LabMechanic

The non-popular or non-widespread aspect of a "prompt" (e.g., "derive the geometric product given axioms/definitions ...") should be understood as a data point that it might lack reasoning skills. As it seemingly parrots what most of its dataset "says". For more details, check my other past comments. Calling such LLMs "AI" seems to be sensationalism at work.

A popular user here on HN also had a similar revelation or sentiment. We will see...

Seems like you are moving the goalpost (i.e., don't expect it to derive a theorem given a set of definitions and axioms from a non-widespread field). It should, in fact, derive any formula given a set of axioms and definitions, but we can't be sure whether randomness fooled us (again). If an LLM can indeed derive any theorem given axioms/definitions, even if we make up a field, then at this point it might not matter:

Ignorance is bliss.

You might see them both "being very, very similar", but that does not mean that it is.

I mentioned it countless times already, but it gets tiresome to repeat myself. Just ask your "AI", capable of "thought", a question (e.g., derive something given a set of axioms and definitions) from a non-widespread field (e.g., perspective geometry, Clifford algebra).

Then you might also arrive at the conclusion that it is a hyped up "Google search results mixer and matcher".

Who knows, but until it can fool me by randomness (like Nassim Taleb suggests), i.e., deriving something given a set of axioms and definitions (from a non-widespread domain), then I might join your "ignorance is bliss" argument.

Also, hype is good for the stock market, so keep hyping it up? At the end, Elon Musk success seems to stem from hype and sensationalism. CNN (i.e., "breaking news" everywhere) has financial success due to sensationalism and hype, so hype is good, I guess?

I am doomed to repeat myself, but I mentioned it countless times in the past and I also mentioned it here: Try asking it to derive something given a set of definitions and axioms from a non-widespread domain (e.g., Clifford algebra, perspective geometry). Then you might see that it parrots what the majority of its dataset says. It is essentially similar to Google, but you don't mix and match the search results, the LLM or the "AI" does.

Yep, or somehow they make it apparent that they indeed "parrot" stuff they had in their large dataset. See my past comments for details. Fooled by randomness? (See Nassim Nicholas Taleb)

Okay, I mentioned this in some of my past comments, but try this:

Ask your "AI" something about a non-popular domain (e.g., Clifford algebra) or even ask it to derive something (e.g., the geometric product) given some axioms and definitions.

Mathematics is precise as it gets, yet it failed to derive the geometric product given a set of axioms (e.g., distributive property) and definitions. Ask it to derive a theorem from perspective geometry.

Or, try something else that is not so popular or widespread (something outside of math), I suspect, that it also does a poor job deriving things from principles (axioms, definitions, theorems).

My 17-year-old "zoomer" sister is unsure about what a browser is. She believes you need a Google app to access the internet. My millennial sisters aren't very proficient with computers. They mostly use Office and a browser. Things like VPNs might confuse them. Are my sisters a rare exception? Who knows? I know other female and male millennials with non-software developer jobs who barely know how to use a computer.

Yep, I tried to persuade my "zoomer" sister to learn how to program, but she declined. :( She is simply not interested.

Yep, following rules like "never ever use goto" or "use design patterns and OOP everywhere" makes no sense, if (typical) software development itself is not exact. Some folks mix up software development with being super precise and mathematical, but it's not that exact. So do repeat yourself, if the alternative is not worthwhile. As a concrete example, I don't need to download an engine (such as Unreal/Unity) to make a retro-styled 3D game. All I need are software rendering algorithms, SDL2 and maybe Rust. That's it.

  There you go. Try it with 4. Still doesn't work? Let's revisit when 5 comes out. 

You appear to assume that current "AI" is able to "understand" and "think". What makes you so sure?

So, it "thinks" and "understands", it does not make (probabilistic) guesses?

Definition: guess

  to give an answer to a particular question when you do not have all the facts and so 
  cannot be certain if you are correct

https://dictionary.cambridge.org/dictionary/english/guess

To your other point:

  Also maybe try asking it something that ordinary humans can be expected to
  accomplish, rather than complaining that nobody has invented
  Ramanujan-as-a-Service yet.

Hmm... is it too much to ask such an "AI" for formal (i.e., well-defined) stuff? I would argue that mathematics or any formal language is perhaps more accurate than human language.

  You will lose the argument eventually, it's only a question of when.

Oh, I hope so, but I am skeptical, as I am still not convinced of an LLM being an "AI".

> Name one thing that HAL 9000 did that current LLMs can't do (for better or worse).

The ability to reason or understand seems to be not there in current LLMs.

See my other comment:

https://news.ycombinator.com/item?id=38819544

Simply put: "fake it 'till you make it" = probabilistic guesses given a large data set = seems to fool us people believing that an LLM "thinks" and "reasons".

  I think the gap is quickly narrowing on the need to fully comprehend most 
  computer programs for most use cases, ...

