AdieuToLogic

> The only nondeterminism in LLMs is parallelism-dependency in the way floating point operations are carried out and in the decoding strategy

This assertion is "oddly" similar to the GPT answer "neural network inference determinism" produced:

  Neural network inference is often non-deterministic due to 
  factors like floating-point arithmetic and concurrent 
  execution, which can lead to variations in output even with 
  the same input.

Surely this is but a coincidence.

Regarding your previous statement:

> However, on a technical level, neural network inference truly is inherently deterministic.

This holds for a vanishingly small set of conditions, none of which include randomness, nor when context and transformers are involved, let alone underlying model evolution (thus making model use over time non-deterministic).

See also:

  Foundations of Large Language Models[0]

0 - https://arxiv.org/pdf/2501.09223v2

> Intelligence can result from learned behavior, but not from explicit programming of rules by humans.

This is incorrect.

Simulated intelligence can and has been encoded explicitly by humans defining rules programmatically in the form of expert systems[0].

0 - https://en.wikipedia.org/wiki/Expert_system

> I feel like I'm fighting a losing battle but I don't see why so many people consider LLMs innately non-deterministic ...

Because LLMs are artificial neural networks[0] (ANN), which are statistical in nature, and thus intrinsically non-deterministic. Pretty much every AI algorithm has randomness involved in its definition and many (most?) incorporate probabilities.

0 - https://en.wikipedia.org/wiki/Neural_network_(machine_learni...

The more semantically rich a programming language, the more it demands mathematical veracity.

The more a language is mathematically sound, the more its linguistic constructs converge to enforce same.

Programming languages which follow this path ultimately support similar capabilities; Applicatives, Functors, Monads, Monoids, and often meta-programming via higher-kinded types and/or intrinsic AST code generation.

Software abstractions are a reification of problem domain understanding. They embody a solution within the mathematical constraints of a programming language such that the collaborations, when done properly, appropriately reflect said understanding.

Discarding abstractions is nothing more than rediscovery of a Big Ball of Mud[0].

0 - https://www.laputan.org/mud/mud.html

> So you do or do not believe video evidence of a vehicle leaving a crime scene should be investigated?

I have expressed what I believe. The question you now pose is a strawman argument[0] unrelated to the subject of stolen license plates and how the act thereof implicates innocent people.

0 - https://en.wikipedia.org/wiki/Straw_man

Very cool poem!

Here is one I wrote a while back, unrelated to LLMs, yet a poem none the less.

  The Rhythm of Time


  The Sun rises,
  The Sun sets.

  Have I checked the mail?
  No, not just yet.

  The Sun rises,
  The Sun sets.

  Have I caught up with neighbors?
  No, not just yet.

  The Sun rises,
  The Sun sets.

  Have I spent time with friends?
  No, not just yet.

  The Sun rises,
  The Sun sets.

  Have I visited with family?
  No, not just yet.

  The Sun rises,
  The Sun sets.

  Have I told those I love I do?
  No, not just yet.

  The Sun rises,
  The Sun sets.

  Have I lost who I am?
  No, not just yet.

  For all I must do,
  Any money I will bet.

  Is to turn away from;
  No, not just yet.

I probably can't. I concede, you win, and please do not feel obliged to respond.

I wish you well in your path in life.

> Must be a misunderstanding here.

No worries, textual communications do not convey the nuances verbal conversation can provide, so misunderstandings are inevitable.

> It might be easier if you just stated your actual view instead of playing twenty questions ...

What I am saying is that criminals do their best to "throw off the scent" of who they are. The purpose of their stealing license plates is to infuse confusion and/or delay in discovering their identity.

> Just to be clear: I would be absolutely fine with law enforcement investigating the appearance of my stolen plates at a crime scene ...

Perhaps you would, depending on the circumstance and timing of said investigation. And perhaps others may not be as confident, including your future self.

Either way, a person otherwise completely uninvolved would now be a subject in a criminal investigation. The second-order effects of this only exist due to intentional identity misdirection, which is, of course, the whole purpose of license plates being stolen in the first place.

>> What I described is a realistic activity criminals do in order to disguise their identity.

> I've never read about this happening.

