The language model is "prompted" for that goal and the language model itself was not trained with that goal embedded into the training process.

The AI does not seem to have been programmed with the basic goal of making a profit.

The prompt probably started with that. The problem is ALL the data a LLM disseminates gets appended to the prompt. That is how an LLM works. Therefore.. by sending over new data you can manipulate the outcome.

For a Linux shell analogy.. the Initial system programming is like a .bashrc, And when your AI talks to people - they get access to a bash prompt. Of course they can manupulate the shell to override directives that came earlier from the preloaded .bashrc.

For a car analogy: You were taught to always come to a complete stop in driver's Ed. But on the road you have another driver shouting at you not to stop at the stop sign; So you get taught to cease stopping at stop signs.

Language models tend to give priorityTowards words that come later over words that come later. Also; your AI inference has a limited working memory size or context window.. Eventually you run out of space, And the earlier words need to be summarized in order to fit the entire combined prompt within the limited time and space for the AI.

If you Want to somehow prevent this; I am afraid the only answer is You will need to combine multiple AIs, and have a true supervisory process.
You need a dedicated AI to read the untrusted user input and Vet that input, and ensure that input is safe before any of it can be h anded over to the empowered AIs for processing. You need extra guardrails and filtering systems to make sure a clever prompt hacker does not persuade the supervisory AI to allow them a sandbox escape.

The Supervisory AI over the empowered AI, and the filtering AIs for user and untrusted inputs need to have their own Independent prompt stacks. The various filtering AIs and supervisory AIs need a number of specialized "panic buttons" that harness True stop controls instead of merely being able to provide suggestions or advise. For example: A supervisory panic initiated a predefined process and blocks further decisions from the impacted unit from being approved until a series of conditions are met to allow a status reset.

It seems like there needs to be a lawsuit initiated against Apple and the retailer for the full value of that account.

it is not a separation of powers issue, in the same way that a congressional staffers refusing to tell the POTUS what their bosses say in closed door meetings is not a separation of powers issue. By your line of reasoning, there is no person in government who does not work at the pleasure of the POTUS except for the elected officials and the justices. Everyone else. Every staffer, every janitor, scientist, soldier, sailor, health inspector is beholden to a single man. And there is nothing anyone else in government can do about it if he decides to fire them. That is not democracy. That is not a balance of powers. That is dictatorship. you are advocating for dictatorship.

Congress has the ability to delegate its authority to regulatory agencies they create, in the same way that the POTUS can delegate his authority to Cabinet members. When the POTUS tries to undermine the independence of the FTC (or any other independent agency), he is not taking back to him powers that were his. He is violating the very separation of powers you claim to care about, by seizing control over congressional authority simply because the job of the agency requires some level of 'execution'. Regulations are derived from, and therefore an extension of the law, which is the responsibility of Congress. The president was originally envisioned, not as the leader of the government, but as the executor of the will of congress with oversight from the court.

Take a look at the list of (~40) independent agencies in the US government. There are A LOT!
https://en.wikipedia.org/wiki/...

Now think about all the things that those agencies do day to day. Think about the technical expertise necessary to perform these important government functions. And then think about the skillset of the average congress critter.

Congress is full of politicians, which means folks whose primary job qualification is the ability to convince rich people to part with their money for the purposes of running a campaign. And their secondary qualification is being able to read the political winds in their district, and portray themself as the "Right kind of person" for their constituents. Nothing about financial acumen, or experience with Trade, logistics, economics, etc.

Fact is congress DOES write the laws, which are a form of high level mission statement. And then they delegate implementation to subject matter experts. And then they empower those experts to do their jobs with a degree of freedom from the whims of politics. It is by no means perfect, but it is far better than the patronage system of old, or the crony capitalism of the current trump administration.

An analogy is home building. If you want to build a new home, you sit down with an architect to design your home. They decide, with your input, where the kitchen should go; how big to make the family room; how many bed rooms, floors, etc. Then, they have to hand those plans off to an engineer to figure out how to actually build that house safely and within code. Architects can draw anything they want on paper, but only an engineer can tell you how to actually build it, or if it even can be built. Congress serves as the architect for our laws. Federal agencies serve as the engineers. Figuring out HOW to implement the design in a practical way. They also serve as the home inspectors, to make sure that the builder actually followed the blueprints the engineer approved. Since we can't trust our politicians not to be in the pock of the home builder.

Unless you are a lawyer I would suggest you not presume to judge the constitutionality of delegated powers and independent agencies

the Supreme Court has upheld the constitutionality of both independent agencies and delegated powers in the past. This court may not, but that has more to do with their focus on giving the Trump administration as much leeway as possible, and nothing to do with what the constitution actually allows. After all, they have upheld his appropriation of similarly explicit congressional powers related to spending.

Follow the money.

If you cannot see the difference between "we will set the best policy we can, even if it is something the president does not like" (how independent agencies operate), and "we will let the president set policy, and simply implement his will, regardless of how good or bad it will be for the country" (which is how agencies that are not independent operate) then you are not a serious person.

Independent does not mean unbiased. So your whole straw man about "does an unbiased person exist" is irrelevant.

In this context, an "Independent" agency, is one that is not specifically under the direct control of the executive branch, despite serving an executive function within its area of responsibility. It means that while the executive branch has some administrative responsibilities for the agency, it does not call the shots. Those appointed to lead the agency, by congress and the president, are not shackled to the whims of political theatre. It is a way to help an agency focus on good policy, and not on good political optics (at least to a degree).

...but we've already been mislead so many times before.

All I know about Volkswagon is they're a car company who deliberately cheated on their emissions tests.

No surprise their demand is falling away due to past outright illegal conduct.

Their license to manufacture a single new unit should have been cancelled the day this was found out. So I don't feel sorry for them.. surprised They did not have a complete shutdown imposed by the government sooner.

EULAs can not legally apply to a minor, just have your neighbor's kid set it up.

If you knew about the EULA, then hazarding to allow the minor to click accept still counts as you accepting it.

Judges are not keen to entertain "workarounds" like the kind you are describing. You can't avoid being deemed to have accepted a EULA by deliberately causing it to be accepted, no matter what method you pick.

I mean if such antics would work in practice; people could just let a cat click randomly; hex edit or NOP out the dialog display function from the executable, or figure out which bit written to disk or flash will bypass the prompt; etc. The concept of a clickwrap license would become a joke.

It's drivers for their AI chips.

they sold it, as most people would to erase their streaming service credentials.

Your appraisal of consumer security awareness is way too optimistic.

Most people would just sell the TV. It's uncommon to seem them also reset to default aside from tested units sold by some secondhand stores that clean up used gear before selling.

Streaming services; assuming an old TV was even used for those; normally detect if a device has gone unused or moves to a different ISP or geolocation and cancel the device token requiring a revalidation. The old TV's reason for being for sale might even be that Netflix, etc, revoked their compatibility with it due to its age or outdatedness.

Anyway. There is never any presumption the seller factory defaulted their TV before selling it.

The manufacturer can log the EULA acceptance, and they will most likely be able to report on exactly the date, time, and IP address when someone clicked Okay. In the case of a dispute; the onus would be on the manufacturer to show evidence that the customer agreed. That is if the customer disputes the alleged fact that a EULA was accepted by them.

Get a new TV and never, EVER let it connect to the network.

Be really really careful. Manufacturers keep coming up with more and more ways to get it just enough internet access to talk to home even if you don't want it too. Hidden cellular modems. New mesh networking protocols like Sidwalk. Bluetooth. Aggressive wifi autoconfig. Ethernet over HDMI. etc