I initially planned to run a chess tournament for LLMs but they are not good: besides obvious mistakes, they output incorrect moves, get stuck in loops by repeating the same moves and the smaller models fail to output valid json frequently. I thought the reasoning models like o3 mini might be good, but they are an incremental improvement in chess.
Feedback and suggestions for other games to explore welcome.
The article mentions, he is going to run the marathon, looking forward to what he can do in that distance. I feel it's only a matter of time until someone breaks the 2 hour barrier in an official race. Lot of people thought it would be Kelvin Kiptum, unfortunately he passed away in an accident.
No, it is politics. Apple stopped advertising on X over politics, not technical reasons. The Democrat-run media has successfully convinced a large number of highly conformist individuals that we've got another Hitler on our hands, and a lot of these people opine about their paranoia extensively on HN.
The impact this organization had was incredible. I doubt they would have been able to do this work if they were based out of any other country, which makes me wonder how the US legal system, regulators and law enforcement in general are not extremely corrupt. What reasons or incentives make the system work in the US? Of course there are many instances of corruption and injustice, but in comparison to almost any other country, it seems to work surprisingly well.
I think there is a really nuanced check-and-balance system that has extreme visibility for the federal legal system:
- investigators need approval from a prosecutor to move forward with investigations, and ultimately have to present their evidence in sales calls to their boss/peers. It’s a lot of red tape.
- prosecutors have bosses and reputations to uphold, they don’t want to take on risk.
- judges act as a procedural review for the prosecutor and watchdog for civil liberties
- the defense is red teaming the prosecutor and investigators for fraud etc
- the appeals court acts as a second level review for everyone + original judge
There are many reasons for this but primarily its simply that it is a wealthy country with a functioning legal system. US courts are generally fair; if biased. This has changed with many recent rulings by the SCOTUS. But there exists a culture of generally respecting laws (and an apparatus to enforce those laws).
Great work! When I use models like o1, they work better than sonnet and 4o for tasks that require some thinking but the output is often very verbose. Is it possible to get the best of both worlds? The thinking takes place resulting in better performance but the output is straightforward to work with like with sonnet and 4o. Did you observe similar behaviour with the 1B and 3B models? How does the model behaviour change when used for normal tasks that don't require thinking?
Also how well do these models work to extract structured output? Eg- perform ocr on some hand written text with math, convert to html and format formulas correctly etc. Single shot prompting doesn't work well with such problems but splitting the steps into consecutive api calls works well.
That's a good point. We don't see that in our experiments because it's all in the math domain. However for OAI it's plausible that training for o1 might conflict with standard instruction training, leading to less human preferred output style.
In this paper and HF's replication the model used to produce solutions to MATH problems is off-the-shelf. It is induced to produce step-by-step CoT-style solutions by few-shot ICL prompts or by instructions.
Yes, the search process (beam-search of best-of-N) does produce verbose traces because there is branching involved when sampling "thoughts" from base model. These branched traces (including incomplete "abandoned" branches) can be shown to the user or hidden, if the approach is deployed as-is.
OpenAI recommends using o1 to generate the verbose plan and then chain the verbose output to a cheaper model (e.g. gpt-4o-mini) to convert it into structured data / function calls / summary etc. They call it planner-executor pattern. [1]
I don't understand their api not being intended for individual use [0], are developers supposed to use this subscription only? The haiku model is pretty good for the price + available large context, and the opus/sonnet models are in the league of gpt-4, so I would have liked to pay for the api. I've moved on to llama 3 70B as a daily driver and it works really well! The only issue is it's small context size and it doesn't work great if you give it a lot of files to work with. Currently I'm forced to break down problems a lot more than I had to with gpt-4 or the claude models.
I know I can access the claude api through 3rd party sites + official partners, but there's no incentive to go through the trouble when llama 3 and gpt-4 apis work great for my use cases.
I heard of people using the GPT-4 API for personal use because it's a lot cheaper than paying for the subscription since it's pay-per-use. Maybe they don't want people to do that.
The idea of system 1 and system 2 had a profound impact on me. While specific conclusions in the book were reported to be based on low quality data, it doesn't take away from the fact that it gave me a new mental lens to look at things and understand people's behaviour.
