Prerequisites
This book assumes that you have some experience programming in Python
and are familiar with the fundamentals of machine learning. The focus will
be on building a strong intuition rather than deriving mathematical
equations. As such, illustrations combined with hands-on examples will
drive the examples and learning through this book. This book assumes no
prior knowledge of popular deep learning frameworks such as PyTorch or
TensorFlow nor any prior knowledge of generative modeling.
If you are not familiar with Python, a great place to start is Learn Python,
where you will find many tutorials on the basics of the language. To further
ease the learning process, we made all the code available on Google Colab,
a platform where you can run all of the code without the need to install
anything locally.
Besides ethernet cables providing more stability to your network connection than WiFi, it isn't mentioned in the article. Not many people know about this: most houses have coaxial cable lined throughout. If your router is located in a different location in the house, you can use the coaxial cable to help create a network. You need two MoCA adapters (coaxial to ethernet) on both sides (one at the router/switch/gateway and another at your computer). Connect it all up… bingo, bango, bongo, and you got stable and fast internet! This is how I have the most stable connection in my house :)
Also remember that even if you can't or don't want to connect your device via Ethernet, you might be able to use MoCA to easily move your WiFi router to a better location, or add a second router/mesh node with a reliable wired backhaul instead of a crappy radio backhaul, and that might make all the difference.
I'm just building a new house this year and Ethernet to every room is standard in this area. In a different state in the US coax is standard in new construction. (We just moved) It varies by location.
I've lived in both Australia and the UK, and the only coaxial cables in any home I had/have have been direct point-to-point connections from the roof-mounted TV antenna to the corner of the sitting room where the TV was intended to be placed. Not terribly useful for computer networking.
Cable TV (from cables under the road) is much more common in Denmark than in the UK.
Digital, broadcast TV in the UK is mostly received by radio broadcast or satellite broadcast, so the house would only have the appropriate cables from certain rooms up to the roof. Thinking of the houses I lived in, it wouldn't have been practical to put a router of any kind up there.
From recent UK statistics:
- 95% of households have at least one television
- 30% receive TV over satellite
- 13% by cable
- 80% by IPTV
- 48% by terrestrial radio broadcast
- 96% have broadband internet
That's 13% of houses using their cables for a TV subscription, so more than that number have the cables available, but some big chunk probably had it installed to a single room in the 1980s and have never expanded it — the kitchen and the kids' room use terrestrial broadcast.
Nope. As you say yourself: Possibly an antenna connection.
Exactly that. When we redid the floors i our apartment we renewed the electrical wiring and added ethernet and coax. The coax was intended for radio/tv.
There is however an overlap of people who have the foresight to add coax cabling also add ethernet a well.
If I was was re-doing today I would add more empty pipes and fiber.
Cabling ahead has a very high WAF (Wife Approval Factor) when adding new gear :-)
Most unreliable network I ever had the misfortune to use (not to install) was coax thin ethernet when I worked at the BBC - termination problems were terrible. Twisted pair for the win as far as I'm concerned or better, WiFi. I don't think I have any coax in my fairly modern UK house.
You are talking about two cometelt different technologies. You are talking about 10BASE2 Ethernet which runs over RG-58/U coax cabling and as you mention requires termination at each end. It ran at a speed of 10Mbs. It could be unreliable for a variety of reasons such as connections coming lose, someone deciding to move their computer disconnecting a cable and breaking the continuos connection between stations that is required or a large number of Ethernet collisions because if either misbehaving nics or too many stations on a segment.
Currently there is MOCA hardware which supports speeds up to 2.5Gbs. The standard for 10Gbs has been released but no hardware for it is currently available. At least not to consumers. MOCA runs over the coax that is often already installed in homes to support cable, satellite or over the air antenna TV. It uses different frequencies and thus can coexist with these on the same cable. MOCA is not Ethernet. It is a half duplex shared medium protocol using time division multiplexing. It was originally developed to distribute IP TV without the need to run additional wiring in a house. Today it is mostly used to bring broadband internet connections into a home or to bridge Ethernet connections through a home. Different frequencies are set aside for each purpose and so both can be done at the same time. It is very reliable. I use it to extend my network to several out buildings on my property which had coax run to them many many years ago.
