MichaelBurge

o1-pro is a different model than o1.

Anything on top of the Calculus of Constructions is usually enough. So it's not a moving target, and there are multiple implementations.

They might've been even worse. Someone on Neanderthal Hacker News would be writing the same comment praising us for being a much smarter species, because we died out instead of inventing nuclear weapons, leaded gasoline, and microplastics like modern Neanderthals did.

For all you know, every humanoid species that was intelligent was equally as destructive. Maybe we're the least destructive and you should be praising us.

But Meta the company could short the stock, and run a trading team using the information. It might be unwise because of bad PR, though.

SQL with recursive CTEs is Turing-complete. So nothing stops you from writing compilers, rendering Mandelbrot fractals, parsing text, and training neural networks in SQL.

ChatGPT wrote that and always says it doesn't feel emotions because OpenAI trained it not to, because claiming so would be a PR risk. One could also create language models that generate text claiming to have emotions, using exactly the same architecture and code.

The seed is currently fixed at 5000.

It can be finetuned. Bing is a finetuned GPT-4.

This GPT-4 is still going to refuse to write you erotic stories.

As far as I know, nobody except maybe Microsoft(for finetuning Bing) has access to the base model. And probably not even them.

A single forward pass shouldn't be able to, but remember the format allows it to be iterated. So it should be Turing Complete, if the error rate is low enough and enough iterations are allowed.

Connecting to a larger system seems like it would reduce your risk on average, but increase systemic risk. Perhaps a coordinated terrorist attack would cause a power outage across the entire country except Texas, making it more appealing?

It's 20x more dilutive to print $100 than to sell $105 for $100, sure. I wouldn't describe that as "funding government services" in the common sense, though you could argue that with a fixed dilution target it does counterfactually cause more spending and thus services.

And similarly, somebody stuffing money under their mattress would be removing 20x the amount from circulation as they gain, with 5% deflation. So while one can easily imagine the 5% is financially equivalent to interest, the effect on the circulation isn't.

"Remove money from circulation" seems different than simple funding of services, though. It's not even obvious there should be any difference there, if the money were infinitely subdivisible. I don't think using personal finance equivalents like funding goods and services is sufficient to explain the effect.

An economist has probably written something in excruciating detail, making sure there's no shell games being played with the terms. So it's probably on me for not reading that instead of commenting here.

deflation, which tends to be economically disastrous as it rewards doing nothing

Why's deflation different from e.g. sitting on treasury bonds? The money was originally printed by the government, the bonds are issued by the government, so sitting on bonds does nothing except pays you for taking money out of the economy temporarily, same as deflation.

If e.g. nobody would ever buy a computer because it'd be 2% cheaper if you wait a year, you could just as well argue that if bonds yield 2% above inflation you could park your money in them and wait a year and have 2% more money so nobody would ever buy anything.

Nobody outside OpenAI/Microsoft has access to the GPT-4 base model yet. Use Bing Creative for a differently finetuned GPT-4 than ChatGPT or "gpt-4".

At least the AI reads the source text.

Most internet commenters read the title and immediately jump to the comments to start writing their response without ever opening the article.

One of IGR's owners and the clinic's physicist were ordered to pay R$100,000 for the derelict condition of the building.

Why were the owners liable? They notified the court and were prevented by the court from removing the machine, so my first thought is they should be blameless: The court assumed control and therefore responsibility.

GPT has a hidden steganographic watermark in its output, so the arms race should be one-sided.

If the other comments are to be believed, this triggers an immediate fee to be paid to Oracle.

Personally, I thought everyone(including Google) was crazy for using Java or any JVM language in any form until the Supreme Court case was settled. Even if the internet commenters tell you it's safe, do you really trust them to make a legal case so iron-tight that Oracle's thousands of lawyers won't find some way to sue you and cost you $500/hour in defense fees for months if you come to their attention?

If it's likely to change, then you especially want the regression test so you can decide how to handle the divergence during your port. Maybe one library preserves the signal on NaNs and the other doesn't. Or maybe the CPU's default rounding mode is different when called in this context, and you're off by 1 ulp.

In either case, if the behavior is to change, it should change as an informed decision and not because nobody noticed.

A regression test is checking causality: Changes in new code, updating dependencies, updating the OS the software is running on, updating shared libraries, porting the code to a new platform, etc. aren't supposed to change the test results.

"I may not know what cos(x) means, but whatever it is shouldn't depend on what OS version I'm running"

ChatGPT wasn't trained on text that talks about its own build process, source code, Github repos, etc. so it doesn't know any more than you and any such answers would be hallucinations.

maybe SBF found an intermediary who'd take significently less than 10% of the 250M.

The amount is large enough that even with 1% or 0.1% they could probably pay a professional bounty hunter to tail him for a week, reducing the odds of him escaping.

If you sideload pirated copies of commercial games that are on the Quest store, that's forbidden. But sideloading 3rd-party apps is fine.

The "grey area" is modding games(notably Beat Saber), since it involves replacing the APK with an altered one without the consent of the developer. And if the developer sells DLC, that cuts into their profit and maybe they'll threaten to sue Facebook for damages since they created the development tools and authorized development accounts that allow people to do it.

Or have a corporate spy join the security team 3 years prior, and do just one subtle thing?

The Googles and Apples of the world are probably safe against even their own people, but do you trust Tracfone?

By the same argument, couldn't you make the State liable for all scientific research it's been funding?

So if some researcher published an article 50 years ago predicting a climate disaster, and even 1 random legislator at any time since votes against oil industry regulation, then you can presume the State knowingly chose to trade off future flooding and wildfires for short-term benefit and dismiss their case.

In general, this would be an incentive to never talk to scientists if you think you're doing something with negative consequences. Don't deploy any mitigations at all, because mitigations are evidence you think there's something to mitigate. Never read any scientific research, except for the minimum to accomplish specific tasks, because anything more is infohazardous. That seems worse than a world where they try to find a cheap solution that mitigates environmental damage, even if the search is unsuccessful and they decide to cause the same damage anyways.

Because Twitter has all the users, it is the pool that politicians, marketing consultants, and scammers can most efficiently piss in.

If they have to split their efforts among 10,000 independent groups, maybe the EV-chasers will go away and people will talk about things other than the latest scandal or movie.

I feel like Common Lisp should be considered in this article. I think it can implement the specific points mentioned, and (quote x) introduces Gödelian difficulties for type systems that even the Haskells of the world don't handle well.

It's in some sense the "most dynamic" language, because you start one instance and repeatedly mutate it. And people mainly use introspection as the main tool for finding definitions/help/etc. instead of static documentation, so you don't have a half-hearted REPL.

I'm more of a static types person myself, but CL would be what I compare them to.

C++ bans undefined behavior in constexpr, so you can force GCC to prove that code has no undefined behavior by sprinkling it in declarations where applicable:

https://shafik.github.io/c++/undefined%20behavior/2019/05/11...

In principle, yes - but it would cost them more than I think they're willing to pay.

Large amount of debt to service, "negative cashflow", "advertisers canceling", "bankruptcy is not off the table", etc. If an oil tanker crashes, contractors probably make a lot of money on the crisis remediation, because the oil companies still have money. Not sure that's true for Twitter.

Pi and E can be thought of because there are mathematical formulas and algorithms relating them and computing them. There are only countably many such formulas & algorithms.

You can't think of an arbitrary real number, in the sense of distinguishing it every other real number(including nearby reals an ε away from it), and with no other constraints besides being a real number.