Among those reading this text, of course, there are many specialists. And, of course, everyone is well versed in their fields and appreciates the prospects of different technologies and their development. At the same time, history (which “teaches that it teaches nothing”) knows many examples when experts confidently made various forecasts and missed oh-oh-oh-very much:
- “The telephone has too many flaws to be seriously considered as a means of communication. The device is of no value to us, ”the experts wrote. , then the largest telegraph company in 1876.
- “Radio has no future. Aircraft heavier than air are not possible. X-rays will turn out to be a hoax, ”lit in 1899, and you can, of course, joke that British scientists lit it back in the XNUMXth century, but we will measure the temperature in Kelvin for a long time, and there is no reason to doubt that the esteemed lord was a good physicist.
- “Who the hell wants to hear the actors talk?”, - said about the sound film , who founded Warner Brothers in 1927, is one of the best film experts of the day.
- "There is no reason why someone needs a home computer" - , founder of Digital Equipment Corporation in 1977, shortly before home computing took off…
- In our time, nothing has changed: “There is no chance that the iPhone will receive a significant market share,” wrote the CEO of Microsoft in USA Today. in April 2007 before the triumphant rise of smartphones.
It would be possible to joyfully laugh at these forecasts, if your humble servant himself, for example, did not make quite serious mistakes in his field. And if I had not seen how massively many and many experts are mistaken. In general, there is a classic "never happened, and here it is again." And again. And again. Moreover, experts and specialists doomed to make mistakes In many cases. Especially when it comes to damned exponential processes.
Oh, this exhibitor to me
The first ambush of exponential processes is that even knowing how fast they grow (for the same time interval, their parameters change by the same number of times), at the household level it is extremely difficult to imagine such an increase. A classic example: if we move one step forward, then in 30 steps we will cover 30 meters, but if each step grows exponentially, then in 30 steps we will circle the globe 26 times (“Twenty-six times, Carl !!!”) equator:
Source:
Question for programmers: what constant are we raising to a power in this case?
ResponseThe constant is 2, i.e. doubling at every step.
When a process grows exponentially, it results in rapid massive changes that are clearly visible to the naked eye. An excellent example is . In 1900, on Fifth Avenue in New York, one could hardly see a lone car among horse-drawn carts:
And just 13 years later, on the same street, you can hardly see a lonely horse-drawn cart among the cars:
We see a similar picture, for example, with smartphones. Nokia, which rode one wave and was a leader by a wide margin for a long time, but could not fit into the next wave and almost instantly lost the market (see with market leaders by years) is very instructive.
All computer scientists know , which was actually formulated for transistors and has been true for 40 years. Some comrades generalize it for vacuum tubes and mechanical devices and claim that it has been in effect for 120 years. Exponential processes are conveniently depicted on a logarithmic scale, at which they become (almost) linear and it is clear that such a generalization has the right to exist:
Source: This and the next two graphs from
On a linear scale, growth looks something like this:
And here we smoothly approach the second ambush of exponential processes. If growth has been like this for 120 years, does this mean that our exhibitor will remain at least another 10 years?
In practice, it turns out that it is not. In its pure form, the growth rate of computing has been slowing down for several years, which allows us to talk about the “death of Moore’s law”:
Source:
Moreover, it is interesting that this curve can not only straighten out, but also go up with renewed vigor. Your humble servant . Yes, there will be other calculations (inaccurate neural networks), but in the end, if inaccurate abacus and mechanical calculators have expanded the scale to 120 years, then neuroaccelerators are quite appropriate there. However, we digress.
It is important to understand that exponential growth may stop for technical, physical, economic and social reasons (the list is incomplete). And this is the second major ambush of exponential processes - to correctly predict the moment when the curve starts to leave the exponential. Mistakes on both sides are very common here.
Total:
- The first ambush of exponential growth - the indicator is growing unexpectedly fast even for specialists. And to underestimate the exponent is a traditional mistake repeated over and over again. As real harsh professionals said 100 years ago: “Tanks, gentlemen, this is fashion, but cavalry is eternal!”
- The second ambush of exponential growth is that at some point (sometimes after 40 or 120 years) it ends, and it is also difficult to accurately predict the moment of completion. And even Moore's law, at the deathbed of which many technical journalists left their hoofprints, . And it doesn't seem like much!
