rodonn
rodonn on Hacker News
You can't easily evaluate the success of any single forecast, but as soon as you have a collection of forecasts there are a lot of ways to evaluate which forecasters are more accurate and which are well calibrated in their predictions.
For example you can look at the calibration of the forecasts: Take all the outcomes where you estimated ~10% probability. How many of them actually happened? If you are an well-calibrated forecaster, the actual number should be fairly close to 10%.
Brier Score:
I agree that it won't be an unbiased source of information, but should be roughly same level of trustworthiness as clicking on the link and reading the information on the advertiser's website.
I totally agree about the subscription being worth it for most heavy users, but there are a lot of people in the developing world for whom $20/month is hard to stomach. I think it's good that those people have an option to get more access to AI with lower out of pocket price.
Yeah, basically this. Some people really do click on ads that are relevant to their interests and might want to learn more or be able to ask questions about the product. I'm not confident about how well it will work in practice, but I could imagine cases where this would be more helpful than clicking on the ad link and needing to explore the resulting website to find the information I want.
The challenge is timing. Power is a relatively low share of the cost of a modern data center, but the length of time it takes for approvals and construction for solar means it would substantially slow down the construction vs. using grid+natural gas turbines which are faster to build and get approval for.
I'm 100% on board with charging data centers more for their power usage. They can afford to pay the extra costs. They would much rather pay in dollars vs. in time.
The problem is that I don't know before I read a doc whether or not it will be useful and valuable.
If someone wants me to spend my time and attention on something they have shared, I would like them to demonstrate that they put a proportionate amount of time and effort into its production.
"profitable on inference" means "marginal costs of inference are lower than revenue". It is very common to distinguish between upfront costs vs. marginal costs when judging the economic viability of a business.
You mention "debt servicing", but OpenAI has no debt. All the money they have raised is equity not debt.
There's been a huge amount of improvement in coding agent effectiveness since they ran that experiment. In a more recent follow up experiment, METR found 20% speed up from AI assistance and says they believe that is likely an underestimate of the impact.
https://metr.org/blog/2026-02-24-uplift-update/They are working on making a new measurement approach that will be more accurate.
You can use the "cameo" feature only with users who have gone through the cameo creation flow. Sama has an account and created a cameo likeness of himself.
When you create your cameo you can choose who is allowed to make videos using it: "only me", "people I approve", "mutuals", or "everyone".
That's hiring for a role buying ads, not selling ads. i.e. this is ChatGPT buying ads on other platforms to promote ChatGPT.
The Clayton act explicitly includes partial acquisition as still being covered. "No person engaged in commerce or in any activity affecting commerce shall acquire, directly or indirectly, the whole or any part of the stock... [where] the effect of such acquisition may be substantially to lessen competition, or to tend to create a monopoly."
There may be some other regulations that are avoided by a partial acquisition, but it doesn't bring it wholly outside of the relevant antitrust laws.
It might reduce scrutiny, but not completely prevent it.
Clayton act says
"No person engaged in commerce or in any activity affecting commerce shall acquire, directly or indirectly, the whole or any part of the stock or other share capital and no person subject to the jurisdiction of the Federal Trade Commission shall acquire the whole or any part of the assets of another person engaged also in commerce or in any activity affecting commerce, where in any line of commerce or in any activity affecting commerce in any section of the country, the effect of such acquisition may be substantially to lessen competition, or to tend to create a monopoly."
Depends on where you live. The average in San Francisco is $0.29 per kWh.
These drugs (mostly) don't allow you to eat more unhealthy food, instead they make it easier to have the self control to avoid over eating / choose healthier foods.
The profit maximizing price after a 30% cut `q(price) * (price - 30%
price)` is not the same as the profit maximizing price without that cut `q(price) price`.
As long as the demand curve is downward sloping, there will be some pass through of Apple's cut to customers, though the fraction that is passed through will depend on price elasticity.
This is why targeted advertising is so much more effective than un-targeted. On Facebook/Instagram I frequently see ads for products I was not aware of that are relevant to my interests.
I haven't run into any friction for (1), since worst case you just call `.to_pandas()` at the end of your pipeline before you start plotting. For any plotting apis that rely on the direct column vectors, no conversion to pandas is required.
For (2) I like to do:
1. `from polars import col` which at least shortens each use by a few characters
2. For columns that I use very frequently, define a "constant" for them e.g. ID=col("id").
This lets you do df.groupby(ID) instead of df.groupby(pl.col("ID")).
Another advantage of defining these column "constants" is that it makes it much easier to refactor to rename all usages of the column (without needing to check whether each string "id" is being used as a column name vs. something else)
You don't 'need' tracking for ads, but it has a huge impact on the the amount of revenue per impression the ads will generate, because tracking enables you to show ads to the people those ads are relevant to.
There are a lot of business models that are feasible with tracking for ads, but are not feasible to offer for free without tracking.
If you had a system that was (a) oriented towards achieving some goal and (b) understood the world it was operating in, then allowing someone to "pull the plug" would interfere with it achieving the goal it was trying to optimize towards.
There has been some attempts at researching the question of how to design a intelligent system that is "corrigible" = willing to allow humans to change the goal it is set to optimize. This is unfortunately still an open question where no great solutions have been found that seem to be reliable when faced with a highly intelligent and capable AI system.
If you are interested in reading more, a few relevant search terms are "Off-switch game" and "corrigibility".
They key is that it is a tax on the value of the land not the size, so large plots of land would face a much smaller tax per acre compared to prime real estate in a city.
It is true that a LVT will have a larger percentage impact on the market value of undeveloped land who bought land for speculation without any investments in making the land more valuable. This will especially affect people who bought up lots of land speculating on the value going up. This includes people who bought up lots of land for farm/lumber/mining, but also people who buy up urban plots and keep them as a flat land parking lot while waiting for the value to go up.
Any change in taxes will have some winners and some losers. For example an increase in traditional property taxes hurts anyone who bought expensive houses. An sales tax/consumption taxes hurts anyone who saved up financial assets. An increase in income taxes hurts people who invested in education to increase their labor value.
The question is which is the least unfair way to tax to meet the needs of society. Most of those other taxes will actively discourage valuable investments that make society as a whole more prosperous. Property taxes discourage building. Wealth taxes discourage saving. Income taxes discourage earning and investment towards future earning. In contrast a tax on land value, doesn't cause there to be any less land. It does discourage speculation in future increases in land value, but that speculation is mostly zero sum (it might make the individual speculator richer, but doesn't increase the wealth of society as a whole).
I don't understand how you think that building up defensive capabilities is provoking war. Making the cost of invading higher makes it less likely the PRC will try to invade. The US's long standing goal has been to preserve peace in the Taiwan strait by making the cost of attempting to invade sufficiently high to deter PRC from making the attempt at invading
https://www.defense.gov/News/News-Stories/Article/Article/35...
I'm pretty sure the author is misinterpreting what she saw.
Here is a thing that Google does do:
The user searches for "children's clothing".
An advertiser has created a search ad for the keyword "NIKOLAI-brand kidswear". If the advertiser has enabled "broad match" for their ad, then their ad will be eligible to be shown on the user's page in one of the advertising slots, even though there are no words in common between the user's query and the advertiser's keyword.
Does anyone have evidence of Google manipulating the user's query to affect the organic search results shown to the user?
GPT stands for Generative
Pre-trained Transformer.
The "main" training step using huge amounts of inputs is called pre-training. The idea is that after that pre-training, you might fine tune the model for your specific use case.