They’re not, and will never be in their current form and architecture.
Compilers are mechanical and engineered to produce a correct output. A compiler emitting incorrect machine code is exceedingly rare, and considered a bug. They have heuristics and probabilities in them, but those are to pick between a set of known-good outputs.
An AI is a bag of weights outputting a probability of the most plausible token that follows [1]. It is inherently probabilistic in nature and its output is organic (by design, they’re designed to mimic human speech), as opposed to mechanical like a compiler.
A compiler follows hard rules. An AI does its best.
And to be fair, AIs are no better than human in this regard: humans are pretty bad at generating correct code without mechanical tools to keep them in line (compilers, linters, formatters). It’s not a wonder we use the same tools to keep LLM output in line as we do humans. (And, to be fair, LLMs are better than humans at oneshotting valid code).
[1]: to those that tell me this vision of an LLM is outdated: nope. The heavy lifting is done in the probability generation. Debates about understanding are not relevant here, and the net output of an LLM is a probability vector over raw tokens. This basic description can be contrasted to a compiler whose output is a glorified Jinja template.
Not OP, but it means nothing, because it's not "effectively" becoming a compiler.
Think about it from an information theory standpoint:
A compiler takes at least the exact amount of information it needs to produce a result, and produces exactly that result every time (unless it's bad at its job or has a bug).
An LLM always takes far less information than would actually be needed to fully describe the desired output, and extrapolates from that. It fetches contexts and such to give itself a glut of assumedly relevant information, but the prompt always contains less information than necessary to produce the code it generates. If it did fully contain enough information, then you've just written a far more verbose version of the program in human language.
Yes, I am saying there's no functional difference (for practical purposes) between a deterministic transformation like a compiler and a perfect probabilistic transformation like an LLM.
We do not have "perfect" probabilistic transformation, and we probably never will (in part because it's hard to know what exactly that even means), but the gap between the two is shrinking every day.
Ergo:
> they're becoming (effectively) more and more similar every day.
These are outlier trades by size alone placed minutes before news broke.
This isn't just people betting on TACO or NACHO or pathological lying and whatnot. These are ultra, ultra high-conviction bets literally large enough to move commodities markets placed minutes before news that, astoundingly, moved the market in exactly the right direction!
I think there actually were drones (even if most reports were aircraft), but that they weren't nefarious (my theory is a commercial LIDAR mapping project) based on the fact that the sitings dropped precipitously when flight restrictions were declared.
I was following it pretty closely for anthropological reasons. There wasn't a single clear photo of a drone, just planes and helicopters, over the most densely populated area in the country. It was the same as UFOs. Not one clear photo. The subreddit kinda went over the deep end and started talking about holograms, sentient orbs, spirituality, because what else can you talk about without any falsifiable evidence? Fascinating example of modern mass hysteria.
Uhh... "those in the know" are the actuaries and if you were to take away the subsidies provided for homeowners and developers to deny basic mathematical facts, the entire area would be totally unbuildable already.
i work in insure tech, in the E&S space, which is where all of the flood and wind polices gets placed. Actuaries have nothing to do with it --- the cost of hurricane insurance comes from Moody's RMS and Verisk AIR, the only two CAT models the carriers and re-insurance companies use. Actuaries price the non-cat risk.
This is mostly a pedantic point that it's not actuaries doing the pricing, but a different set of risk analysts using a different suite of tools, right?
[Edit] The following comment can be read very snarkily but that is not the intention. I'm legitimately interested + curious:
Very interesting! Does it affect the point that insurers (even if not actuaries) put a high price on this risk and that the price is subsequently suppressed by government insurance subsidies?
Local governments have obvious incentives to encourage building, but the state of Florida itself does subsidize flood and hurricane insurance.
If you own a house or building in Florida and have a mortgage, you're required to carry it. Here's how a policy gets priced:
You go to a retail broker with your info. They pass it to a wholesaler, who puts the submission out into the market for quotes. Any carrier or MGA that wants the business prices the CAT and AOP (non-CAT) portions separately. Actuaries build models for the AOP side, while Verisk and Moody's model the CAT portion. Those two numbers get added together, plus some fees — and that's your annual premium.