A simple web search[0] can elucidate the matter.

0 - https://duckduckgo.com/?q=what+is+the+most+common+reason+car...

>> First, I never implied disallowing analysis at a crime scene

> You weren't implying that video evidence of a car should be off-limits to the police?

No, I did not.

A car is not a crime scene. Neither is "video evidence of a car" a crime scene.

> If stealing hair is too hardcore for a criminal ...

What I wrote was:

  Second, "stealing your hair" is a very different thing than 
  stealing license plates off of your vehicle. This should be 
  obvious.

If you cannot see the difference between someone taking an attachment from a vehicle unbeknownst to you and being in your presence to take hair off of your head, then there is no reality which we share.

> ... just use your example and imagine that, instead of having the plate captured on camera, they find it at the scene of a vehicular homicide.

Then that would be entirely different than your original assertion, which was:

  The misuse of license plate data wouldn't make my top ten 
  list of potential powers that law enforcement might abuse.

Wouldn't it?

> Should that license plate be investigated, or should it be off-limits?

What are you even talking about? You said license plate data "wouldn't make my top ten".

When did I raise the question about license plates being investigated or off-limits?

> Almost no day-to-day policing is done with warrants!

Just because I feel bad for taking advantage of the above statement, here is a snippet from "Warrants in U.S. Law: The Ultimate Guide":

  Imagine your home is your castle, and the U.S. Constitution 
  has built a protective wall around it called the Fourth 
  Amendment. This wall protects your right to privacy and 
  security from government intrusion. Now, imagine a police 
  officer wants to enter your castle. They can't just knock 
  down the gate; they need a special key. A warrant is that 
  key. It's not just any key, though. It's a special 
  permission slip, signed by a neutral judge (a magistrate), 
  that says the government has a very good, specific 
  reason—called Probable Cause—to believe that evidence of a 
  crime is inside your home, or that you have committed a 
  crime. This document acts as a crucial check on government 
  power, ensuring that law enforcement can't search your 
  property or arrest you on a whim. It's a fundamental 
  safeguard that forces the government to justify its actions 
  before it invades your privacy, protecting every citizen 
  from unreasonable searches and seizures.[0]

Judicial warrants are the cornerstone of U.S. criminal law.

Policing enforces U.S. criminal law.

0 - https://uslawexplained.com/warrant_law

> Almost no day-to-day policing is done with warrants!

By this logic, there is no need for "day-to-day policing" to have warrants for:

  - phone taps
  - SMS messages
  - GPS history
  - email
  - banking information
  - justified arrests
  - court approved searches

The result of the position you advocate is known as a Police State[0].

0 - https://en.wikipedia.org/wiki/Police_state

>>>> Abuse by police officers and other individuals with access to these systems is the answer to the stupid but common question "why should I worry about surveillance if I have nothing to hide".

>>> I never hear that question from Flock supporters. The question I hear is "why shouldn't we be pulling over stolen cars?"

>> These are the same Constitutional question.

> They're not remotely the same Constitutional question.

They absolutely are the same Constitutional question; surveillance by the State without a warrant supported by probable cause, verified by the judiciary and issued thusly, does not have "an upside" of coincidentally identifying criminality. Remember this portion of the 4th amendment:

  ... and particularly describing the place to be searched,
  and the persons or things to be seized.

How does ubiquitous surveillance used as evidence whenever possible not violate this section of the Constitution, if not others?

And how could the question "why shouldn't we be pulling over stolen cars?" be answered without "why should I worry about surveillance if I have nothing to hide" being aligned?

They are the same Constitutional question - to accept one is to accept the other.

> Should we not allow the analysis of hair found at crime scenes because someone might steal some of my hair and place it there?

First, I never implied disallowing analysis at a crime scene. What I described is a realistic activity criminals do in order to disguise their identity.

Second, "stealing your hair" is a very different thing than stealing license plates off of your vehicle. This should be obvious.

> Someone framing me in this manner is incredibly unlikely, would be disproven by the other broad investigative powers afforded to law enforcement, and, at least in my opinion, is ridiculously less likely than the cameras photographing real criminals.

