AI 2023. Meet ChatGPT. - page 205

New comment
 

You could add comparing LLM to some depersonalised liberalist politician.

He speaks very well and beautifully, and at first we think that he is highly intelligent. But when we start to dive into the details, we are surprised to learn that the guy is a cuckoo. No intelligence.

And this is a very dangerous thing, when beautifully spoken or written words are perceived by the untrained masses as the presence of artificial intelligence.
 
Maxim Dmitrievsky #:

I felt point 1 on myself, well, in terms of becoming interested in all these semantic connections and how we ourselves think with words (actually, we don't think with words). In general, LLM is a paradise for linguists, in terms of studying languages and meanings.

As for purely language(s) - it's a solid A. Well, that's how they were originally conceived.

Imagine, you have learnt to express your thoughts better than others (at least better than Andrei), and you are already rated by others as having a higher intelligence! Although you just learnt to use the structure of the language well.

LLMs can be used for many things, but you can easily do without them. They are cool, but not much needed. Useful, but books and even Wikipedia are more useful (in my opinion).

In practical application - highly overrated by bloggers and adverts.

I've read people who have a really cool speech and I'm far from them. And so - thanks, of course).

And also, we in the thread agreed not to get personal, otherwise we erase posts. No one is an exception.

 

I'll be posting the first part of chapter four today.

However, I decided to make some changes to the previously approved plan. Serious consideration of the psychological effect of conversational AI on the individual's subconsciousness has been postponed to the end of the story, as the topic is complex and requires deep analysis. An obvious indication of "deification" or"grail-mania" syndrome is obviously not enough. Let's try to make something in the spirit of Jung or Blavatsky (archetypes, etc.), but in the context of the market exploitation of the weaknesses of the human subconscious)). Let's add disclosure of the reason why OpenAI so zealously undertook for conversational AI, and not for some other (it's not for nothing). And let's finally solve the mystery of who Musk really is.

But I will add all this in a dosed manner, gradually building a complete picture.

However, first I will focus on the issue of cheaper robotics and factory automation. Using simple examples of emergencies, accidents and breakdowns, we will trace the limits of automation capabilities. After all, the weak link here is not a robot, but a factory with all its nuances. Let's look at them closely and in great detail. Literally at the level of extracting a screw stuck in the machine, which requires the work of a whole team of specialists.

It will be interesting.

 
But before continuing, a thesis summarising the latest material....
 

Main theses of the third chapter:

Incomplete list of production processes and process steps:

  • Extraction of minerals.
  • Processing of raw materials.
  • Logistics.
  • Production of semi-finished products.
  • Production of components.
  • Assembly of the final product.
  • Commissioning.
  • Set-up, maintenance, diagnostics, preventive maintenance.
  • Disposal.

The list gives a general idea of the logical sequence of actions leading to the emergence and operation of a technical product.


Further...


Technical infrastructure:

  • The product arises due to the technical infrastructure that ensures the flow of technological processes.
  • Without technical infrastructure it is impossible to create a final product. Man is a biological creature and focuses on the final product, forgetting about the producing infrastructure.
  • The more complex the product, the more complex the infrastructure and "trickier" the processes in which it is assembled.
  • The more intensively the production infrastructure works - the more accidents, breakdowns and abnormal situations occur.
  • Accidents, breakdowns and emergencies require quick and correct decisions in the face of unpredictable escalation.

Technological processes:
  • Flow production is fraught with problems and breakdowns due to high loads on automation.
  • Production automation is the weakest link in production. Everything that moves breaks down. The main causes are production loads, operating errors, improper maintenance, and wear and tear.

  • The key problems in production are always related to process imperfections.
  • Process failure has 3 possible causes: imperfect technology, poor quality of materials, human factor.
  • Theunprofitability of improvements is an economic reality for many manufactures.
  • Autopilot of the manufacturing process is seen as possible until the first accident. Accidents in production are inevitable, so autopilot is impossible.
  • The causes of accidents are physical, software, human - poor quality of materials, wear and tear, loads, deviations and errors of input parameters.


Continued tomorrow...


 

For some reason, the possibility of disproportionate development of different areas is assumed - AI develops very much, but production technologies remain at the current level.

For example, the current approach to working with metal is extremely inefficient. We can assume that instead of it there will be powder metallurgy, combined with 3D printing and supplemented with the technology of collecting and processing everything broken back into powder.

 
Aleksey Nikolayev #:

Somehow, the possibility of disproportionate development of different areas is assumed - AI develops a lot, while production technologies remain at the current level.