From there, the insurers buy reinsurance on their portfolios. The reinsurers run those same models, do their magic, and come up with their own price.
Just an example, because no major hurricanes have hit the south east in a while, premiums are down 30% right now. All of the insurance companies are getting squeezed.
Way too many Americans either don't know or disbelieve that a substantial chunk of the body politic, and now our elected and military leaders, actually literally believe this type of stuff.
IMO any eschatological beliefs whatsoever should be 100% universally disqualifying for any political or military position, no matter what book title or special ancient zombie character they're filed under.
“Leaders” who believe this kind of stuff don’t end up running developed states. It’s the leaders who know how to make use of morons who believe this kinda stuff who do.
Eh, no. Trump of course has zero actual ideology, but there's pretty solid reason to believe e.g. Hegseth and Mike Johnson actually believe this type of stuff.
We should have a one-time buyout for flood zones: pay someone enough to buy a median home somewhere similar and turn the land into a nature preserve (let mangroves return to protect Florida coast, etc.). Put a cap on it so we’re not buying new mansions for a few rich people with beach houses but otherwise keep it simple so people aren’t impoverished into becoming a drain on society.
I have no expectation that we’ll be willing to invest in our neighbors, though.
I thought the government should have done this for all the beach houses that were destroyed by hurricane Sandy. Buy people out and prevent a house from being built there ever again.
Agreed, building on a flood plain is incredibly stupid. The city of East Grand Forks demolished all of the buildings in the flood plain portion of town after the 1997 Red River floods and turned it into a park. https://en.wikipedia.org/wiki/Greater_Grand_Forks_Greenway
The way McDonald's does this that they essentially offer discounts that you can only get by using the mobile app, and the mobile app can dynamically price your cart however they wish. All they (or any other business) has to do is to increase their non-mobile prices to squeeze out non-identifiable shopping.
This is hilariously first-order-effects thinking...
If any store used dynamic pricing to expand their margins, the others would just do the same and compete away those margins once again, with the marginal gain being handed back to consumers.
Dynamic pricing on personal data is bad I think, but temporal dynamic pricing is actually very good for everyone and I hope it doesn't get thrown out by some reckless legislation-writing.
What happens in practice that there are only so many grocery stores where consumers can choose to shop thanks to corporate mergers and lax antitrust enforcement. So if all of them raise their prices at the same time then those consumers are out of luck.
Now technically it would be illegal for the grocery stores to collude in price fixing like that, but they'll hide behind the fact that all of them will buy their surveillance pricing data from Google [1].
Google will tell all of the competitors exactly how much they can charge you for your eggs, and you'll get the same price everywhere.
That's the ideal dream scenario, but in reality the market isn't that efficient. Lots of markets gouge their customers and due to power imbalances the customers can't really do anything about it. The free market solution to this is just generally to let people suffer.
The grocery market is. Margins sit at 1-2%, there is absolutely no reason to believe dynamic pricing would change that. Grocery stores are one place where free markets have created incredible consumer surplus because competition is high.
I don't see why consumers wouldn't just pick the stores that have better prices for them. Why would they know/care/need to know what the prices are doing for other people?
The issue is that consumers who can pick better stores will get better prices, while those that can't will get gouged. Imagine someone who works multiple jobs and only has time to shop at the grocery store closest to them.
The algorithmically-driven store will start by randomly showing them some random prices and seeing how they respond. If they are willing to accept high prices, the store will keep charging them more. If they leave the store and go somewhere else, the store will revert to lower prices. The store will discriminate against people who won't comparison shop for whatever reason (busy, limited access to transport, rich enough that they don't care, etc.)
However, the store's extra margins won't lead to lower prices for other consumers, even in a fully competitive market. Raising prices for consumers who won't comparison shop will do nothing to change the marginal cost of serving a consumer who does, so this won't change the competitive dynamics for those consumers and they won't see lower prices.
> In a competitive environment (which grocery stores are), this excess capital gets reinvested into beating the competition.
That's not true though. Excess capital does not necessarily get reinvested in an efficient market. If that were the case, companies in relatively efficient markets would spend a very small portion of their free cash flow on dividends and share buybacks, which is not the case.
