karpierz

We have entire fields that have been trying to solve the alignment problem for autonomous + intelligent agents for centuries: economic and political science.

The author is noting that in this particular case, Valve made a mistake in releasing the debug symbols for a particular build, as noted by them releasing them in a 2017 build, and then removing them in a subsequent build.

The author did not say that "shipping with debug symbols is always a mistake", they noted that it can happen unintentionally and provided evidence of that in this case.

I believe the parent poster's point is: LLMs are more effective when they use more memory, meaning the less they are forced to compress the training data, the better they perform.

Is the implication here that everyone who could be affected by a transaction should have to consent to it?

In which case, sure. Hence why the government regulates transactions.

> Most mathematicians believe that not all sequences can be produced algorithmically, so this leaves some sequence too complicated to have a finite Kolmogorov complexity.

Isn't this just true?

The set of algorithms is countable, and the set of sequences of real numbers is uncountable.

> I've seen plenty of programmers who are very educated in history, philosophy, economics etc.

I've seen plenty of programmers who think they're well educated in history, philosophy, economics, etc. But actually being educated is rare. Much easier to read stuff like "Guns, Germs, and Steel" uncritically and just quote it at people.

> They put him under house arrest for having an astronomy theory they disagreed with.

No, they put him under house arrest because he wrote a book about for the Pope is a moron for not agreeing with him, after the Pope explicitly asked him to write a factual account of the theories at the time.

Tyrannical? Sure, like any monarchy. But not anti-science.

What are some examples?

I'm not saying Elon succeeds by accident. I think he's a very smart man who despite that, manages to think he's smarter than he is.

I think that his unique talents are not in engineering but sales. He's great at selling his vision to some people who will then bleed for him. He's great at fundraising and convince people to fund his ventures.

I don't think he's an engineering prodigy like Carmack or Wozniak. I don't think he's a product genius like Jobs. I think he's a fantastic salesman who finds bets that will pay off if you can keep at them for long enough, and uses his fundraising talents to keep his companies afloat until they do pay off.

> either he's a really good engineer or he's really good at hiring them, which means he's a really good engineer.

Or someone else at SpaceX is good at hiring engineers? Is your conclusion that everyone who owns a successful engineering company is a really good engineer?

The difference is that normally, the CEO answers to the board every few months. And the board doesn't tell the CEO what to do; the point of a CEO is that they can be trusted to do the right move. If they can't, then they serve no purpose and will be fired.

In this structure, Elon is embedded in the decision making loop directly. The CEO normally is responsible for the company, but instead, Elon has decided that the CEO will deal with making money while he makes all of the decisions as to how the product works. Which essentially cuts the CEO of of any decision making aside from business partnerships, making her more of a VP of Sales than CEO.

He has the title 'Chief Engineer' at SpaceX, a company which he owns.

Turns out, when you own the company, you can make the company call you whatever will best serve your ego.

> I did see significant benefits, the ones you describe like a silent mind, attentiveness, patience, for a period of time. Then those things faded away, I stopped taking, and experienced side-effects as my body normalized to not taking.

This is exactly how someone without ADHD is affected by Adderall. The medication has two functions: increased alertness and focus, which you build a tolerance to, and improved executive control, which is only relevant for folks with ADHD, and your body generally doesn't build a tolerance to.

Sorry, they've crawled trillions of pages, and narrowed it down to an index of 100s of billions. Conveniently, the link answers your question of "can you have PB sized indices?" to which we can clearly say, yes.

What if some child wants to work in the coal mines? What if some worker wants to be paid in company scrip?

Unravelling worker rights is just a race to the bottom.

Google has definitely indexed over a trillion pages.

There are a lot of different ways to interpret the word intelligent, so let me rephrase:

When you say "intelligent", what do you mean exactly?

What might help is describing what specific interactions give you the impression that GPT-4 is intelligent?

Ah, if that's what you mean then there are plenty of intelligent systems out there.

I've used Google search for decades and it's been able to answer questions better than humans ever could. Same for Google Maps, though arguably they're the same system at this point. My calculator is far more intelligent than any human I've met, at least when it comes to adding large numbers. My compiler can detect even the slightest syntax error with impeccable accuracy. Microsoft word has an incredible vocabulary. Wikipedia knows more historical events than any human dead or alive. And so on.

Shit, users thought Eliza was intelligent in the 60s.

If what you really mean is that LLMs are cool and useful, then sure. Just say that instead of couching it in some vague claim of intelligence.

Where did you get that the LLM was as _intelligent_ as a human?

All we've shown is that LLMs are as useful for parsing legal text as the average human. Which is to say, not.

A dog is also as useful for parsing legal texts as the average human. So is a rock.

Probably not.

I'm not sure that pointing out that LLMs are as useful for parsing legal code as the average human is something to brag about though.

How do you define numbers in a meaningful way without involving operators?

Without operators, all you have is a set of objects which you can't tell apart.

Assuming that all you want to do is assign, then sure. If you want to log the change or do any additional conditional logic, then that wouldn't work.

> you must suffer and live ascetically

What do you think veganism is?

If you take "any task" to mean literally any conceivable task, then sure.

If you take "any task" to mean "any practical task" or "any task a human would conceivably want to have done", then no free lunch doesn't apply.

> Maybe we should revisit those decisions. Even individuals revisit decisions, it is insane that we don't revisit certain decisions.

Maybe you could revisit those decisions by passing laws that directly change them, instead of threatening to keep the decision in place and just not allocate the funds you promised for it?

> That's not an actual algebraic definition in the same way you can express a peano successor function for say, all natural numbers.

I guess it's not clear to me how you define an "actual algebraic definition" of a function.

Can you provide a formal definition for this? Like some rules that I could use to determine whether a function is algebraically defined.

I think it'll clear up whatever misunderstanding I seem to have.

Here's a simple function from positive integers to primes.

Let f : Z+ -> Z+

Let f(0) = 2

Let f(n) = p, where p is not divisible by f(n') for all positive integers n' such that n' < n, and p <= p' for all positive integers p' such that p' is not divisible by f(n'') for all positive integers n'' such that n'' < n.

> (I believe) that's not true for the reals (someone correct me if I'm wrong about that).

It is also true for the reals. R^2 can be trivially mapped to R.

Here's one mapping from R to R^2:

Let b_{n} b_{n-1} .. b_0 . b_{-1} b_{-2} ...

Be the binary representation of the real 'x'. For example, if x = 2.5, then:

b_1 = 1

b_0 = 0

b_{-1} = 1

Have x map to the following value in R^2 (y, z) where

y = b_{n} b_{n-2} ... b_{0} . b_{-2} b_{-4} ..

z = b_{n-1} b_{n-3} .. b_{1} . b_{-1} b_{-3} ..

So 2.5 -> (0, 1.5)

1 -> (1, 0)

1.75 -> (1.5, 0.5)

etc.

Are you saying that with that prompt, an injection attack impossible, or that you haven't figured out how to get one to work?

You aren't required to make recordings. But if you made a recording, then you can't slate it for deletion. By disabling history, they're setting it up so records are deleted after 24 hours, instead of kept indefinitely.

In the court's eyes, it's analogous to writing notes to each other, only to burn them the next day.