For example, the current approach to working with metal is extremely inefficient. We can assume that instead of it there will be powder metallurgy, combined with 3D printing and supplemented with the technology of collecting and processing everything broken back into powder.

In my time, I often passed by one of the city bookstores where old Soviet technical literature was sold for pennies - shabby volumes in unsightly covers, with wrinkled corners and yellowed pages. I remember curiously picking them up and leafing through them. I was very interested in technology back then. The books were about microelectronics, reinforced concrete structures, automation, industrial chemistry, the general structure of technical systems, mechanics, machine tools, logistics. There was a sea of books and each one contained incredibly complex material. Formulas and diagrams on every sheet. Each book had several authors - engineers, scientists.

Then I understood the level of training of specialists and developers of technical processes, as well as the degree of complexity of the material they considered.

The point is that conversational AI is a useless in engineering talking toys, frantically developing due to purposeful increase of computing power by sellers of processors and video cards. And they develop this direction actively because the market asks for more talking toys. But for metallurgy, industrial chemistry or reinforced concrete constructions, what is the use? There, computing power and vocabulary do not decide).

With all the "power" of modern AI, it cannot solve any task facing specialists and is useless for front end developments. To suggest/draw something obvious and average - please, but to find a new solution - sorry). To solve high school students' problems - sure, to make a discovery in the scientific field - no way.

I think we are approaching the point where the computing power of computers can no longer solve anything. I will explain why in more detail below.
 
Реter Konow #:
In my time, I often passed by one of the city bookstores where old Soviet technical literature was sold for pennies - shabby volumes in unsightly covers, with wrinkled corners and yellowed pages. I remember curiously picking them up and leafing through them. I was very interested in technology back then. The books were about microelectronics, reinforced concrete structures, automation, industrial chemistry, the general structure of technical systems, mechanics, machine tools, logistics. There was a sea of books and each one contained incredibly complex material. Formulas and diagrams on every sheet. Each book had several authors - engineers, scientists.

Then I realised the level of training of specialists and developers of technical processes, as well as the degree of complexity of the material they considered.

There was always a time gap in Soviet books. The technical specialists described the current state of affairs well, while the writers described the distant bright future. The not-so-distant future was always left out of consideration or the representation of it was inadequate. A typical example is the famous passage about the unnecessity of the Windows OS by Soviet engineers in one book.

Powder metallurgy has been developing quite a bit recently - for example, last year the states conducted the first test launch of a rocket, most of which (over 80%) was printed on a 3D printer. With the right level of development, this technology will change everything much more than even the introduction of the electric motor (the second industrial revolution).

 
Aleksey Nikolayev #:

There was always a time gap in Soviet books. Technical experts were good at describing the current state of affairs, while writers were good at describing the distant bright future. The not so distant future was always left out of consideration or the idea of it was inadequate. A typical example is the famous passage about the unnecessity of the Windows OS by Soviet engineers in one book.

Powder metallurgy has been developing quite a bit recently - for example, last year the states conducted the first test launch of a rocket, most of which (over 80%) was printed on a 3D printer. With the right level of development, this technology will change everything much more than even the introduction of the electric motor (the second industrial revolution).

I read a bit about powder metallurgy. Certainly an interesting technology.

In general words: The technology is promising, complex and expensive. It has been developed since the times of ancient Egypt and Incas. Mass production of products in powder metallurgy starts from the middle of the 19th century. Was also well known and developed in Soviet times.
Key disadvantages: high production cost and complexity of technological processes. On the plus side, it allows smelting parts with unique physical properties that are difficult to produce by other methods.

In the context of the above remark about disproportionate development of technological fields, when computing power grows daily and production technologies lag behind or stagnate, I don't see how AI can solve the problems of powder metallurgy. After all, it makes absolutely no difference to it how smart it is.

That is, we are approaching a limit where process physics and economics are not improved by the growing capabilities of virtual simulation and digital optimisation. When the effective management of the manufacturing process can't shift entirely to AI. When the manufacturing process delivers so much data that there is both a lot and little at the same time.

About this further...

 

Well we're talking about metal 3D printing on printers, which has only recently emerged. This is a very obvious opportunity to combine AI with technology. If this is complemented by nanotechnology, the potential opportunities will increase many times over.

I am not saying that this is a matter of the immediate future, but the speed of change since the 19th century is quite impressive, especially compared to the speed of biological evolution.

1...198199200201202203204205206207208209210211212...214
New comment
Reason: