If you're misusing LLMs to solve TC^0 problems, which is what the paper is about, then... you also don't need the slop lavine. You can just inject a bunch of filler tokens yourself.
You’re not doing anything wrong. This isn’t a bulletproof idea. It can work, and this is what a lot of people end up with to manage complexity, but there’s a critical point beyond which things collapse: the agent can’t keep the wiki up to date anymore, the developer can’t grok it anymore.
Managing complexity, modularity, separation of concerns, were already critical for ensuring humans could still hold enough of the system in their brains to do something useful.
People who do not understand that will continue to not understand that it also applies to AI right now. Maybe at some point in the future it won't, not sure. But my impression is that systems grow in complexity far past the point where the system is gummed up and no-one can do anything, unless it's actively managed.
If a human can understand 10 units of complexity and their LLM can do 20, then they might just build a system that's 30 complex and not understand the failure modes until it's too late.
> People who do not understand that will continue to not understand that it also applies to AI right now.
I think this is mostly a matter of expectation management. AIs are being positioned as being able to develop software independently, and that’s certainly the end goal.
So then people come in with the expectation that the AI is able to manage that, and it fails. Spectacularly.
The LLM can certainly not manage any non-local complexity right now, and succeed in increasing the technical debt and complexity faster than ever before.
I understand freedom of speech and I understand she's free to speak but there may be consequences. I understand that there are huge complexities in the legal system. I understand you can enter into agreements (part of your speech) that effectively gives away your speech. But if you step back and look at this situation, it's just fucked up that a corporation can do this to you. If freedom of speech is supposed to be inalienable, these types of agreements should not be legal.
disclaimer: She lives in the UK and I'm speaking from a US perspective.
The corporation did not do this to her. It was a two party agreement. She bears just as much blame for the agreement as the corporation. She entered into it willingly. And that does and should have consequences.
Morally speaking I think the company is reprehensible. But nor do I think contact law should be changed because of it.
The antidote to a power imbalance is to recognize that there is no power imbalance and go about your life that way.
Pretending there is one lands you in an imaginary trap. Build a society where we recognize that and you build a society where the imaginary trap disappears.
You're the one pretending here. The economy is unfortunately designed around most people relying on an income stream that remains at the whims of someone else.
The UK is far worse, with draconian libel laws where the burden of proof is on the defendant. Originally designed to stop uppity commoners from challenging the aristocrats, now used by oligarchs to silence journalists.
> I understand she's free to speak but there may be consequences
nit: this isn't generally a valid analysis. Rather, it's a common refrain used by people undermining freedom of speech while pretending to support it. This trope is often even trotted out in full-powertalk mode where it's applied to consequences coming from the government itself.
You just lost your job, through no fault of your own, or maybe because you did the right thing and blew the whistle on illegal and/or unethical behavior. You don't know how long it will be before you find a new job or how you are going to pay the bills until then. Your employer offers you some money to tie you over, and maybe some resources to tide you over until you get a new job, but you have to agree to a hundred pages of legalese. And you only have a few hours to decide, not enough time to have a lawyer look at it, even if you could afford to pay a lawyer. You are highly stressed, so even if you take the time to read it, you probably don't take it all in, and feel pressured to agree. And your employer also makes vague threats that about what will happen if you don't sign it, like having a hard time finding a new job or maybe having legal action taken against you.
Does that seem like a free informed decision?
Looking at it another way, anti-disparagement agreements are basically bribes to keep quiet even if disclosure would benefit the general population.
I would argue yes. If you have the choice to sell to sell it, you can be forced to sell it.
One can still give up their basic rights if they so choose. The woman in question can cease from disparaging Meta for the rest of her life. A person can opt to enter in to being a slave to another for the rest of their life. I can choose to follow one religion or another or none at all. But one should never have those options taken from them.
In Germany, this sort of thinking is the reason you can't release anything into the public domain. People are presumed to be too stupid to be trusted with the decision to renounce their copyright and so they are "protected" from this possibility.