Exponential Processes and Capturing Markets
If we talk about visible changes around us and about the market, it is interesting to see how different technologies conquered the market. The most convenient way to do this is on the example of the United States, where for more than 100 years different types of market statistics have been kept relatively accurately:
Source:
It is very interesting and instructive to watch how the proportion of homes with wired telephones grew smoothly, and then sharply sank by a quarter in years. . The share of houses with electricity was also growing, but it sank much weaker: people were not ready to give up electricity, even when there was not enough money. And the spread of home radio hardly felt the big economic crisis at all, everyone was interested in the latest news. And, unlike a telephone, electricity or a car, the radio does not have a fee for use. By the way, the rise of personal cars, which was interrupted by the Great Depression, did not recover until 20 years later, wired telephones recovered after 10 years, and electrification of houses after 5.
It is clearly seen that the spread of air conditioners, microwaves, computers and smartphones took place much faster than the penetration of new technologies before. From a share of 10% to 70% growth often took place in just 10 years. Technology from the turn of the century often took more than 40 years to achieve the same growth. Feel the difference!
From amusing personally for the author. Check out how washing machines and clothes dryers have grown quite in sync since the 60s. It's funny that we have the second almost unknown. And if in the USA, from some point on, they were usually bought in pairs, then our guests very often ask the question: “Why do you need two washing machines?”. I have to answer with a serious look that the second one is in reserve, in case the first one breaks down.
Also note the drop in the share of washing machines. At that moment, public laundries were very common, where you could come, load the laundry into the car, wash it and leave. Cheap. Similar items are still very common in the United States. This is an example of a situation where the business model of a particular market changes the share of technology penetration and the sales structure (expensive professional machines with anti-vandal performance are better sold).
The acceleration of processes is especially noticeable in recent years, when the mass penetration of technologies has become “instantaneous” by the standards of the early 20th century (in 5-7 years):
Source: (the chart is interactive!)
At the same time, the rapid rise of one technology is often the fall of another. The rise of the radio meant pressure on the newspaper market, the rise of the microwave reduced the demand for gas ovens, and so on. Sometimes the competition was more direct, for example, the rise of cassette recorders drastically reduced the demand for vinyl records, and the rise of the CD drastically reduced the demand for cassettes. And the torrent killed them all with the growth of digital distribution of music, industry revenues fell by more than 2 times (the graph is surrounded by a mournful black frame):
Source:
Similarly, the number of photographs taken is growing exponentially, moreover, recently, with the transition to digital, the growth rate has greatly increased. Therefore, the "death" of analog photos was "instantaneous" by historical standards:
Source:
Full of drama who ironically invented the digital camera and missed the exponential rise of digital photography is extremely instructive. But the main thing that history teaches is that it teaches nothing. Therefore, the situation will repeat itself over and over again. If you believe the statistics - with acceleration.
Total:
- There are many predictive benefits to be gained by studying the acceleration and deceleration of market growth over the past 100 years.
- The rate of innovation is increasing on average, which means that the number of false predictions will increase. Be careful…
Let's go to practice
Of course, you think that all this is quite simple, understandable, and, in general, it is not too difficult to take all this into account in forecasts. It's you in vain ... Now the most interesting will begin ... Buckled up?
Recently, Igor Sechin, Executive Director of Rosneft, spoke at the St. Petersburg International Economic Forum, where, in particular, he said: “". Does anyone doubt that Sechin is an expert in his field? I think no.
At the same time, in recent years, the share of alternative energy has grown by about 1% per year, and the growth of the share has accelerated:
Source: (This method of calculation was chosen - without large hydropower, since it just gives the current 12%).
And then - a task for the 3rd grade. There is a value that in 2017 was equal to 12% and grows by 1% per year. In what year will it reach 16%? In 2040? Have you thought well, my young friend? Note that by answering "in 2021" we are making the classic mistake of making a linear prediction. It is more reasonable to take into account the exponential nature of the process and make the classic three predictions:
- "optimistic", taking into account the acceleration of development,
- "medium" - based on the fact that the growth rate will be like the best year in the last 5 years
- and "pessimistic" - based on the fact that the growth rate will be on average the same as the worst year in the last 5 years.
At the same time, even according to the average forecast, 16.1% will be achieved already in 2020, i.e. next year:
Source: author's calculations
For a better understanding (of exponential processes), here are the same plots on a logarithmic scale:
They show that the average scenario is quite a trend, even if you look from 2007. In total, the value predicted for 2040 will be reached with a high probability next year, maximum in a year.