Of $7.2B free cash flow in 2025 they spent $3.9B on capital expenditures and about $3.6B on dividends and buybacks (those numbers don't add up because of things like loans, sale of assets, and stock issuance).
Additionally, even if companies in competitive markets did reinvest all their excess profits from personalized pricing, the benefits would only accrue to consumers that the algorithm thinks are price-sensitive.
You can't see you mean? Consumers pick things that are reasonably convenient, achievable and known to them. These things are all subject to exploitative tactics by grocery stores.
There's also price-fixing, which famously occurred in Canada recently.
Not to mention cornering a market like Walmart would and removing consumer choice entirely.
You were not asking about dynamic pricing. You were asking why you weren't able to see the reason that consumers don't always find & choose the cheapest option. The person you were responding to was explaining the unethical realities of food retailers. There's no reason why these would disappear with dynamic pricing.
No, the entire conversation is about a change toward dynamic pricing. My comment about pricing is that dynamic pricing doesn't change the competitive dynamic between retailers. You are chiming in and saying "well there are other competitive/noncompetitive dynamics at play," which I agree with and never disputed and am not talking about.
> the others would just do the same and compete away those margins once again, with the marginal gain being handed back to consumers
This is an assumption that doesn't necessarily have to happen. Some markets remain uncompetitive, otherwise you would see every market collapse to 0 margins if this were always true.
The point of personalized pricing, which is what this bill addresses, is to identify consumers who won't shop around and give them higher prices.
You can't compete that away because the consumers being hurt aren't sensitive to competition (for example, they don't have a car and therefor can't get to a different grocery store). It's an inherently anti-competitive practice.
I agree that temporal dynamic pricing might be good in some circumstances, but this bill doesn't ban it.
Right, anti-competitive practices are where a business tries to limit competition. And dynamic pricing is anti-competitive because businesses are essentially algorithmically identifying customers whose business they don't need to compete for, and then charging them non-competitive prices.
> If any store used dynamic pricing to expand their margins, the others would just do the same and compete away those margins once again, with the marginal gain being handed back to consumers.
How would you do it if pricing is dynamic and changes every day?
By the time competitor finds out about the price, you might have already reduced it, making it look like theirs is more expensive even after they applied discount.
Dynamic pricing based on personal data is not even a market, let alone a perfectly competitive one. Temporal dynamic pricing can mean almost anything, so might be ok (early bird lunch deal) or pure evil (bottled water now costs $100 because there is lead in the tap water).
The point of pricing water to that level is that it would induce other people who have access to bottled water to bring it to that market, as is desirable
No, the point is to selfishly profit-maximise. I'm not trying to be difficult in saying that. The thing you describe is not the intent, it's the hypothetical effect. It may or may not do that (I don't think it typically does, take toilet paper during COVID for example).
Yes, the point for the individual setting that price is to selfishly profit-maximize. The point for us accepting a system that does this is because it signals to other water-bottle-holders that there is a dire need nearby and pays them to meet that need.
I don't think the example of a meme-driven pseudo-shortage of a paper good during a once in a lifetime global pandemic (causing both supply and demand shocks and significant information problems) is a very good point.
You assume that other people can simply bring bottled water to market & compete with discoverability and access to customers with established players?
Or is your point that all people in a market with leaded water should be paying $100 for pure water because it is inherently worth that much per the market.
No, I assume that if anyone can bring bottled water to market, they should have a strong incentive to do so whenever there is a strong need for more of it.
But they (everybody) can't. Bringing bottled water to market requires a clean source, rights to acquire it, and a manufacture & distribution network. Plus retailers. As well, these things are often blocked to newcomers because of existing deals with big players.
> But they (everybody) can't. Bringing bottled water to market requires a clean source, rights to acquire it, and a manufacture & distribution network. Plus retailers.
I didn't say everybody. I said anybody who can.
What you describe is exactly why it's important to have an incentive for the people who do have those resources to employ them towards getting bottled water to this lead-poisoned region...