Have you ever been conned into releasing something into the public domain? Me either. Its not a real problem. But signing over the rights to some corporate party? That happens all the time, and is permitted in Germany. Germany is being very stupid here. They're letting abstract reasoning about principles blind them to common sense (many such cases in German history.)
Washington state does this with some parts of tenant law. There are certain tenant rights about habitability (heating, etc.) that are protected by law, and cannot be nullified by a rental/lease contract, even with consideration. The law states that the tenant legally cannot waive those rights.
Both non-disparagement and non-disclosure agreements should—just as many jurisdictions have for non-compete agreements, which do not even implicate free speech the way the others do—be sharply limited as a matter of public policy (non-disparagement even moreso than non-disclosure.) Both are routinely used to inflict public harm for private gain, and government enforcement of either is in tension with freedom of speech; while there is a legitimate case to be made that non-disclosure agreements within certain bounds have a certain degree of necessity in enabling legitimate business, this is a much harder case to make for non-disparagement agreements, at least for ones that are not temporally bounded within an active business relationship.
Depends on what type of non-disclosure. Disclosing technical guarded and not publicly known technical know-how - I am ok with those. Disclosing that boss treats people like trash should be allowed and I think lawmakers should have enough intelligence in their brains to make laws accordingly.
> Disclosing technical guarded and not publicly known technical know-how - I am ok with those.
I would love to see NDAs for trade secrets limited in a way that incentivizes companies to rely on patent protection instead, where the system is set up to ensure that knowledge eventually becomes public record and freely usable by anyone. It would be very interesting to see how eg. the tech industry would change if trade secret protection were limited to a meaningfully shorter duration than patents.
If you don't believe that people should be able to sell themselves into slavery, you should start by offering your list. If you do believe that people should be able to sell themselves into slavery, then unlimited freedom of contract is a basic freedom for you.
What you shouldn't do is pretend not to understand.
I'm not the one making a positive claim. I haven't even claimed such rights exist so why on earth would be the expectation be that I list them? You've assumed that I believe in this shared fiction.
We sell ourselves into a form of slavery every day. Some would argue that is a big driver of our current society and way of life.
Freedom of speech is far from inalienable. Non-disclosure agreements are most relevant, but every country on earth also has at least some regulations regarding hate speech, threats, incitement, purjury, or defamation — not to mention security clearances or state secrets.
And that's where the complexity arises in this argument that I don't know how to resolve. In the case of this woman vs Meta, to me it doesn't "feel" legal that one disparaging comment costs $50K. It feels that there's something wrong here that should not be allowed despite her entering the agreement. Maybe I don't believe she should have been allowed to enter the agreement.
But I understand that my point of view doesn't match legal code. Just feels fucked.
To be clear, I do agree with everything you've said here — I just disagree that freedom of speech is an inalienable right, and I don't think there's ever been a time or place where it has actually been considered one.
If it were up to me, I would require non-disparagement agreements to be standalone contracts, and cap the damages a company can claim to the amount they paid you to sign it. Once that number is met, the contract is void. That way the company only gets as much leverage as they're willing to pay for.
The majority of people will self-alienate themselves in exchange for power or even just survival
Think of a person digging their own grave under threat of immediate murder (tons of well documented examples). This is the maximum self alienation: do work to make life easier for your oppressors.
In my 41 years it seems like the majority of people are content digging their own graves
There's a basic list on that page. There are many LLMs out there that you can discuss this with if you want to waste your own time. I'm not going to waste any more time with this thread. You have an attitude that says "Debate me but you'll never convince me". If you'd like to learn something, there are many resources on the internet available to you.
Only for a very narrow definition of slavery. Arguably constructing society such that it costs so much to just exist (for example, by artificially restricting housing supply) and thus you have to work is not all that different to slavery. I would say the dollar is but company scrip with better PR.
Well now you're equivocating. We've established one that you can't deal in in a specific country. _Should_ is quite a different question. You can't establish should by establishing is.
Well that would seem to make the rights in question not particularly inalienable. In fact if we're talking about the US slavery _is_ legal in certain contexts. So it's definitely not inalienable. Only in the context of voluntary agreements between private citizens.