Thanks for that - very informative. I must admit I haven't done anything with wired networks for many years! But I still like twisted pair, and wifi :-)
Depending on how modern it is, you might well have "telephone extensions". In principle this is Cat 3 cable strung in a tree shape, but in practice the electrician is probably buying wholesale cable - carrying one reel of Cat 5 for all jobs is easier than owning a Cat 3 reel and a Cat 5 reel and bringing the right one for each job, the price is usually either identical or within pennies. So there's an excellent chance it's Cat 5 cable anyway†.
Now, Cat 5 cable is a perfectly good telephone cable, but it's also Gigabit Ethernet (over reasonable distances, you don't live in a mansion). The tree shape won't work for networking, but the individual cables buried in walls or elsewhere are basically just right there already. You just hook the existing cables to new Ethernet shaped faceplates. I am literally writing this from a wired connection in a bedroom, nobody built this to have Ethernet, they built it to put a phone in the main bedroom, but it's 2025, nobody owns a wired phone, everybody needs Internet.
† Also the network cards can't tell, they will try to achieve 1000 Mbit/s and chances are they succeed even if the cable isn't actually rated Category 5. I have retro-fitted modern switches to an ancient building (the old Mountbatten chip fab at the University of Southampton, before it burned down) and in 90+% of cases this "upgrades" the connection to Gigabit because the Cat 3 cable pulled a decade or more earlier was good enough.
Coax ethernet was often deployed as a bus: 20something computers that must be directly connected to a single cable going from room to room. If one of the plugs was unplugged or not tightened correctly, everyone on that line suffered. This happened a lot, people moved computers around. And the tech looks simple.
That is a different situation than installing a fixed, point to point connection.
If you do this doing forget to add a moca filter where your cable enters your house, otherwise you will leak your Ethernet signal to anyone who is on the same cable.
Funny, besides preventing eye strain, I thought it saves battery on my iPhone 13 mini, which uses an XDR OLED display. They'll have to continue the same experiment with OLED displays like they mentioned
Yeah, this makes "perfect sense." The IRS is the primary federal agency that generates net revenue with a positive return on every dollar invested. When the IRS receives more funding, it typically enhances tax enforcement efforts and increases the total revenue collected. It doesn't seem logical for DOGE to reduce staff when those employees contribute more revenue than they cost, resulting in a positive return on investment.
Unless they believe it's possible to maintain the same revenue collection with fewer people through increased automation, what are their motives? Laying off probationary employees still developing critical skills undercuts the IRS's long-term revenue goals. Increasing the performance standards for probation rather than terminating these new hires would help build a more effective workforce that generates more significant revenue. IMO, Elon and friends are just trying to reduce future tax enforcement and get away with more
This is exactly why it’s stupid to trust Musk & Doge, even if you agree with their approach. It’s a basic fact that it’s a huge conflict of interest. Musk has several companies who stand to benefit from his influence in government, and he can easily personally benefit from changing the IRS like you say.
It’s fundamentally not possible to trust a person in that position — no matter their intentions. (And Musk’s intentions are pretty likely related to the personal power he gets from this situation — why else would he risk jumping on board with a controversial administration?)
It’s not trustworthy on a fundamental level. If you want the government to change like this, you should have voted for people who are cool with transparency and accountability.
Because of how corporations and state are tightly fused in China's governance.
> A Leninist system features an authoritarian regime in which the ruling elite monopolizes political power in the name of a revolutionary ideology through a highly articulated party structure that parallels, penetrates, and dominates the state at all levels and extends to workplaces, residential areas, and local institutions.
Yes. These are not comparable political systems. In the US, the information you share can be accessed by law enforcement with the approval of a judge if there's a crime suspected. But in cases where the government improperly accesses your data, they actually destroy their own case against you, because anything from that poisoned tree of evidence can be thrown out in court. Even when governmental power is abused in the US, it is nothing like the routine surveillance and suppression that chills free thought and speech in a totalitarian dictatorship like China.