To be fair, Sechin is not alone in making such a “mistake”. For example, BP (British Petroleum) oilmen make an annual forecast, and they are already being trolled that they, making forecasts for years, again and again do not take into account the exponentiality of the process (“Derivative? No, have not heard!”). Therefore, every year for many years they have had to raise their forecast:
Source:
Closer to Sechin forecasts (parties with a large weight of oilmen, evaluate the pipes at the root of the Russian section of the site). In principle, they do not take into account the exponentiality of the process, which leads to an order of magnitude error for 7 years, and repeat this mistake systematically:
Source: (the site itself , by the way, very good)
Their forecasts with more recent data look especially funny (you also read “well, when will they stop, finally !!!” in their curves?):
Source:
This is indeed counterintuitive, but when predicting many processes, it is more efficient to take into account not a linear forecast for the previous period and not a linear forecast based on the current derivative, but a change in the speed of the process. This gives the most accurate result for similar processes:
Source:
In English-language literature, especially in business analytics, the abbreviation CAGR is constantly used ( - the link is given to the English-language wiki, and it is characteristic that there is no corresponding article in the Russian-language Wikipedia). CAGR can be translated as "compound annual growth rate". It is calculated according to the formula
where t0 - starting year tn is the end year, and V(t) is the value of the parameter, which is supposed to change according to the exponential law. The value is expressed as a percentage and means how many percent a certain value grows per year (usually some kind of market).
The network has a lot of examples of how to calculate CAGR, for example, in Google docs and Excel:
- (suitable Russian text )
Let's conduct a short master class under the motto "help Sechin", taking the oil BP data (as an estimate from below). For those who are interested, the data itself is , you can copy yourself and count differently. Global renewable generation is growing rapidly:
Source: Hereinafter, on black charts, the author's calculations
The scale is logarithmic, and it is clear that all regions have exponential growth (this is important!), many with exponential acceleration. As expected, the leaders are China with its neighbors, which has overtaken North America and Europe. Interestingly, the penultimate one - the Middle East - is one of the most oil-rich regions on the planet, and it has the highest CAGR among all (44% over the past 5 years (!). It's not surprising to see growth by an order of magnitude in 6 years, and judging by the statements of their officials, they are going to continue in the same vein. Back in 2000, the former oil minister of Saudi Arabia wisely warned his OPEC colleagues: “The Stone Age did not end because the stones ran out,” and it seems that they took into account this wise thought 10 years ago. The CIS (CIS), as we see, is in last place. Growth rates, however, are not bad.
CAGR can be calculated in different ways. For example, let's plot the CAGR for every year since 1965, for the last 5 years, and for the last 10 years. You get such a curious picture (in total for the world):
It is clearly seen that, on average, exponential growth accelerated and then slowed down. Moskovsky Komsomolets and other yellow media in this case usually write something like “The Chinese economy is falling”, meaning “the fantastic growth rates of the Chinese economy are slowing down” and tactfully silent that they are slowing down to such rates that others can only dream. Everything is very similar here.
Let's try to predict production in 2018 using data up to 2010, taking CAGR'1965, CAGR'10Y, CAGR'5Y and a linear forecast from 2010 relative to 2009 and relative to 2006. We get the following picture:
| Linear'1Y | Linear'4Y | CAGR'1965 | CAGR'10Y | CAGR'5Y | |
| Renewable production in 2018, forecast based on data up to 2010 | 1697 | 1442 | 1465 | 2035 | 2429 |
| Relationship to the real in 2018 | 0,68 | 0,58 | 0,59 | 0,82 | 0,98 |
| Forecast error | 32% | 42% | 41% | 18% | 2% |
Characteristic moments - none of the forecasts turned out to be too optimistic, i.e. missing everywhere. In the most optimistic scenario with a CAGR of 15,7%, the undershoot was 2%. Linear forecasts gave an error of 30-40% (a period was specially taken when, due to a slowdown in growth, their error is smaller). Unfortunately, it was not possible to add the Sechin model, since it is not possible to restore his formula.
As a homework exercise, try hindcasts by playing around with different CAGRs. The conclusion will be obvious: exponential processes are better predicted by exponential models.
And as a cherry on the cake, we present the forecast from the same BP, whose exponent (“Beware, specialists are working!”) is replaced by a linear growth in the forecast:
Source:
Note that they do not count hydropower at all, which is a classic renewable energy source. Therefore, their assessment is even more conservative than Sechin’s, and they give 12% only for 2020. But even with the undervaluation of the base and the stop of exponential growth in 2020, they have a share of 2040% in 29. It doesn’t look like Sechin’s 16% ... It’s just some kind of trouble ...
