I have to wonder how far behind the news this AI is going to be, or if they have somehow fundamentally changed things. Current AI's spend a lot of time being trained on a static corpus of knowledge, and then only have a relatively small context window of extra information that they can be given to respond with new information (e.g.: your recent chat history).

But stock prices are fluctuating all the time, and often enough the recent fluctuations are more important to the future of the stock than previous ones. So how often are they going to re-train the AI in order to keep up with changes in the market. And I have to assume the this AI will only be trained with the numbers (stock prices, EPS, etc), and will not be watching news in order to respond to that.

I am curious to see how they have solved those issues, or if this will only work in a more general sense of information a fixed amount of time in the past.

Given that China doesn't allow everyday citizens unlimited access to the internet, we can assume the only ones allowed out are bad actors like badbot, so blocking China entirely would be a net benefit for the entire world. We'd have to get the VPN operators to cooperate, which is near impossible since they'd sell their own mothers for a quick buck.

Because I've tried using LLMs to generate code and I've seen the results. They are not usable. They *resemble* valid code, but they typically throw exceptions and raise errors, they can't pass unit tests, and they don't correctly handle edge cases. AI-generated code is a mess that *superficially looks right* but isn't fit to purpose.

There is a meme going around about the fact that you can tackle a normal coding task by spending 3 hours to write code and 1 hour to debug and test it, or you can use CoPilot to spend 15 minutes to write the code and 8 hours to debug and test it. That matches my experience.

A lot of what I write only gets in commands through HTTP, and all the resources it works with are through S3 or database connections. In some cases they use local disks, but only as a cache to deal with files that are larger than memory (less common for me right now).

So as long as network connections are possible, most of my work would be better done in systems like this.

I know someone who worked at a startup that was trying to do exactly this: watch as workers put fast food into the bags in order to catch mistakes on orders. The startup had some problems and laid people off, but they were actually starting to work with a chain on testing the in-develoment project.

The whole point was to catch mis-picked-orders before they went out the door, so that customers (primarily at the drive-though) would not be angry at the company (costing either future business, or expensive soothing of the customers). It seemed like a decent idea, but one the company was having trouble make work seamlessly.

Given that machine learning has proven to have virtually no moats around business models, even if this is not the same company, it is likely to get to the same point as that one. To me this seems like a solveable problem, if one that it is going to take a bit of pain to get to the point it is really useful.

"Open Server" might be a great distributed idea. We need a piece of software that individuals and companies could host and run. This software would have a configurable list of open source projects the host company wishes to support. When the software is running it would join the distribution network for those listed projects. This would be like an Open CDN sort of thing. Lots of details would have to be worked out around security and validating downloaded files and stuff. But the idea sounds kind of cool.

I think you have blinders on here. An easy example would be the Chinese high-speed rail system, or the Japanese system, or even the rail systems found in much of Europe. While not all of the U.S. is suitable for this (e.g.: the plains states), the eastern seaboard would benefit enormously. But we can't get our act together and have a pitiful system. And the Chineses have created a comprehensive high-speed rail system covering an area roughly the size of the U.S. (including Alaska). Amtrack's coverage of the U.S. is pitiful, and slow everywhere.

As far as I can tell the U.S. buys all of its trains from either Japan or Europe, so we don't even have a hand in advancing the state-of-art.

In another area, the U.S. has not led in semiconductor production for most of my life. The best fabrication machines come out of ASML in the Netherlands, and the rest of the process is much more advanced in Taiwan. Right now the U.S. is pouring taxpayer dollars into this area to try and create plants that will be a generation behind when they come online, and we are paying the Taiwanese to build them (and staff them, largely). The only place we lead in that area is in the CPU design space... but we can't build them.

We need to stop resting on our laurels and pretending we can catch up whenever we want to. I am not sure that Thorium reactors are where we need to concentrate, but this attitude has to go.

Apple's FaceID has always used an IR dots system to get a 3d map of the face. So a flat picture has never worked to fool it.

There are a number of implementations on Windows and Android that use a standard or IR flat-camera with no dots system, so those could be fooled by flat pictures.

So I think you probably saw one of those other systems, not an iPhone an FaceID.

I don't understand why you are surprised by this. The NoGIL version is expected to be slower, as is Golang.