This has nothing to do with the founding fathers. The Ancient Greeks talked about natural law. The UN passed the Universal Declaration of Human Rights. 193 countries have ratified at least parts of it.
Again, I beg you to at least read a paragraph or two off Wikipedia.
The specific term 'inalienable' is heavily associated with the founding of the US. The others are different things but not very different in substance, i.e. ultimately some guy claimed these are universal rights. Wikipedia is not going to make appeal to authority work any better as an argument, I'm afraid.
> Whereas recognition of the inherent dignity and of the equal and inalienable rights of all members of the human family is the foundation of freedom, justice and peace in the world...
Is the US not in the UN now or something? The whole UNUDHR was an Eleanor Roosevelt project. She literally drafted the documents! At least look it up before being rude. You need to get the knowledge before applying the sass.
No I just mean in the sense that I give over the rights to my own words. I can't repeat them outside of the context that I've agreed to. They were both examples of the same kind of agreement. They'll keep those rights well after I'm dead, by the way.
You're not giving over the rights, you're selling the right to profit from them under contract.
You can argue that contract law is essentially a battle of relative political and economic power, and IP and employment contracts will always be unfair unless limits are set by statute and enforced enthusiastically.
And personally I would.
But generally you're signing away the rights to specific text, not the insights or commentary in that text, and if you freelance there's nothing to stop you making your points through some other channel, and/or some other text.
If you're a full-time employee then the usual agreement is that your words (code) are work product and owned by your employer, and you're in that situation because your political and economic power is relatively limited.
> You asserted that free speech was an inaliable right, they provided an example showing it's not.
No.
"Inalienable right", like the "right to bear arms", has never meant you get to do anything with it. Free speech doesn't extend to defamation; free expression doesn't extend to murder; freedom of the press doesn't extend to sneaking into the CIA's archives, freedom of movement doesn't apply to jails.
I'm of the opinion that arbitration clauses and non-disparagement agreements of the scope involved in this particular case are unconsionable, because they unreasonably infringe upon such inalienable rights.
(I don't agree - re-read my wording carefully - but some certainly take that position. My point: those who do still tend to take the "but there are limits!" position on, say, home-brewed nukes.)
In each case, though - constitutional right, human right, inalienable right, natural right - the fundamental concept of "sometimes two people have rights that conflict, and society has to resolve this" applies.
We're talking about whether people should be permitted to sign away their right to speech. I think you've conceded that such is permissible at least for a limited duration. Shall we quibble the permissible durations, or are you done?
Sure; we have to resolve conflicts between two sets of people with rights sometimes. The inalienable right to free speech doesn't extend to defamation and fraud; the inalienable right to freedom of movement doesn't apply to jailed murderers.
Unconscionability is a bit like obscenity; hard to perfectly define, but sometimes quite clear.
Those are not inalienable rights either, they're legal rights. Here, courtesy of Cornell law:
"An inalienable right is a fundamental entitlement inherent to every person that cannot be sold, transferred, or taken away by the government. These rights, often called natural rights are considered essential, cannot be surrendered by the individual, and are not dependent on laws."
Just Google "is x an inalienable right" next time.
I'm not sure most people are that naïve that they can't differentiate between "any computer that acts smartly" (how the term "AI" is used) and the word chatbot. Of course, LLM is even more precise
Tangential question: what do you call transformers-based models that generate images or videos? Are they LLMs? They're not really "language" models. But there's not really an easy term for them. Maybe "image models" and "video models"?
I'd call them video/image generators because I don't know the word for the underlying technology. I can't imagine they're LLMs because, indeed, the second L is for language
Today I learned, Stockfish moved to neural network on 2023. I knew that it was just a minmax with alpha beta pruning and a really good eval function. Now its not.
> I knew that it was just a minmax with alpha beta pruning and a really good eval function. Now its not.
It is still "just" a minimax with alpha beta pruning, except the eval function is now a neural network. NNUE, to be more specific.