It is clear that Sechin is a smart person. I am an applied mathematician by profession, not a power engineer, so I can’t adequately answer the question about the reason for such a serious error in Sechin’s forecast. Most likely, the fact is that this situation really smells like a decline in oil prices. And at (for those who have not listened to this song by Semyon Slepakov - look) for a not very clear reason, a stable exchange rate for the sale of crude oil abroad, and not refined products. And if you strongly distort the forecast, then this saves (for a while, one must think) from unpleasant questions. But as a mathematician, I would prefer to see a systematic error at least at the level of the gentlemen from BP who have not heard of derivatives. I don't care, I'm on the same ship.
Total:
- As all officers know, in wartime conditions, the value of the constant π (the ratio of the circumference of a circle to its diameter) reaches 4, and in special cases even up to 5. Therefore, when it is very necessary, the forecast of specialists demonstrates ANY values necessary for the authorities. It is advisable to remember this.
- Exponential processes are best predicted using the compound annual growth rate, or CAGR.
- Sechin's forecast at the St. Petersburg International Economic Forum can be assessed as an extreme degree of disrespect for the audience or as gross manipulation. By choice. Let's hope that there are brave people who will ask unpleasant questions. For example, why is petrochemicals all over the world very profitable, while Russian state-owned companies are investing tens of billions in the "pipeline" and the export of raw materials, and not in it?
- And finally, I would like to hope that one of the readers will make a page about in Russian Wikipedia. It's time, I think.
Solar power
Let's fix the topic of exponential processes. The latest BP chart shows how the share of the "sun" jumped sharply in 2020, and even the conservative BP believes in its future. Interestingly, an exponential process is also observed there, which, like Moore's law, has been going on for more than 40 years and is called Swenson's Law:
Source:
The general meaning is simple - the price of the module falls exponentially and production grows exponentially. As a result, if 40 years ago it was a technology with a cosmic (in every sense) cost of electricity, and it was suitable mainly for powering satellites, then today the cost per watt has already fallen by about 400 times and continues to fall (soon by 3 orders). The average CAGR on value is about 16% with growth up to 25% in the last 10 years, which happens infrequently.
As a consequence, this also causes an exponential growth in installed capacity and generation:
Source:
Growth by 10 times in 7–8 years is very serious (calculate the CAGR yourself, get 33–38% (!)). Laughter is laughter, but if it is not stopped, then only solar energy will generate 100% of the world's electricity needs in 12 years. This must be fought resolutely. To somehow slow down this disgrace in the US, last year Trump imposed a huge (for other markets) 30% import duty on solar panels. But the damned Chinese reduced prices by 34% by the end of the year (in a year!), not only zeroing out the effect of duties, but also making it profitable to buy from them again. And they continue to build fully robotic factories producing tens of gigawatts of batteries a year, lowering prices over and over again and increasing production volumes. Kind of a nightmare, you know.
The fall in the cost of batteries is such that in recent years they have not only become competitive without subsidies, but the frontier of their cost-effective use is rapidly moving north in the northern hemisphere, covering up to hundreds of kilometers a year. Moreover, yesterday it was important to direct the batteries at the optimal angle and all that. It takes 3-4 years, and for the same price, a large area of batteries can be installed simply on the vertical southern facades. Yes, they are less effective, but they need to be washed less often and easier to mount. And at the same installation price, lower cost of ownership is more important.
Again, the Achilles heel of solar energy is uneven power generation, especially in conditions where storage efficiency is far from 100%. And then it turns out that with such a rate of decline in the cost of producing one megawatt, very soon not only low and medium storage efficiency is covered (i.e., it is possible to store in a less efficient, but cheaper way), but also the cost of installing batteries (i.e. for those same money, we can install not only so many megawatts of generation, but also so many megawatts of storage “for free”, which radically changes the alignment).
Total:
- Swenson's law is approximately the same as Moore's law in terms of validity, however, the CAGR is smaller. But exactly in the next decade, its effect will be most noticeable.
- This is a completely separate topic, but thanks to the rapid development of the "sun" and "wind" in the last 3 years, some insane billions have been invested in industrial energy storage systems. Naturally, Tesla is here , which showed . Gas workers . However, the fun has not yet begun, as several technologies threaten to overtake Li-Ion in the fall in storage costs. However, this is a completely different story, we will be interested in their CAGR in a couple of years (now it is fantastic, but this ).