I'm sorry, but your idea of how the US works is a complete fairytale. You need to get a serious reality check on how the US actually works in real life. The law in the US is applied selectively (depending on the profiles involved, severity of case, political backdrop, etc). There's plenty of corruption, misaligned incentives, and corporate meddling. I can't count the number of cases from the past 30+ years that demonstrate this.
Also weird how people pretend Snowden wasn't just trying to draw equivalence between the US and the dictatorship where he currently resides, on behalf of said dictatorship.
> "CCP did not invade Iraq, Libya, Afganistan, bomb Syria or support the Palestinian Genocide."
1. There has been no genocide in Palestine.
2. CCP meddles in other countries to equal if not worse degrees - both militarily and politically/economically. Routinely imprisons and erases millions of own citizens. Works to annex territories that aren't part of China (today). Funds and arms Russia, Iran, Syria...
You seem like the kind of person that selectively applies and practices their morals, depending on whether the story aligns with your agenda.
"There's an obvious problem with the concept of training on user prompts; how would training on a bunch of questions cause it to know the answers?"
I imagine by analysing the chat?
If the user says thanks in the end, or gives a thumps up, it likely was a useful and correct answer, that could be included in further training. Or at least considered for future training and I cannot imagine them not considering and experimenting with it.
User queries were at least historically useful to train smaller models from larger models. You need to know the kind of questions real people ask to train a model that’s good at answering those questions
Back when I started using LLMs for writing code I would type out long, gently phrased explanations about why it was wrong, as if I was teaching a pupil, hoping it would help. I'm sure a lot of us did. If they can parse and mine those prompts, they'll have a nice little metacorpus to build on.
Now I just tell it to stop being stupid over and over until it does a good job. I wonder if it would improve the model to keep all of the beratement in the training data.
Edit: Apparently a 'metacorpus' is a swollen nematode ass. My sincerest apologies, bros.
The bigger question is what ELSE are Anthropic/OpenAI/et al. doing with your data? Training is just one of many ways to exploit users’ data. Some of the other possibilities are truly chilling.
This is actually dumb. I go by he/him pronouns and I put them in my signature, because my first name is gender-neutral. I find it annoying when people assume I'm female over email correspondence. Not considering people with different genders or who have a desire to use different pronouns, I'm sure there are other people like me in the federal government. What's next? They'll tell people they need to change their name to match their assigned birth gender?
A chunk of the federal employee base will have something like "LTC" or "Dr." in front of their name, which again confers no indication of gender or preferred pronoun.
I don’t understand why you’re going through all these weird lengths to solve a problem that pronouns more optimally solve and cover more use cases (e.g. nonbinary people).
interesting. this might be able to be exploited by those who wish to convey their pronouns without breaking this new pronouncement, at least for those who have chosen one of the traditional binaries.
Yeah, I mean, it's not going be a one size fits all solution here though. Some people get really finicky with titles. And with about 2.3 million employees, simple rules aren't going to work, there's too may execptions.
Just with the simple trans stuff, estimates are about 0.5% of people fit that description. So about 11,500 employees. Not all of them are fully out, so you're looking at a lot of people that don't fit that bill.
Some people really do not want to be a Mr. or a Ms./Mrs. I think it goes back to bad childhoods.
Some people really do not think that they are a Mr./Mrs. anymore, that they are Dr. or Col. or Rev. now and just reject the Mr./Mz. out of hand.
I have older people in my life that are super particular about the Mrs. thing and just use Mz. and always have.
Also, you still have a lot of women, especially older and in the south, that will take their husband's name as their formal title (Mrs. Dr. John Q. Doe). And those that I've met that do this are very particular about it.
Again, there's a lot of people here and I think leaving it up to the particular person on the other end of the conversation is the only workable method. It's a 2-way street afterall and you have to respect that other human on the other end.
In the very least, you've gotten a lot of info about them and their personality that you can then use to your advantage.
This is what my colleagues in Vietnam have been doing for decades (well, the two decades I've been doing business there).
Vietnamese names are gendered but Westerners have no idea which genders go with which names. (Is Duy male or female? How about Duyen?) So, at least in my experience, they've always just put Mr/Ms in their email so people know which it is.
