Every computer is a general purpose computer. What would you do to force me to participate exclusively in your nanny-net? I suspect any answer to that question requires an incredible amount of coordination and participation.
The real problem with the internet, as I see it, is centralization. This is a product of monopoly, which is the core feature of copyright. A truly better internet would replace the authoritative structure of copyright with a truly decentralized model.
As far as I can tell, the only hard problem left in decentralized networking is moderation. No one wants to browse an unmoderated internet. The problem is that moderation is structured as an authoritative hierarchy, so it's not compatible with true decentralization.
I propose we replace moderation with curation. Every user can intentionally choose the subset of internet they want to interact with, defined by attestations from other users, all backed with a web of trust. This way everyone is the highest authority, and users can help each other avoid content they are disinterested in.
Trump proved to the Republican party that a politician only needs one thing to win: engagement. What's the most engaging policy? Bullshit.
Bullshit gives you free advertising. Bullshit makes all of your political rivals fight you, which makes them look bad. Better yet, your political rivals get busy fighting the bullshit itself, leaving you free to shovel more. There is no such thing as winning a fight in politics: you only win the vote.
Most voters in the US are convinced that there is nothing more to politics than bullshit anyway. This division is rooted in our two-party system. Because of first-past-the-post voting, this is a mathematical inevitability: every vote is a vote against whoever you dislike more strongly. Remember: dislike is not generated by valid criticism as well as it is generated by bullshit engagement. Politics in the US is a team sport, and nothing more.
Of course they can! They just did right there. You don't have to accept their argument. You can be a heartless sycophant and reject it out-of-hand because their argument doesn't follow the rules of the very framework they are arguing against. That's your choice.
Progress is being made in both directions on this front. If you truly believe in the law, then argue for it. Considering the fact that the current POTUS is a convicted felon, and the supreme court has practically eliminated all checks on his power, I'll be taking arguments for lawlessness seriously.
There is no objective truth. Everything is arbitrary.
There is no such thing as "accurate" or "precise". Instead, we get to work with "consistent" and "exhaustive". Instead of "calculated", we get "decided". Instead of "defined" we get "inferred".
Really, the whole narrative about "AI" needs to be rewritten from scratch. The current canonical narrative is so backwards that it's nearly impossible to have a productive conversation about it.
Even if there isn't an underlying problem, the only real way to change a habit is to replace it with a new one.
I have noticed that usually people who make it their mission to stop doing a thing are replacing that thing with the mission itself. This strategy is always bound for failure, because the moment it starts to work for them is the moment they end the mission. This is when, instead of reevaluating their strategy, they punish themselves for the failure to not do. The cycle repeats, and the person spirals into rumination about their stress.
I didn't just stop biting my nails. I started trimming them instead.
An LLM is not a database. There is no significant amount of information in a model that can be accessed 100% of the time. This is because it's a mystery to the user what collection of tokens will lead to a specific output. To get a predictable result from an LLM 50% of the time is very significant.
This doesn't tell us for certain whether or not the model was trained on a full copy of the book. It's possible that 50-token long passages from 42% of the book were, incidentally, quoted verbatim in various parts of the training data. Considering the popularity of both the book itself, and derivative fan-fiction, I would not be surprised. I would be less surprised to learn that it was indeed trained on a full copy of the book, if not several.
The more meaningful point here is that the ability to reproduce half a book is the same sort of overt derivative work that is definitely considered copyright infringement in other circumstances. A lossy copy is still a copy. If we are to hold LLMs to the same standard as other content, this isn't very easy to defend.
Personally, I see this as a good opportunity to reevaluate copyright on the whole. I think we would be better off without it.
1. Strong copyright to prevent competition from undercutting their related businesses.
2. Exclusive rights to totally ignore the copyright of everyone that made the content they use to train models.
I personally would much prefer we take the opportunity to abolish copyright entirely: for everyone, not just a handful of corporations. If derivative work is so valuable to our society (I believe it is), then I should be free to derive NVIDIA's GPU drivers without permission.
Everything is derivative. This boundary you are defending between originality and slop is extremely subjective at best. What harm is slop anyway? If originality is so objectively valuable, then why should its value be systemically enforced?
At the intersection of capitalism and copyright, I see a serious problem. Collaboration is encapsulated by competition. Because simple derivative work is illegal, all collaboration must be done in teams. Copyright defines every work of art as an island, whose value is not the art itself, but the moat that surrounds it. It should be no surprise that giant anticompetitive corporations reflect this structure. The core value of copyright is not creativity: it's rent-seeking.
Without copyright, we could collaborate freely. Our work would not be required to compete at all! Instead of victory over others' work, our goal could be success!
We know what the world looks like without copyright and that world has far fewer works created and very few artists who can do it full-time absent patronage or independent wealth.
Banning the nonsense that is character copyright and shortening copyright back down to a reasonable length of time (say, 20 years) would still enable the creation of more culturally-relevant derivative works without pauperizing every artist.
How could we possibly know that? Copyright has existed since before the industrial revolution even started. What you described is not really that far from reality today: most artists are not really making a living. The words "starving artist" have not even begun to lose their meaning. Every artist I know has been failed by copyright. The value a copyright creates is not applied to the art: it's applied to the moat around the art. The only certain beneficiaries are the giant corporations that use their collected moats to drown out small competition, including artists.
The copyright laws that existed prior to the industrial revolution only existed only in a small number of countries. A large swath of the planet had no equivalent.
Even British Colonial America had no copyright, save a handful of exceptions, as the Statute of Anne did not apply to the colonies.
That's the difference between bias and logic. A statistical model is applied bias, just like computation is applied logic/arithmetic. Once you realize that, it's pretty easy to understand the potential strengths and limitations of a model.
Both approaches are missing a critical piece: objectivity. They work directly with the data, and not about the data.
Tesla wants to be isolated from criticism. Why? It's not just because this will show how bad they are relative to competition. The real reason is much more concerning.
Criticism of Tesla would deconstruct their dualist narrative. Tesla has sold the public on the notion that "good enough" self-driving is objectively safer than human driving. Anyone who accepts this narrative can consider the failure of human driving safety as an ultimate bad, which implies that Tesla's automated driving alternative is an ultimate good. This dogmatic thinking hinges on Tesla's vague assertion that automated driving in general is statistically safer than human driving in general. As soon as people engage with any criticism of this narrative whatsoever, the dualist perspective is lost, and the narrative itself falls apart.
The real problem with the internet, as I see it, is centralization. This is a product of monopoly, which is the core feature of copyright. A truly better internet would replace the authoritative structure of copyright with a truly decentralized model.
As far as I can tell, the only hard problem left in decentralized networking is moderation. No one wants to browse an unmoderated internet. The problem is that moderation is structured as an authoritative hierarchy, so it's not compatible with true decentralization.
I propose we replace moderation with curation. Every user can intentionally choose the subset of internet they want to interact with, defined by attestations from other users, all backed with a web of trust. This way everyone is the highest authority, and users can help each other avoid content they are disinterested in.
reply