Somehow the idea of perpetually paying property taxes and land value taxes doesn't sound appealing to me, especially since businesses already pay taxes. I don't understand the argument of designing a system to hurt a specific business type such as low value businesses. If there's a loophole such as lack of sales tax for car washes, fix that, but let the playing field remain even. If desirable high value businesses aren't able to compete with car washes, isn't that the market doing it's thing? Introducing additional property and land value taxes might discourage low value businesses, but what are the 2nd and 3rd order effects of such a change?
> I don't understand the argument of designing a system to hurt a specific business type such as low value businesses.
The argument is that good business spots are a limited community resource that it makes sense to tax, like radio spectrum. If you can make good use of the space you're taking up, go ahead, but you should compete fairly with other uses of the space. If anything taxing space use makes more sense than taxing profits; a profitable business is probably one that's serving the community well, whereas a business that takes up space and doesn't generate much profit is no good for anyone. From the article:
> “A car wash does not provide a lot of jobs for the community, and they take up a lot of space,” Broska said. “If you want to invest your dollars into a car wash, then God bless you. But at the same time, I’m responsible for 17,500 people and have to be cognizant of their wishes.”
> the largely automated facility wasn’t the best use for a prominent Main Street site
> a profitable business is probably one that's serving the community well
This argument reminds of the argument googlers to explain why placing paid ads ahead of organic results is better for the user: they say thay if someone can pay more for an ad than means that can get more money from the user, and therefore the user likes it more. Lol
No, profit is profit, it doesn't mean anything else.
*In a healthy competitive market with a consumer-base that's well educated on potential consequences of their purchases.
I do agree with the principle that a space-hogging marginally profitable business is detrimental to a community. Just that the opposite is not necessarily true; profitability does not imply beneficiality.
Humans do not fit the model of "rational self-interested agent" commonly applied for economic models. Gambling and addictive substances are two hugely profitable business sectors that would not exist if it we're remotely accurate.
I'll also preempt someone's inevitable assertion that the burden of verification should lie on the consumer. In informationally antagonistic environment, it's absurd to expect each individual to individually vet every service and product. That's a phenomenal waste of labor that favors only well-funded organizations practiced in deception. Any rational group would pool resources and have a single org do the research and share it with everyone. Oops we've reinvented a government.
> Humans do not fit the model of "rational self-interested agent" commonly applied for economic models.
They generally do -- the misalignment comes from analyzing people's behavior according to presumptive interests which have been externally attributed to them, instead of observing behavior in order to ascertain what people's interests actually are.
> Gambling and addictive substances are two hugely profitable business sectors that would not exist if it we're remotely accurate.
No, gambling and addictive substances exist because people enjoy them. Large numbers of people exhibit a manifest preference for short-term pleasure over long-term stability; expecting such people to act in ways that pursue long-term stability over short-term pleasure is itself irrational.
> I'll also preempt someone's inevitable assertion that the burden of verification should lie on the consumer. In informationally antagonistic environment, it's absurd to expect each individual to individually vet every service and product.
Unfortunately, your attempt at preemption has failed. Only the consumer has the relevant criteria necessary to determine how well a given good or service fits his own particular needs or desires. Being rational, most other people intuitively use the experiences and advice of others as Bayesian indicators of product suitability or unsuitability (even if they don't know what Bayesian indicators are), but they're still using those external resources as tools with which to make their own decisions.
> Any rational group would pool resources and have a single org do the research and share it with everyone. Oops we've reinvented a government.
No, you've reinvented Consumer Reports. Except for the "single org" part, anyway -- there's no single determination that could be applicable to all people all the time, so people will naturally develop a variety of parallel solutions that apply different criteria to the evaluation process.
I take it you view drug / gambling addicts not as people with a mental health issue making irrational decisions, but rather fully rational people that prefer "short-term pleasure over long-term stability"?
Absolutely. People make rational decisions to fulfill the motivations they actually have. But sometimes people, being complex creatures, have multiple conflicting motivations, where fulfilling one impedes another, which leads to psychological and emotional distress. So mental health does come into it, but as a matter of reconciling conflicting parts of ones own psyche, not as a matter of overcoming irrationality.
Or, to put it another way, the irrationality is a matter of having contradictory desires in the first place; choosing to act upon one and dismiss the other resolves the irrationality. The fact that some people make the trade-off in the opposite direction that you would doesn't make them irrational, it just demonstrates that people are different.
I disagree... the diff lies in the definition of "humans" vs. "group". It's like the quote in the movie MIB "Kay : A person is smart. People are dumb, panicky dangerous animals and you know it"
No, it does not. That the users end up paying more in no way means or should be implied that they _choose_ to pay more.
