Global recession over the end of Moore's Law - page 14

 
СанСаныч Фоменко:
Do you mean Trump? Or the people behind him?
Don't take away my last hope for your sanity.
 
Dmitry Fedoseev:
Don't deprive me of my last hope of your reasonableness.

I'm totally with your reasonableness, for my own mercantile reasons, but here's the reality...

Let's wait... I do wish you were right.

 

Some facts about the global mobile industry (service providers) in 2020:

  • Number of subscribers: 4.6 bln
  • Number of smartphones: 5.9 billion
  • Sales: $1.4 trillion
  • Percentage of global GDP: 4.2%
  • Number of jobs: 29 million (direct and indirect employment)
Taken from: http://www.gsmamobileeconomy.com/GSMA_Global_Mobile_Economy_Report_2015.pdf

"The increasing level of maturity in developed markets, combined with the recent strong growth in developing markets, means that there will inevitably be a slowdown in global subscribers. Over the six years to 2014, unique subscribers grew at a CAGR of 7.6%. This figure is forecast to slow to 4.0% over the period out to 2020. Revenue growth is forecast to slow further over the coming years, with a CAGR of 3.1% per annum through to 2020, down from just over 4% in the period 2008-2014. This reflects the ongoing impact of factors such as market maturity, competition and regulation."

And what happens in 2020-2025 when mobile internet speeds have peaked and transistors have stopped shrinking in size?

Claiming that if one mobile industry is affected by a recession then the rest of the economy won't feel it is wrong. Look at what happened in 2008. Theoretically only the banks should have felt the recession due to the drop in mortgages sales. The financial sector's percentage of GDP is about 8%, compared to 4% for the mobile industry. But the recession has affected all sectors. It's like dominoes: increased unemployment in one sector, leads to lower sales in all sectors, investment in all sectors falls due to investor fear, unemployment spreads to other sectors, etc. Look at the Dow or S&P indices. You can clearly see two declines, 2000-2002 and 2007-2009, almost identical in magnitude by the way. The first decline was caused by the .com bubble. At the time, too, one could argue that .com was a very small % of GDP. Market and investment is a delicate matter, depending on the emotions of investors who can send the market down 50%.

 

A flawed analogy comes to mind.

The Chinese Communist Party's "one family, one child" policy, which began in the 1970s and is being phased out this year, has not led to a recession in China. On the contrary, the enormous GDP has increased impressively every year. Why is this so? I think that an only child is getting more in terms of education, health, culture. The backlash has grown from subsequent generations. The Chinese nation has become, shall we say, "smarter".

It is the same with hardware. To make devices more efficient you don't have to produce as many transistors per unit volume as possible. You can improve the software.

Now the possibilities of devices are quite satisfactory for a person in 2D and 3D modes. But if he wants to get on the Internet with all his senses, then yes, some kind of revolution in processors is needed.

 
Комбинатор:
And again, past it. Now there's a surge of interest in neural networks. msqrd is a clear proof of that

A typical local view of reality.

The reality is somewhat different:

There is indeed a surge of interest in msqrd, but it does not affect the overall picture and what is interesting is that all this "interest" is for some reason mostly from Belarus, where the Combinator is stationed. It must be a coincidence ;):

 
khorosh:

But there are programs that beat world chess champions. Of course, these are highly specialised programmes, but if at the same high level a set of programmes could be developed that could perform a wide range of different aspects of human mental activity, then it would really be artificial intelligence. And if such a chess program has been created, I don't see any fundamental limitations that would prevent this task from being solved.

The most difficult thing, of course, is to teach the programme to think up something new (to invent). But not everyone can become an inventor either. Although there are algorithms in inventing too, if you think about it.

Do not confuse chess with the gift of the gods. Chess is a game of combinatorics. Here the computer will always be a hundred points ahead. There is a game called Go, which the computer learnt to beat humans back in the late 80s. However, it has nothing to do with AI.
 
Vasiliy Sokolov:

A typical local view of reality.

The reality is somewhat different:

There is indeed a surge of interest in msqrd, but it does not affect the overall picture and what is interesting is that all this "interest" is for some reason mostly from Belarus, where the Combinator is stationed. It must be a coincidence ;):

You're just not aware that the new round of popularity of neural networks is being searched for by the keywords deep learning or deep learning. These are multilayer neural networks :) They are awesome at solving pattern recognition problems, for example. Better than a human.

 

Modern computer architecture was founded by Thuring and Von Neumann in the 1930s and 1950s. They also laid down a set of axioms within which any computing device operates. Specifically, a modern computer is a pro-version of a Turing machine, or a deterministic automaton, with all the consequences that follow. One of these consequences: it is impossible to check the computability of a given problem. The computer will either solve it, in a finite number of steps, or not. But it will not be able to check it itself. At a primitive level, "program freeze" is just a consequence of this limitation. On the other hand, relying on a number of propositions, or formal assertions, the computer can go from assertion "A" to assertion "B" in a finite number of steps, thereby emulating the thought process. But it would only be an emulation as it is constrained, I repeat, by the axioms of the Turing machine and von Neumann's theory of finite automata. And AI requires that very leap of thought that goes beyond those axioms. It is thanks to this aspect of thinking that one is able to operate with infinite concepts and find new hypotheses, which cannot be arrived at by logical reasoning.

This may not seem obvious to everyone. Many eminent scientists used to think otherwise. Among them was the eminent mathematician David Hilbert. He founded a trend in mathematics known as constructivism. According to this approach, mathematics was to be based on a few fundamental axioms from which absolutely all mathematics could be proved. It was suggested that something automatic and similar to von Neumann's finite state machine (this was before the work of Turing and von Neumann) should be used as a 'solver'. However, this attempt failed. It soon turned out that in mathematics there was a whole host of assertions which could be neither proved nor disproved. They cannot be arrived at by logical operations. However, this does not mean that they are false, abstract or non-existent. It is simply that the current knowledge of mankind is insufficient to understand them. When the time comes, they will also be proven in other axioms that expand our understanding of the world around us. But right now they are waiting to happen and computers, however powerful they may be, will not help here.

In summary: AI will never be created, at least within the framework of modern computing, for the simple reason that the limitations of the Turing machine and the von Neumann finite automaton are imposed on modern computing devices. All modern computing is inherently deterministic and only works when there is a straight path of a finite number of stones from statement A to statement B.

 
Alexey Burnakov:

You're just not aware that the new round of popularity of neural networks is being searched for by the keywords deep learning or deep learning. These are multilayer neural networks :) They are awesome at solving pattern recognition problems, for example. Better than a human.

(chuckles) Okay. Okay, I'm not gonna argue about neurons. Cause I really don't know anything about it. It's just that we're talking about AI. It's a dead end. Why I think that it is impossible to create AI on the basis of modern computing devices written above. Neurons are good, but it is still not AI.
 
Getting back to the topic of the topic of the topicstarter. One thing I do not understand is why Vladimir equates the increase in hardware computing power with the demand for computer hardware. Why does the new smartphone have to be more powerful than the previous one? For example I don't know how many megahertz my CPU has in my smartphone and I don't care about that at all, neither do the vast majority of other users. It is not the speed that is more important now, but the services and programs provided. Cloud capabilities and other features.