What makes you assume that current "AI" "comprehends"?

I wish it could, but it appears not to.

As an example, ask it to code up something from a "less popular domain" (e.g., Clifford algebra).

You might notice that it appears to not have enough data to make good guesses. Some "AI" hype folks appear to move the goalpost then.

However, maybe the stock market and some companies benefit from all this sensationalist stuff.

I wish I had a HAL 9000 beside me, but I am sorry Dave, I cannot let you do that.

There's a "contagious" sentiment going around: C++ is obsolete; use Rust.

Just search for HFT jobs or any software-based automotive positions, and you'll still find C++. I am confident that some experienced people don't care about the technology used; all that matters is domain knowledge and getting things done.

However, I'm concerned about "rookies" (those aspiring to become programmers); they might face decision paralysis because some high-level people share strong opinions, and they seem to follow them blindly.

I've sensed some common dogmatisms: "never use goto," "OOP is the way," "FP is the way," "use design patterns everywhere," "always avoid code duplication," ...

Are you telling me that software development (as typically done) is an exact science (like math)? Why are you striving to abide by those rules as mathematical theorems (things that require rigourous proofs), when those rules break or make less sense in specific cases?

I'm here to remind those folks to learn to solve problems in different domains. Language is a tool; use it as one.

Arguably, that is far more important than syntax and language details.

As Naval suggested, learn fundamentals (e.g., math, domain knowledge) and leverage them (e.g., through programming).

Here's another listing:

  Proficient in at least one of C/C++/Objective-C/Swift
  Experience with Python/Lua/Bash

https://jobs.apple.com/en-us/details/200525862/system-softwa...

Even if Rust gets more exposure in the industry, no big deal, once you know the hang of programming, you can just hop to another tool (i.e., programming language).

Indeed. We seem to assume semiconductors and computers are unlimited. It is just not practical to use your dog, a raven, or yourself as a Turing complete machine. You can't run Google on your Dog or yourself.

Hence, keeping "unused" electronics around might be wise, given the so called "unpredictable future" and supply chain interruptions.

Arguably, old and slow tech is still better than having nothing, as you do not have to do these boring computations by hand.

A layman question:

If Faraday cage holes are smaller than the gamma ray wavelength, does it protect against gamma rays? Or is a thick lead wall needed for M-Disc protection? Is it even possible to construct such small cage holes?

gamma rays = electromagnetic waves

Preserving an M-Disc against a nuclear blast seems futile, as the high temperatures can turn even stone into dust. Is there even a practical way to engineer survival in a nuclear blast? (Any physicists? Engineers?)

Here's a guide on making a Faraday cage, but can it also protect against gamma rays?

https://www.wikihow.com/Make-a-Faraday-Cage

I get your point, which boils down to "your money, your choice".

I am trying to bring across that having warranties for digital stuff makes sense, just like we have them for physical products. It's about fairness. Companies shouldn't delete things (especially paid ones) without a good reason; it should be reasonable, not just based on their whim.

The "sole discretion" thing is not fair. It's like saying companies can do whatever they want (e.g., polluting your neighborhood, risking your health). Imagine a world without any rules (like no government, no police). We'd probably go back to "smaller tribes" because surviving alone would be tough. A "big tribe" (i.e., a country) is okay, but a global one is a bit too much to handle. Even the US is divided up into smaller states.

Anyway, all I am arguing for is: there should still be some rules (i.e., warranties) to keep things fair.

C was my first language as a teenager. It is my native (formal) language, if you prefer.

Considering OP's question:

After C, I'd suggest looking into the Lisp family of languages like Scheme or Racket. I discovered them not long ago and wish I had found them sooner.

The reason is simplicity, ubiquity (Clojure), maturity (way older than C, see Lindy effect).

Then after Lisp, Python. Done.

There is nothing really to be learned thereafter. Programming languages come and go, but they have certain elements in common:

  - I/O
  - repetition (recursion)
  - conditional execution
  - arithmetic

Then there is stuff on top of that: OOP (which means data and code belong together. You can do that in C too:

https://www.state-machine.com/doc/AN_Simple_OOP_in_C.pdf

After C, Lisp (e.g., Racket) and Python, I would focus on common data structures and algorithm topics:

  Data Structures
  - Linked Lists
  - Trees, Tries & Graphs
  - Stacks & Queues
  - Heaps
  - Dynamic Arrays
  - Hash Tables
  Algorithms
  - Breadth-First Search
  - Depth-First Search
  - Binary Search
  - Merge Sort
  - Quick Sort

If you want to go to the extremes:

Lambda calculus and the formal language around Turing machines. Turing machines only need to read/write memory, move around memory, execute a finite program, have (near) unlimited memory.

Excel seems Turing complete, so maybe start programming with Excel. Don't!

Then there are what I call "industry languages", programming languages that you cannot keep in your head easily and probably need IDE assistance with:

C++, Java, Scala, Rust, etc., are production environment languages. Learn them later in life.

