This is an infrastructure design and monitoring mistake. Can't be traced to just Microsoft, it is a very simple way to avoid responsibilities.

An institution as important as a hospital, must have a partitioned network, with security levels that avoid a mistake in a zone to destroy the work of another one. And the data and related processes must be organized accordingly with their importance and sensitivity.

Also, good and enough contingency plans must exist.

Everything else is to play as children with sand.

We can't do the things "just because" we can.

All these almighty tools that use stolen information from experts around the world, must be directed to resolve the unknown, not to replace the stablished solutions, destroying in their path the life of people that have been improving their craft on a lifetime work. And here is the big question ... why I do need to do that?

Anyway, when a "current" AI solution is created, we know nothing about it. We have no idea how it works inside and how could evolve in the future. There is no easy method to tune it in very particular directions, neither or to transform it to be perfectly adapted to the daily nuances of life. The usual methodology here is to replace it with a new solution, but following this direction we have no idea if the already resolved characteristics will work the same with the new version. Humanity behavior is not so easy to track.

The answer is not to replace but to complement. It is not to need less engineers, but to help them to work better products, to advance faster but improving the life of more people. When what we think is only about how our product can do what others already do, we destroy their lifes and trash the marvels that this new technology could be working on instead.

And all this look as a huge applied EGO, making the result even more bizarre. Really, why do we need to follow this path?

I really don't know why people are paying that level of attention to a person that believes he has all the keys of future.

OpenAI has done some impressive things but at an alarming cost, but monetary and social. And he not only describes the end of the current world, but also he is asking to break any rule that is in front of his company to work as he likes.

Technology is a vital part of humanity, but to advance without taking into consideration the side effects just because of the sake to arrive to some dreamed place we really don't need, and destroying many things in your path to do so is wrong.

Also, we don't advance by brute force, using huge quantities of computing power. That is not a technological improvement. The magic would be to design something that produce really useful things with a very limited quantity of power ... that other thing that these modern "AI" are doing is to cheat.

Yes ... humans, and humans in quantity, are troublesome.

The important thing is to have a strong person in charge, that can decide quickly the right things. This way no convoluted, complex, expensive and not so useful thing will be done. And to expend enough time deciding "WHAT" needs to be done, debugging on paper until the result is the right one. Then, go to find who could do that, even if could be done internally just purchasing the "right" technology for the task at hand.

But to start without a clear end ... this is a nightmare.

You are right in the sense that complexity can be deceptive, 100% with you.

But I think it is important to pursue simplicity all the time. I have been working with big enterprise servers, mini computers, all the way to raspberry pi and arduinos, and it is so special the way you can unfold the need of power when you distribute and go to edge computing. These are modern technologies that can help to think smaller, to process more near the source, to let less in a central behemoth.

And right now I am writing this on a brand new Apple Mini M4 - with 16G RAM and 256G storage - yes, I didn't go for the hungry storage ones because I can attach an external storage unit for a fraction of the cost. In fact, 4TB cost me less than the 1TB Apple model.... you look for options and you find them. And I am testing a beautiful BananaPi F3 with an octacore RISC-V SoC, 16G RAM, 128G eMMC, and a 1TB Kingston NV3 PCIe 4.0 NVMe ... around $250 with an acrylic box (some taxes and courier, but this is my reality in Costa Rica, not yours); and although I can't compare it with a modern last generation server, it is in fact 3 times fast than an HP Integrity with Itanium (the latest models they did), and the storage is around 100 times faster.

My point here is that fatted systems just erase your advantage and eats your money as ants eating sugar. What you can do with industrial level systems and well designed Just Enough Functionality coding on sub $1000 equipment and open source, is equivalent to what you did 20 years ago on $30000 machines with $100000 software licensing. To expend $50 million dollars on any type of HR system is just nuts. And even if you rethink what you did 20 years ago, that technology is enough for your current needs .... better engineering at hand.

Our life became just ridiculous, with a machine in your pocket more powerful than the equipment sent Voyager beyond our Solar System, to read today's gossip. How can be that? What happens to us?

I have decades of experience in all sorts of software, and I can't see what is the advantage of creating these huge, complex and expensive systems.

It is like some sort of drug. We need to expend more, more, more, we need to use bigger and bigger systems, with huge quantities of storage, when the real tasks are really small and simple.

What it is needed is to stop and check what really work, and to develop "something" that can deal with future natural capacity grow, using preferable small cheap devices that can be replaced by any brand ones when needed, not to be attached to particular technology providers. Today is $50 million, tomorrow will be $500 million and counting if they follow that path.

For me this is as to see a TV cartoon about how to control the world.

As if every possible invention and discovery only can be done in an Occidental country or, in particular, in the USA.

The generative AI that needs all these powerful chips is just one of the many possible types of AI, that have been developed recently ... there is a huge quantity of different ways to do the things, and I am sure many of then without requiring all that fat way of wasting power.

And look about what happened in Russia with the commercial sanctions on food ... now they make their own European type of cheese, and in the future when sanctions stop, they won't need to import these products anymore. Who is really being sanctioned?

Knowledge and Expertise are not limited to a couple of countries. They are around all humans everywhere. Then you invest a lot of money in one direction and suddenly other person discovers something that makes your invention to be just trash.

The AI is really important, but current affairs have all the look of a bubble.

A new bread of feature phones is needed. The smartphone is beyond control by now.

Experience is important and not having it is not free of problems.

This is not about how easy is to drive a submarine ... it is about what you know to do in an emergency.

A good solution is just to cover the led with black tape.

A better one .... for the device providers that acquire the data, DON'T USE THE LEDs on the device to show data activity. In fact, don't use them at all, and save some bucks.

Right now, I am working on a notebook with an i5 and 32GB of RAM (upgradeable to 64). And this machine works better than the same brand with an i7, because both are notebooks, and when the i7 uses more power, the CPU is forced to reduce the speed to save power and reduce heat, while the i5 is happy working at nominal speed. So ... why the extra $$$ for the i7?

At the end those iX just say that the chip have more or less inactive processing units inside. You must deal also with another number and a generation to understand what type of power you are dealing with.

AI is wonderful "by concept" ... but all these XXX-GPT are creating a toxic environment for the technology usage, making it unusable. It was a good test, a good extremely expensive test, but that's all ... it is not a good replacement for other practices as have a lot of failures mixed with good answers. And it is a very difficult to control tool; it just have many variables involved without a clear methodology to assess its trust level.

What could happen is that the run to have "the right" alternative will damage all them, and the good name of AI for a lot of years.

The problem is not that ChatGPT can lie... it's that, sometimes, it could tell the truth.

And ... what about the new Sodium based batteries? Yeah, they are very new, but at least they are not based on lithium.

We can't forget that humans continuously are chasing for technology improvements. So, the "idea" that a "battery" is something wrong is false. The real problem is the wrong type of battery with the wrong care in the wrong environment. And also, there is gravity energy storage (https://qz.com/1355672/stacking-concrete-blocks-is-a-surprisingly-efficient-way-to-store-energy) or other "different" ideas like that for such tasks as to store huge quantities of energy in big places.

Did you know that there is a Raspberry Pi OS for x86_64 machines? He didn't say that he is running Linux on Raspberry Pi, he just said Raspbian.