Also, the dark pattern where it shows the interface and lets you enter a prompt/set settings, but then pops up the 'create account' dialog when you press submit is pretty annoying.
>"Massgrave, the group behind Microsoft Activation Scripts, has noted that Windows setup sends hardware info to Microsoft and receives identifiers back that are later used for Store access and licensing. Blocking GDID assignment breaks both activation and UWP apps."
I think probably it sends the gdid back to Microsoft as part of telemetry/updates/MS account periodic re-auth, which lets them know the current IP address, and then once the government has Microsoft's logs and the various target websites' logs, they can correlate based on IP address. I don't think it's actually sending the gdid to the web sites. Maybe.
This serves to further illustrate that nobody should be using Windows for anything that involves the need for privacy. And doubly, triply, and morefold so, nobody should ever sign a Windows machine into a MS account for any reason.
The article is not about "should people be allowed to buy this product because it's potentially dangerous/addictive/etc" but "Should the company be allowed to sell this product because it consists of acetaminophen plus two useless ingredients and is basically a scam".
This is the Platonic ideal of an LLM-generated blog post; it's pages long, it uses a lot of flowery language, and communicates almost zero actual information. I can't disagree with the premise, because the premise melts away when you try to pin it down, like cotton candy washed in a stream by a raccoon.
Congratulations, person who wrote the prompt that caused a machine to generate this article, you have truly achieved some sort of perfect something.
I agree; the situation "young people know all about computers" was maybe a twenty-five year phenomenon; before that, there weren't enough computers for it to be generally true, and at some point you no longer needed to know all about computers to successfully use a computer for the sorts of things the general population does.
The issue with this is that we don't know how it works. Generally speaking, we know how the level of abstraction that we were born with works. We might have some understanding of one or two previous levels, but that decreases the farther down you go. We might understand the next level, and some of the next after that, but eventually people will be making things that we don't have the context to understand without having to unlearn a lot of what we know now.
I'm old enough to see this process in action; I used to be young and in possession of esoteric knowledge that made me infinitely in demand and now most of the things that young people have esoteric knowledge about is things that I don't particularly care about, and I'm left with a lot of finely honed skills to solve problems that have mostly been abstracted away.
Isn't most of our economic growth based off inputs of energy (in the form of labor, electricity, etc) that ultimately derive from the sun, though, rather than primarily non-renewable non-recyclable resource extraction? The sun is technically a non-renewable resource, I guess, but only on cosmic timescales.
This has always been a thing with IT advice, though - the more complex a system and the outcome, the harder it is to clearly define "better" or "worse". Add in the fact that LLMs are intensely and emphatically non-deterministic and LLM guidance basically becomes gardening advice.
Heck, even the 'benchmarks' are mostly somebody's attempt to crystallize their vibes with varying amounts of success.
Right, but that's still external to the LLM, it's just a KV cache that's stored on the provider side for performance reasons, so that the client doesn't have to re-send the whole chat history with every subsequent call in the conversation.
It still generates every response using the model's pristine state with every new API call; whether the context is provided from the client or from a colocated cache server doesn't really change that.
The fact that a LLM is essentially immutable would be my biggest argument against consciousness or self-awareness.
It's a big file with a bunch of coordinates describing spatial relationships between tokens. When you give it a prompt, it uses those relationships to generate a string of tokens that is a statistically likely response to that prompt, then it stops. It's not changed by the experience. It doesn't remember anything. It doesn't sit around thinking on its own.
Even if the model itself were extremely complex, it's hard to imagine a definition of consciousness that includes something that doesn't remember and can't change.
First and most importantly, it's not really about LLMs, it's about AGI, and the second does not necessarily follow from the first; LLMs in their current state are pretty clearly not AGI, and most of the LLM-world progression in the last few years has been about better tooling/interfaces, refinements in training data and techniques and people learning how to use LLMs effectively rather than the huge leaps in fundamental capability that we saw in earlier years. It seems more likely that at this point, when AGI comes, it will be something entirely new or something that LLMs are only a component of, rather than "we built an LLM with ten trillion parameters and suddenly it became God".
Second, it's not even really about AGI, it's about AGI superintelligence. And more than that, it's about affordable AGI superintelligence, assuming that such a thing won't cost billions a year to operate.
I've found that the AI overview is usually right but confidently wrong enough of the time that I don't trust it. The interface that you get with the 'AI Mode' button (which I assume is just Gemini with very low compute settings), however, is usually pretty solid for well-documented queries.
I think this touches on the core difference between good and bad use of AI; using AI as part of the process vs cutting and pasting LLM output.
Use AI as part of the research process, to help understand a concept or problem. Use it to format data, or as a part of the design or brainstorming process. Use it to build manageable portions of code that you can read and understand before committing. But if the output doesn't go through your brain somehow before you unleash it on the world, that's really no different from a seventh-grader Googling the subject of his homework and then cutting and pasting the entire text of the first result, headers and all, and turning it in.
Was it the actual raw chain-of-thought? I know GPT-5 will emit thinking tokens, and while they're an interesting insight into the 'reasoning' process, they're apparently pretty heavily sanitized presumably because the raw thoughts could reveal proprietary training info that's part of their moat.
I think the strongest argument against AI consciousness is that there's no persistent internal state, no feedback, and no change; the 'conversation' you're having is a series of one-off API calls where each subsequent call is provided enough information about the previous calls for it to generate a plausible response.
If we (very, very generously) assume that an LLM is a structure capable of conscious thought, then it's still not conscious - we've created a representation of a brain which we turn on for a fraction of a second to generate text and then return it to the representative state. There's no opportunity to develop consciousness, it's a brain trapped in stasis.