I highly advise anyone who is curious about chess engines, but hasn't heard about NNUE to read about it. I find this technology absolutely fascinating.
The key idea is that a neural network is structured in a way that makes it very cheap to calculate scores for similar positions. This means that during a tree search, each time you advance or backtrack you can update the score efficiently instead of recalculating it from scratch.
I mean, it still is. Now it just has a really good neural net-based eval function. Don't be fooled: it's not that stockfish just has "a really good eval function", and that's the only thing that makes it as strong as it is. The actual tree search is _incredibly_ sophisticated, with boatloads of heuristics, optimizations, and pruning methods on top of alpha-beta.
What are you on about? Is this about how when people use an English term in a particular way, then their listeners/readers begin to use it too? If so, yes, it's a self reinforcing mechanism called lexical dissemination, and I'm very curious to hear about how you'd disrupt it.
Honestly the PELOSI act is a step in the right direction, and should be extended to include all kinds of cryptocurrency and real estate. Also the president and his extended family should be included in it.
The only problem is in the name. It's clearly just to score political points. I'd be surprised if it gets approved.
There are more recent versions with better names, like the Restore Trust in Congress Act[0], and one introduced a few days ago to ban congresspeople and the president (as well as their immediate family members) from betting on prediction markets[1].
I can only imagine the GP mentioned the PELOSI act (which, as I recall, was cosponsored by Nancy Pelosi and blocked by Republicans) because it was designed to be partisan messaging.
One theory is that the talking points are seeded by a set of paid supporters on platforms like Twitter, Facebook and Reddit. These people live in low income countries and can use LLM + broad directional instructions to mass produce comments in support of the regime.
The talking points that are successful are then reinforced by genuine regime admirers, enter the canon and spread. There’s no verification mechanism for bad or wrong ideas, since we’re in a post truth society.
The goal is to uphold the regime. The system trying to be stable and defend itself from the fallout of its actions. The actions are actually guided by an ideology plus personal interests, so they can’t be optimal.
I'm not sure citing a Qatari propaganda site with a well documented history of employing members of terrorist organizations as "journalists" is helping make your point.
And here I thought aljazeera was a respectable journalistic outfit. They certainly seem to be doing better than most of the corporate US media by my estimate.
> And here I thought aljazeera was a respectable journalistic outfit.
While they may do some legitimate journalism they are well known for highly biased articles especially when it comes to anything relating to the middle east.
> They certainly seem to be doing better than most of the corporate US media by my estimate.
In what way? News organizations in the US like CNN certainly have higher standards than aljazeera.
> Perhaps you should back up your mud slinging?
Journalists who did work for aljazeera even held hostages in Gaza[0].
Holy shit, you're right! The country murdering thousands of captive civilians and spreading proven lies about the status of hospitals as terror bases reported it, so it must be true! I'll never trust Al Jazeera again, now that I know they take hostages.
That's just a really bad argument that falls flat on its face and discredits you.
Let's the the documents you're speaking of (and not those fake IDF spreadsheets which list some of them as being Hamas members when they were 6 or 7 years old).
Bari Weiss's CBS said a journalist who was killed was someone they worked with or some stupid shit when they were clearly a colleague or possibly more. I don't remember enough to be more specific, but that type of minimalism is rampant.
Israel kills journalists and civilians. Oh shit, civilian:
You're right, execs keep trying to fit the LLM square peg into the "inteligent agent" round hole.
Developers use it, for groking a codebase, for implementing boilerplate, for debugging. They don't need juniors to do the grunt work anymore, they can build and throw away, the language and technology moats get smaller.
The value of low level managers, whose power came from having warm bodies to do the grunt work, diminishes.
The bean counters will be like when does it pay for itself. Will it? IDK, IDC.
Validation efforts likely become more necessary, so costs rise in another area. And product managers find they still need someone to translate the requirements well because LLMs are too agreeable. Cost optimization still needs someone to intervene as well.
I know there's an attempt to shift the development part from developers to other laypeople, but I think that's just going to frustrate everyone involved and probably settle back down into technical roles again. Well paid? Unclear.
reply