If cheaper options are made less accessible or clearer, and customers are intentionally mislead to more expensive products, as a result they will pay more too.
Also, there is a cost associated with searching. Consumers may intentionally forego the effort for perceived low marginal gains (especially in nominal rather than percentage terms, e.g. "I'm not going to waste my time to save a quarter." even if the quarter is a significant percentage difference). This is one of the factors in the success of Amazon. People "value" convenience.
But people will not pay more than a product is worth to them. The fact that they are willing to purchase the product at a higher price point indeed does imply that that price point is still lower than the consumption utility of the product for them.
Do you always click in paid ads before the organic results? No? Oh because you personally don't prefer them? Oh you mean they're preferred by the minority that does click ads, over alternative paid ads?
> The argument is that good business spots are a limited community resource that it makes sense to tax, like radio spectrum.
I'm not sure I follow the reasoning here. How does the existence of economic scarcity imply that it makes sense to tax anything?
> If you can make good use of the space you're taking up, go ahead, but you should compete fairly with other uses of the space.
But that's already inherent in the nature of scarcity -- the more demand there is for a scarce resource, the higher the price is. So businesses making use of high-value prime real estate are already paying more for it. The law of supply and demand already does what you are proposing. How does paying additional fees to a separate institution with its own perverse incentives add anything to the equation?
> the largely automated facility wasn’t the best use for a prominent Main Street site
This was the personal opinion of a local bureaucrat who thought his personal opinion should be policy. That's why the company is suing.
> I'm not sure I follow the reasoning here. How does the existence of economic scarcity imply that it makes sense to tax anything?
Land in the right place is not merely economically scarce, it's economic land.
> But that's already inherent in the nature of scarcity -- the more demand there is for a scarce resource, the higher the price is. So businesses making use of high-value prime real estate are already paying more for it. The law of supply and demand already does what you are proposing.
Supply and demand doesn't work for land because there's no new supply. No matter how much the price goes up, people aren't going to make more.
> Land in the right place is not merely economically scarce, it's economic land
I don't see how there's anything particularly special about land that warrants construing it as something fundamentally different from any other scarce resource.
> Supply and demand doesn't work for land because there's no new supply.
I'm not sure that I agree that there is no new supply of land in an economic sense -- developments that increase the productivity of land use are functionally equivalent to those that increase the physical supply -- but regardless, the law of supply and demand operates the same whether the supply on the market is new or old.
“I don't see how there's anything particularly special about land that warrants construing it as something fundamentally different from any other scarce resource.”
I’ve overheard enough conversations between high net worth individuals to say there’s absolutely unique qualities about land as an asset class or they wouldn’t hold so much of it.
read some solid reasons for land focused tax regimes long ago but couldn’t remember deets on why they were compelling so googled helped me remember …
‘Immobility: Land doesn’t move, making it a stable tax base. (Important for local services like deciding whether a town can afford the new school(s) or a road).
Scarcity: No more land is being created, so taxing it efficiently is crucial. (Kind of like capital itself)
Non-distortionary: Taxing unimproved land doesn’t distort transactions.
Local Funding: LVT may be an effective way to fund local government since land cannot be moved to avoid taxes’ (1)
(1) which BTW is the main reason often given why ordinary folks have to pay higher payroll taxes vs capital gains the wealthier brackets ‘pay’ because you know capital will just up and move somewhere else if we ask too much of them.
Well just tax the land, and if the capital holders want to move all their wealth out of the community, fine, but good luck extracting wealth from a local community without owning any land near it.
So yeah. Tax the damn land already. Especially to fund local government and services, which you kinda need to have a functioning society.
I clicked through because I’ve wondered often why there’s a dozen of these new washes in my community with ten more on the way and figured it was some financing/tax write off hack.
> I’ve overheard enough conversations between high net worth individuals to say there’s absolutely unique qualities about land as an asset class or they wouldn’t hold so much of it.
Every asset class has its own uniquely defining characteristics in financial terms. There are unique qualities of stocks, of bonds, of index funds, etc. But I don't see any fundamental difference between land and any other scarce resources in real economic terms, and definitely nothing sufficient to treat land differently as a fundamental philosophical principle!
> Immobility: Land doesn’t move, making it a stable tax base.