NoGIL will probably never be faster than "regular" Python on most code. It is only when you have a lot of truly highly parallel (i.e.: little interaction between threads) that anyone is expecting an eventually speedup. The version that exists in the Python main-line now is mostly there to allow for the gradual shifting of some fundamental aspects of Python to allow for the eventual speed-up in that highly parallel code.

And Python is never going to be as fast as Golang, for many of the same reasons that Rust will probably always be 30% or so faster than Golang: they are engineered for different purposes. Python is a dynamic language that is quick to write small-to-medium projects. Golang is great at making highly parallel things (like web-apps), and Rust is for when you really need things to go fast (and are willing to spend the extra engineering time to make it work).

Sure, people use the languages for other purposes. And sometimes you use them together (e.g.: Python as the glue, and C or Rust extension for the hot-loop). And there are pleanty of times where people use the "wrong" language for a project (either because it is the hammer already in your hand, or because the scope started small and crept to be big... see Facebook and PHP as an example of the latter).

This is a revenue tax, not a profit tax.

Lyft takes the credit card payments, and then pays the drivers. The amount they pay the drivers for that transaction sometimes exceeds 100% of that payment.

Lyft is arguing that their revenue should not be the amount of monty they take in, but that the payments to drivers should be removed from that, and the drivers should be charged the taxes for that (but since no driver makes more than the $2 million in excepted revenue, there would be none).

But if you work this same thing slightly differently: if it were a store that was selling widgets, it would be hard to argue that the money it uses to buys the widgets should be removed from their revenue. The store has the revenue from those sales, and the widget manufacture also has revenue ultimately from those sales.

Argue all you want against revenue taxes, but I don't think Lift has a leg to stand on here.

I agree with almost everything you say, but I think you are missing one important note that your opponents will bring up: there are many primarily economic migrants who are coached to apply for asylum. There are many people abusing this system.

Of course, the solution to that problem is exactly the one you have already outlined: speed up the system to quickly and decisively make decisions. Unfortunately, the Republican establishment (now those surrounding Trump) have decided that there are a couple of easy ways of doing this:

• Make the process so terrifying and horrible for people that they are scared to try it
• Make the bar for asylum entry so high that no one can clear it

As an example of the first see the family separation policy. That was entirely about scaring families so that they would not make the trip. Of course that logic is never going to work for most of the people making these harrowing journeys to the U.S. souther boarder: their lives in their home countries were spot horrible that we are never going to match that.

I think you are missing the whole point with the "people who denied biology" part. People who are transgendered are not denying biology, rather they are resisting the roles that people fit into in society based on that biology.

And even that is a bit over-blown. In most cases they are not resisting those roles (arguably non-Binary people are), but rather declaring that what is between their legs should not define their role in society.

Do you actually disagree with that? Do you think that the particulars of your birth should determine your role in society? We have, as a society, already declared that we don't like nobility-by-birth, or skin color, to be the determining factor (though wealth of your parents is still a factor). And we have even declared in our laws that your national origin is off-limits there (unless you are a non-citizen... related, but different topic).

But gender, and biological sex, is the issue we started grappling with in the 1960s, and are still working our way though (not that we have skin color completely done either).

This is tough for some people to get through. They are used to biological sex determining gender (full stop), and the idea that this does not apply to everyone is tough for some. The Democratic party has decided that they will take a stand on this, like they did on Civil Rights, like they did on Gay Rights. To me it is clear that they are on the right/moral side of all of these issues.

Democrats are happy with white christian men, they are just not pandering to their fears of no longer being the group solely on top.

And to be clear: I am a straight, white man married to a white woman, with kids. I don't have a dog in any of these fights directly. But I can see the moral right of all of them, and support the fight.

I know...former smokers are the least charitable critics of current smokers. But still, I feel real pity for anyone still sucking at Meta's sour teat. The algorithms are going to learn what tickles their limbic system and feed them that sour content like the proverbial rats hitting the heroin button. Please just leave the platform now — life is better without it.

I won't defend the number. I agree it is truly, mind-boggling obscene. But this is an oversimplification. It is very likely this needs to interact with a bunch of other byzantine federal systems like the IRS and Social Security. This means the security governance and interagency agreements of the whole project will be unbelievably difficult and full of bureaucracy.

The Physicist one needs to begin: "If you assume it is a sphere..."