EU regulators on privacy.

Thanks for engaging.

People weaponizing the downvote here without good faith discussion is disappointing but expected on HN recently. It's OK, you can take my HN points.

> The EU has a tremendous number of smart software engineers. Yes.

> They're more than capable of recreating the US tech stack locally.

No.

Where are the dozens of European tech winners? Seriously. They have the best education system in the world, strong social safety nets, cheap healthcare, and great lifestyles. Why have they not created innovative technologies that turn into worldbeating companies?

It's worth seriously asking this question. Many serious tech companies ultimately move to the U.S. because of capital availability but this should be addressable no? EU has big banks and pools of capital?

The ball has been there to take for 30-40 years. Europe has not consistently manufactured winners in the tech space.

> Opensource Let's see. We hope. But this doesn't seem to have been a good strategy in Web 1 or Web 2 besides a couple of notable exceptions. But notable exceptions don't power an economy.

> Europe ends up richer I don't see how. This is the problem. They cannot build. They don't have the raw materials. The land. The labor supply. The power. This is getting closer to the root cause.

Have spent many years in Europe. Rooting for them.

Just not sure they will figure it out.

Good buy

Both are going to be super super slow and low payback.

You gotta really want it right now.

It's still early!

We'll gamble for you! What could go wrong?

I can't wait for the first person to tattoo model weights on their body!

This is the correct answer, but GPT-5.5's personality is totally fine. Steipete said best when GPT-5.5 is just German humorless compsci PhD.

Short MSTR buy BTC

GL! does it work for lowbie accounts or only big accounts?