Lots of things don't move in any meaningful way with respect to taxing jurisdictions.
> Scarcity: No more land is being created, so taxing it efficiently is crucial.
I disagree that no more land is being created in an economic sense. As I mentioned above, developments that increase efficiency of particular use cases for land are economically equivalent to the supply of land expanding.
And there are even some cases where new land is even being created in a literal physical sense -- see the Netherlands, for example.
> Non-distortionary: Taxing unimproved land doesn’t distort transactions.
It certainly does distort transactions for unimproved land. And what is "unimproved land" in the first place? Whose definitions apply, and how do we handle edge cases. If I purchase a plot of land for the specific purpose of maintaining it in its natural state as a preserve, is it improved or unimproved?
> Local Funding: LVT may be an effective way to fund local government since land cannot be moved to avoid taxes
But there are lots of other ways of avoiding taxes. Still, this purely pragmatic point -- which correctly understands taxation as a means to fund the necessary operations of government, and not a tool to manipulate behavior or as an end in itself -- explains why a large portion of local government funding already comes from property taxes.
I'm not sure what trying to take what already works and reconstitute it according to Georgist principles (which are logically weak and entail a lot of dangerous implications) brings to the table.
> So yeah. Tax the damn land already. Especially to fund local government and services, which you kinda need to have a functioning society.
Property is already widely taxed, local governments are already funded, and society is already functioning, car washes and all.
> The argument is that good business spots are a limited community resource that it makes sense to tax, like radio spectrum. If you can make good use of the space you're taking up, go ahead, but you should compete fairly with other uses of the space.
If it's really such a prime desirable spot then that would drive up the value of the land to the point that the low value business wouldn't have been able to afford it?
I don't understand saying it's such valuable land and detaching that from what a business was actually able to pay for it.
> If it's really such a prime desirable spot then that would drive up the value of the land to the point that the low value business wouldn't have been able to afford it?
The land continues to appreciate, so the business can hold it based on its value. Especially if they've locked in a mortgage at a low rate. The car wash "business" can profit through land appreciation, which they don't pay any tax on (another flaw in the regulatory regime), and free-ride while the spot becomes more valuable through the efforts of others.
You can't just assert that it makes sense when answering why it makes sense. If you are right, then the owner is losing out on money by operating a car wash, which by the way is their moral and legal right. If you think you can provide more value with the same space – offer to buy it.
> a profitable business is probably one that's serving the community well, whereas a business that takes up space and doesn't generate much profit is no good for anyone.
I think that attributes a lot more to profitability than such a metric deserves. Perhaps in a more narrow and ruthlessly capitalistic sense profitability signifies that a business is doing what it's trying to do well, but it's a big leap to get from there to how well its serving whatever is considered to be a community these days. Among other problems, just because some tiny amount of that money conceivably stays around in the region and people can buy stuff does not mean that the cost of the business being there isn't quite a lot higher, hence the term tragedy of the commons and the hollowing out of let's say America
The theory of (LVT) land value tax is that it replaces other taxes. LVT has less or no dead weight loss so it's a more efficient tax.
> If there's a loophole such as lack of sales tax for car washes, fix that, but let the playing field remain even.
My claim is that the playing field is not currently even rather it is massively in favor for low-capital, low-labor, and low-regulatory businesses (like car washes) and additionally incentives ostensibly designed to encourage real estate development (1031 like-kind, treatment of real estate as depreciating, etc.) are now primarily used to either speculate on existing real estate or build the minimum to gain ownership/interest in speculation. If you take all the cash you have, you can only buy a finite amount of land. If you build a low-capital but profitable business like a car wash, you are only limited by the leverage limits imposed by lenders.
> If desirable high value businesses aren't able to compete with car washes, isn't that the market doing it's thing?
> but what are the 2nd and 3rd order effects of such a change?
Tautology yes, as desired no. Technically yes because the market is shifting towards low-capital and low-regulatory businesses because they have a more predictable ROI. The goal isn't to disfavor the more capital & regular intensive businesses just regulate them. i.e the influx of car washes is the undesired 2nd order effect of some other policy e.g. minimum parking for restaurants and apartments (likely no such rule exists for car washes so now you need less land a car wash vs a restaurant).
Raising regular property taxes (land+improvements) is just easier solution than waiting for far-reaching tax reform like LVT. IMO it's better to correct the market even if it means raising taxes overall in the short term.