If you're okay with companies screwing you over, good for you!

I believe online services and digital goods should have regulations. When I pay for something, I want a guarantee. Cancelling my purchase anytime for any reason shouldn't be allowed.

Imagine it's 1998. You buy software on a disc (a game). Now, in today's digital world, it's like I'm taking back your disc without a refund. That's not right. Even if it is just a "license" you are buying, if you have something (proper) physical, you can still use your product.

I use KeePassXC to manage passwords. Save the encrypted file (Password.kdbx, 3 kB) wherever you want.

Extra points:

For emails, buy a domain like mydomain.com for about $15 a year. This lets you have a stable email address (e.g., [email protected]) no matter which email service you use.

Consider https://njal.la/ or something similar. However, note that:

  1337 has the sole right to at any time amend or supplement
  these Terms. Any such made changes to these Terms will be 
  immediately effective after being posted at the same 
  location as these Terms. 

https://edit.tosdr.org/points/24274

So maybe use another service.

I also use Portmaster when I am on Windows.

Forget defending against the government, they have state authority (violence, jail time), you don't.

See: https://riseup.net/

One-time pad (OTP) or "true randomness" (uniform probability distribution) is your only safe option (as long as they cannot time travel), but do it offline. Don't trust computers, see:

  Intel ME explained (intel management engine) from insider
  https://www.youtube.com/watch?v=1seNMSamtxM

Physics is your only safe haven. Seemingly, everything/everybody has to abide by the law of physics. Even your "almighty" state.

Suppose you've paid for content (apps, movies, music, games, or services). If your seller can terminate access at any time and for any reason, are you okay with that?

The 'don't like it, don't buy it' sentiment appears shortsighted, especially considering that both Android and iOS are operated by two major corporations.

Think of banking apps, for example. Banks usually do not favor a tampered OS.

Consider:

  False Sense of Security: A Study on the Effectivity of
  Jailbreak Detection in Banking Apps

https://ieeexplore.ieee.org/document/8806743

PS: Yes, you might use another OS, but how would you do your online banking with it? Explain.

Another point to your "don't like it, don't buy it"-argument:

Suppose you buy a device; further suppose the state is not involved (no warranty). Your device breaks down the moment you bought it, and there is no refund. How do you deal with that, especially if you are low income?

> "We reserve the right to serve refuse (i.e., garbage) to anyone"

Could not find it either, but clever indeed! :)

> You will now see that in your mind repeatedly!

X)

How do you become a "general problem solver"?

Is it sufficient to be skilled in one programming language, such as Python, and implementing a software renderer or ray tracer, your own neural network, your own HTTPS server, your own OS, your own memory allocator, your own interpreter, your own virtual machine...?

> The language, the stack, etc.. are just details. Given a problem - they can tackle it.

I agree, but are HR people agreeing on this?

For lithium-ion batteries:

  "... keep your battery life somewhere between 40 percent and 80 percent ..."
  
  "Research shows that extremes wear out the lithium-ion batteries ..."

  https://www.eeworldonline.com/why-you-should-stop-fully-charging-your-smartphone-now/

Sony's recommendations:

  "Charge your device just as much as needed. If you leave your device charging
   for a long time after reaching 100%, the battery might be damaged."

  "Avoid total discharges, as these can damage the battery and decrease the
   capacity."

  "Limit exposing your device to extremely hot (37°C or above/direct sunlight)
   or cold environments (temperatures below 0°C)."

  "If you plan to store your device for an extended period, charge it between 
   20 - 80% and then turn it off (Tip: this applies to any device with a
   rechargeable battery)."

   https://www.sony.co.uk/electronics/support/mobile-phones-tablets-mobile-phones/xperia-1-ii/articles/00232835

   https://www.sony.com/electronics/support/articles/00255462

It might not be "news," but God knows how many people are actually aware of terms containing phrases such as "for any reason at any time" and "without notice".

As I mentioned here, Dropbox's Terms of Service seem reasonable. They give you a heads-up. That's the minimum I expect.

I have issue with phrases like "for any reason at any time" and "without notice". If I lose my account, I want a clear reason, not just a general "you violated our Terms of Service" without any specifics.

If I am a terrorist, fine, do not notice me, but in all other cases, I expect you to notice me in advance.

For example, Dropbox seems reasonable:

  "If we do so (discontinuing services), we’ll give you reasonable prior notice
   so that you can export Your Stuff from our systems ..."

  "We reserve the right to suspend or terminate your access to the Services with
   notice to you if Dropbox reasonably believes: ..."

   https://www.dropbox.com/terms

I want to get a heads-up before you close my account. It's better to give some notice rather than just suddenly deleting it out of the blue.

> Should a service be forced to keep accounts running in perpetuity?

No, but maybe wait for 3–6 months for inactive accounts before closing them permanently.