> All sorts of tools try to prevent dangerous/destructive uses
But they don't threaten their users or have an "N strikes and you're out" policy. I take those safety caps off of all the chemicals in my garage because I'm a grown-ass adult and those caps are a pain in the butt. I would not expect the manufacturer of a solvent to show up at my house lecturing me about safety and threatening to ban me from buying his products.
Sure but they would if they could. If they knew idiots were doing idiot things with their products (or evils doing evil things) and did not utilize available methods to prevent them, then the company ends up holding liability. And no, this is not easily signed away in a contract.
Uhh right, but describing that as "dystopian" is frankly hysterical.
It's an obvious corollary of good things (like product liability). Virtually everyone I've heard complain about these safety rails was up to antisocial (at best) stuff. I've never heard a sympathetic use-case. It's objectively good that companies can be held responsible for misuse of their products and that they are therefore incentivized to mitigate misuse.
"My inability continuously attack product guardrails to enable my super esoteric (and probably antisocial) use-case is dystopian" is just... not a compelling argument.
"These safety rails" was referring to LLMs, which have far more nuanced and capable safety rails than chemical caps do, and accordingly also have much more assertive ways to enforce them.
It's the same underlying principle. If I want to ask a software tool what the suicide rate is for my county, I do not expect it to come back with: "Naughty boy! You said an unsafe word! You're getting a strike, and if you get two more, you're banned." This is totally out of the ordinary for a software product, and is absolutely a modern invention. Replace "suicide" with whatever the "AI Safety" obsession word is today.
> If I want to ask a software tool what the suicide rate is for my county, I do not expect it to come back with: "Naughty boy! You said an unsafe word! You're getting a strike, and if you get two more, you're banned."
Did this happen?
I just tested this query in Grok, Gemini, Claude, and ChatGPT and 0% of them admonished me or refused to return an answer.
Just like every single conversation I've ever had on this topic, you have to make up examples that aren't even true. Why don't you just share what you were doing that you feel you were unfairly prevented from?
Which would be more than 0% concerning if I've ever heard (even once) an example of this happening with a query that shouldn't actually trigger something like that, or is so close to such a query, that the false positive is understandable and of incredibly niche value anyway.
OP gave an example of reverse engineering, something that to the LLM looks identical to just hacking. I am totally fine if the incredibly tiny little fraction of people who want to reverse engineer their own systems can't use LLMs to do it, and in exchange top LLMs aren't helpful for the hordes of actual malicious actors who would love a superintelligence to aid their crimes.
No-brainer tradeoff, just like 100% of examples I've ever heard.
I don't think that "dystopian" necessarily goes far enough, this would be one of the rare times where I would call it a fascist mentality - the idea that everything's primary allegiance is to the state and the goals of the state rather than those of the customer or the user.
I want a default that has people empowered, rather than something where it's just another performative smokescreen caused by overzealous product liability. I'll thank you and your kind for needing to distractedly tap the "Agree" button on my car's infotainment every time I start it to confirm that I will pay attention to the road.
"the state" is just shorthand we use for "other people in my community"
> I'll thank you and your kind for needing to distractedly tap the "Agree" button on my car's infotainment every time I start it to confirm that I will pay attention to the road.
Does that actually mitigate antisocial usecases? No? Then it's not what I'm talking about :)
Of course if you wanted to you could just share specifically what totally-reasonable LLM use-case you have in mind that's neutered by this "fascist mentality" instead of dreaming up unrelated instances.
> "the state" is just shorthand we use for "other people in my community"
It's a very different abstraction layer, in the same way as individual cells vs the entity that is you. The entity that comes together from all those "other people in my community" and its priorities are different to the individual desires.
> Does that actually mitigate antisocial usecases? No? Then it's not what I'm talking about :)
Maybe it does? Maybe someone is alive on the road today because they read the message and changed their behaviour. I'm giving an example of something where this liability mindset has created a world where manufacturers are no longer prioritising the desires of their users in order to appease a sense of harm-reduction. And you weren't limiting it to LLMs you were applying it to all sorts of tools.
I think that "reverse engineering" as the OP was talking about is one of those things where maybe 1/10000 uses could actually be harmful. This is not even a high-risk request such as to produce a weapon of some kind where maybe your "antisocial usecases" could be applied.
reply