I like Georgism but if business taxes were replaced would internet businesses that don’t need a physical location thereby pay much less taxes than those that do have need for physical location? That does sound kinda lopsided, unless we’re also doing a land value tax on prime domain name real estate.
Internet businesses have physical locations somewhere. Even if they are drop shipping than someone else is paying the land value tax for the warehouses they are shipping from. The land use exists somewhere at some level of the value chain and that somewhere would then be taxed.
Not all, for example my SaaS company is fully remote and has no physical location whatsoever. I am cool with paying less taxes and I understand that the purpose of the land value tax is more about utilizing a limited resource effectively and not necessarily about some notion of fairness, but it seems weird that someone like me would not need to buy in besides my own home because of the industry that I’m in.
Then the server farm where your code runs will be taxed. Maybe you run it from your house, in that case then it is part of that.
You are still just moving it around at the end of the day. Presumably as you grow more profitable and wealthy you would relocate to higher value real estate. If not then you are spreading the wealth to lower value areas which is also a good thing. Seems like all good things in the end.
> My claim is that the playing field is not currently even rather it is massively in favor for low-capital, low-labor, and low-regulatory businesses (like car washes)
>Somehow the idea of perpetually paying property taxes and land value taxes doesn't sound appealing to me
Eternal wealth to those lucky enough to have been born and bought in the past or born to families who bought in the past sounds plenty dystopian to me; pay up regularly or let someone else who will contribute to society step in.
It's less about land availability. There is plenty of land at least in some countries. It's more about desirability. As those in the real estate trade like to say it's all about "location, location, location."
And for that theoretical tiny category there are tax deductions and other exceptions to the law (for instance in many jurisdictions sites of religious worship are exempt from property tax). The overwhelming majority of owners do not and should not qualify for this.
What eternal wealth would a guy in a hut in the middle of a land-locked forest have, with no infrastructure, and no community? Why is he forced to provide anything beyond self-sustainability?
> pay up regularly or let someone else who will contribute to society [raze your forest to sell wood chips].
Forcing mountain men off their land to work in the factories, sounds plenty dystopian to me. The Industrial Revolution and its consequences have been a disaster for the human race…
> Somehow the idea of perpetually paying property taxes and land value taxes doesn't sound appealing to me, especially since businesses already pay taxes.
And it doesn't sound appealing to me, as an individual. I don't want to feel like a peasant constantly paying a tax to the monarch/state: so at very least the first property an individual owns should be tax free (not annual land value tax / property tax). I happen to live in a country where that's the case (but it's not the reason I moved there): no yearly property tax.
The days of merely protecting one's property are long over. At least in the US real estate taxes pay for many services unrelated to safety of the property. For example, closing a real estate tax funded senior center is unlikely to result in an increase in crime. Maybe the seniors themselves would be more vulnerable without a senior center, but I fail to see a general crime spree resulting. If anything the seniors would be watching other people's property and call the cops at any sign of potentially suspicious activity.
I live in a US state where I do pay property taxes but there is no income tax. Those property taxes and the overall sales tax are the main sources of revenue for the state.
I don't expect to live in an area without paying something to maintain it.
I don't understand myself how LVT wouldn't result in what you see in many "high value" city centers (where business rents are insanely high) - miles and miles of law offices and banks, and not much else.
Taking up land is an externality, it’s not something that can be solved with a supply/demand curve. So, we tax it to encourage efficiency since our usual method if encouraging efficiency doesn’t work.
> If there's a loophole such as lack of sales tax for car washes, fix that, but let the playing field remain even.
What is even the problem here that needs to be fixed? People have found a business model that is able to generate value from low-cost land that might otherwise lie vacant. Great, more power to them.
The idea that people's use of their own property should restricted or manipulated so the government can maximize tax revenue is the epitome of the tail wagging the dog.
Do you have a specific tool selected at the top? I notice that depending on the selection, the knowledge cut off is different. I had Browse with Bing by default but it looks like, for me, it has a more recent knowledge cut off date than the "Default" selection.
Click on 'Play moves' to watch a replay.
I initially planned to run a chess tournament for LLMs but they are not good: besides obvious mistakes, they output incorrect moves, get stuck in loops by repeating the same moves and the smaller models fail to output valid json frequently. I thought the reasoning models like o3 mini might be good, but they are an incremental improvement in chess.
Feedback and suggestions for other games to explore welcome.