The point is that it would be you having to defend yourself for something you did not do. Could you successfully do so? Maybe. It depends on what the crime was and if you could afford a skilled enough lawyer.

And therein lies the rub.

You would have to drop tens-of-thousands of dollars in defense of your not being one of the "real criminals." Because the police are not going to sit there and say, "gee wiz, veec_cas_tant could not possibly be a 'real criminal' because they said they are being framed!"

They are going to say, "we can place you at the scene of the crime."

> The misuse of license plate data wouldn't make my top ten list of potential powers that law enforcement might abuse.

Are you sure about that?

How about if your license plate were stolen without your knowledge, placed on another vehicle of the same make and model, then photographed leaving the scene of a crime?

>> Abuse by police officers and other individuals with access to these systems is the answer to the stupid but common question "why should I worry about surveillance if I have nothing to hide".

> I never hear that question from Flock supporters. The question I hear is "why shouldn't we be pulling over stolen cars?"

These are the same Constitutional question.

In order to determine what cars are stolen, all cars must be inspected. If all cars are inspected, there is a reasonable case that the 4th amendment is being violated:

  The right of the people to be secure in their persons, 
  houses, papers, and effects, against unreasonable searches 
  and seizures, shall not be violated, and no Warrants shall 
  issue, but upon probable cause, supported by Oath or 
  affirmation, and particularly describing the place to be 
  searched, and the persons or things to be seized.[0]

> But I also think people on HN have a real problem in convincing themselves about the obvious righteousness of their anti-surveillance arguments.

See above.

0 - https://constitution.congress.gov/constitution/amendment-4/

> Please have everyone cite their sources and I will cite mine.

I did not ask "everyone." I asked you to do so.

> That Google and Meta employees are not law enforcement completely misses the point I'm making but I never expect people on liberal social media sites to see that

You wrote, and I quote:

  If you think some Google employees doesn't have MORE access 
  than a cop, you are WILDLY mistaken.

That you chose to invoke the trope "people on liberal social media sites" speaks not to a sound position so much as an intrinsic bias evident to all.

I wish you well in your path in life.

> This isn't an answer to the question. Why do you (apparently) refuse to answer the question?

Assuming the question you are referencing is:

  Please compare the number of officers doing such a thing to 
  the number who did not.

I addressed same as being irrelevant as well as a strawman[0] argument.

0 - https://en.wikipedia.org/wiki/Straw_man

> If you think some Google employees doesn't have MORE access than a cop, you are WILDLY mistaken.

This is a conspiratorial position at best. If you can cite from a reputable source similar abuse attributed to those organizations, then do so. Otherwise, the position you are championing lacks both fact and reason.

Additionally, as others in this thread have made clear, both Google and Meta are not police departments authorized to inflict lethal force.

So there's that.

> Please compare the number of officers doing such a thing to the number who did not.

Please compare the number of people who committed burglary at a location to the number who did not.

The act of violating laws has nothing to do with the number of those whom do not. Furthermore:

  SPD’s administrative investigation determined the employees 
  violated department policies governing the authorized use 
  of law enforcement technology and departmental resources. 
  Investigators found that employees were running searches on 
  personal acquaintances and family members. One officer was 
  found to have given someone from an outside law enforcement 
  agency access to SPD’s Flock Safety system, which they 
  weren’t authorized to do.[0]

> Also compare it to the number of incidents where the cameras helped capture criminals and pursue other illegal activities.

Non sequitur and irrelevant to the abuse documented.

> I'm betting that the number of officers who used it this way compared to the value it provides is minute to almost nil.

This is a variant of the "if you have nothing to hide, you have nothing to fear" fallacy. Or a nihilistic perspective of "the ends justify the means."

> It would be interesting to find the number of Meta employees or Google employees who do similar activities.

Again, non sequitur, irrelevant, and an assertion which can only be interpreted as a quintessential strawman[1].

0 - https://www.wtoc.com/2026/08/07/savannah-police-department-f...

1 - https://en.wikipedia.org/wiki/Straw_man

> The biggest thing I've learned from doing stuff like this is that there are no shortcuts. At some point or another, to truly learn something deeply, you've got to dig in to the boring details and do things the hard way.

This is the only path to mastery, or understanding if one prefers. There are no shortcuts to a person achieving deep understanding (a.k.a. "Aha!" moments).

Can a tool such as GenAI be beneficial to someone who already has done the work to understand? Absolutely. But it cannot infuse mastery into a person simply by its use.

Only the time and effort a person devotes can do that.

> There's a quote I forgot that says: "There is a Quality to Quantity".

> I agree with that, it may not be reliable, full featured, well designed, ergonomic, etc. But quantity enables things and thus is a quality in itself. It allows use-cases that benefit from accessible cheap software.

The underlying assumption here is that the quantity delivered is acceptable and does not worsen the customer experience. To continue with the manufacturing domain example; if a company produced widgets quickest and for minimal cost, yet had more lawsuits stemming from their use, is that the kind of enabling an organization should pursue?

> Those use-cases were not viable prior, because cost/time was prohibitive.

Sometimes a cigar is a cigar and sometimes what is thought to be a desirable use-case is prohibitive because, once sufficiently vetted, it is neither cost nor time which disqualifies it.

> I can guarantee that you cannot produce such PRs in the same quantity/quality than a modern LLM.

Who said I would not use a combination of LLMs, scripts, and any other automation technique available?

> They're here, because even "bad" code has uses. We can have quantity over quality. We can generate throw-away programs (where the lack of maintainability doesn't matter). We can solve some problems by brute force (keep trying until tests pass). For people who can't program, even bad code can do more than no code.

If this is an acceptable definition of software engineering productivity, where "quantity over quality" is prized, then hire me.

Because I guarantee I can produce dozens of PRs daily, each having tens of thousands of LoC deltas (pick any language you desire), none of which having research and/or understanding underpinning them, and all easily quantified as being "brute force."

> Yes on a very personal level you look at LLM-generated code and think to yourself "wow this is garbage" but what about the people, as pointed out many times in this thread, that simply do not care? Does that not count as "good enough"?

Aside from the "amateur" qualifier, the Booch quote below provides insight to those questions:

  The amateur software engineer is always in search of magic, 
  some sensational method or tool whose application promises 
  to render software development trivial. It is the mark of 
  the professional software engineer to know that no such 
  panacea exist.[0]

"Code" is shorthand for "encode", which can be defined as "encoding a solution to a problem", which implies knowing what problem to solve, and thus ultimately implying understanding of same.

So "good enough" depends on the stakeholders' definition of that valuation.

IOW, if a person tasked with delivering a solution can instead deliver garbage and still get paid for it, then it is "good enough" for some definition of same. If the people approving a person's remuneration are not satisfied with this type of work product, then the answer as to whether it is "good enough" or not will be quickly answered.

0 - https://en.wikiquote.org/wiki/Grady_Booch#Quotes

> Coding was expensive in the sense that once you decided what to do, it took a few engineers months / years to do moderately complex projects. That's not true anymore.

This conflates two orthogonal concerns; understanding what needs to be done with encoding a solution for what needs to be done. I am reminded of an axiom I have long held:

  When making software, remember that it is a snapshot of 
  your understanding of the problem.  It states to all, 
  including your future-self, your approach, clarity, and 
  appropriateness of the solution for the problem at hand.  
  Choose your statements wisely.

> Therefore the risk of "coding the wrong thing" is less.

I disagree with the proposition that being able to generate "wrong code faster" has anything to do with understanding what ultimately needs to be done. In fact, exploratory efforts having the intent to refine problem domain understanding (a.k.a. "proof of concept" projects) is an established industry technique whose work product is best discarded.

> And finally: Why bother making new software at all when AI can just do the thing you wanted done, directly?

In many ways, the "save costs by LLM-ing everything" mindset is the same logical fallacy companies fell into in the 90's/00's with regard to outsourcing. The end-state was rationalizing outsourcing call centers, then production system support, then software development, then project management, then...

Once the outsourcing raptor came for the accountants (who wrote the cheques) and remaining management (who cashed quarterly bonus cheques), outsourcing became "less appealing."

Indeed they would, as the decision chain I posited is a form of "argumentum ad absurdum"[0].

0 - https://en.wikipedia.org/wiki/Reductio_ad_absurdum