Electric cars
Serious experts wrote in the highly respected Scientific American as early as 1909: “That the automobile has practically reached the limit of its development is confirmed by the fact that during the past year there have been no improvements of a radical nature.” Last year, there were no radical improvements in electric vehicles either. This gives grounds to assert with all confidence that the electric car has definitely already reached the peak of its development.
In all seriousness, most technology has a chicken-and-egg problem. Until the mass character reaches a certain value, it is extremely expensive to introduce a number of innovations, and, on the contrary, until they are introduced, sales are hindered. Those. to overcome "childhood diseases" a certain mass production is needed. And here it is convenient to evaluate innovative technologies by the level of total production per capita:
Source:
I am not an expert and I do not know how electric cars will change in the next 15 years. But this is definitely a very high-tech product, and they change quickly. And the level of current electric vehicles is the level of ICE cars in 1910 and the level of mobile phones in 1983. Changes for the better (for the consumer) in the next 15 years will be dramatic. And that's when the fun begins.
In general, electric vehicles are pushed forward by three factors:
- When you step on the gas, you fly forward like a sports car, and the price is noticeably less than that of a sports car. And electric cars overtake them on short tracks (, , for example, for this reason teslas ) What Russian American policeman does not like to drive fast?
- Refueling is very cheap, if not worth anything. Roman Naumov living in Canada () causes burning irritation, describing how he, an infection, drove 600 km outside the city, spending $ 4 on fuel (or he could not spend at all). Elon Musk, I remember, complained that many wealthy owners of an expensive Tesla drive it to refuel on a free supercharger, a damned freebie. In short, fuel is almost leaving the expense items.
- And all the engineers say in unison that when childhood illnesses are cured, an electric car will cost much less to maintain. That will be much cheaper. Only tires, they say, have to be changed more often, they wear out ...
And, of course, the fact that the car can, in principle, be charged anywhere where there is an outlet is a revolution. That is, if electricity reached the grandmother in the village, you can come to her and recharge, albeit longer. Of course, you won’t be able to drive a country trophy, but 99. (9)% of people come to the village, and then the car still stops. And tomorrow it will not just stand, but consume electricity at a cheap village rate.
Of course, while there are few charges, especially fast ones, but ... let's look at the graph:
Source:
What? Another exponential process? And which! The question is: how will the situation change if in the next 10 years the number of filling stations increases 1000 times (“A thousand, Carl!”)? (This is CAGR=100%, i.e. doubling annually) Sorry, I made a mistake. In the next 8 years 1000 times! (This is a CAGR=137%, i.e. faster than the annual doubling). And two of those 8 years are almost over... And people from the industry say that the next 8 years will not grow by 3 orders of magnitude, but faster, especially with new generation forks. To understand what it will look like, you need to come to China. In fact, there are electrical outlets in most parking lots and they grow like mushrooms after rain in warm weather. And even the inhabitants of high-rise buildings will refuel for a week on a Sunday trip to the cinema or a shopping center (where the car still stands and waits for you for a couple of hours). And shopping centers with restaurants will fight for a visitor on an electric car (in China they are already fighting).
Yes, the price of electric vehicles is too high now. But the battery gives a large share there, and its cost drops like this:
Source:
Yes, they agreed! This is again an exponential process! And the average CAGR is -20,8%, which, as we know, is VERY high. If 5% is 2x in 15 years, but 20% is 10x in 12 years (“Ten times, Carl!”):
It's funny that at this rate in 3-4 years, instead of one battery in your car, you can buy two at the same price. Hang the second in the garage, and it will provide you with a personal supercharger. Come home and refuel. And at the night rate. And the whole house will be fed at the night rate. And power outages in the cottage settlement will cease to excite. And (remembering the CAGR of the "sun") - it will be possible to put solar panels on the roof. There are good savings, so the mass of people will say: “Class! I take! Wrap it up! (mostly in и , Certainly).
The amazing thing is, these exponential processes. In the next 10 years, we will definitely see serious progress in the field of electric vehicles and modern electric vehicles will be perceived as terribly uncomfortable and miserable. No power reserve, no autopilot, you have to carry a bunch of adapters ... Early models, in short.
Total:
- Electric cars sold in China in the first half of 2019 . During the same period, sales of ICE cars fell by 12%. It's not a bell, it's a gong.
- The most hyped among electric cars, of course, is Tesla. But I would draw your attention to the Chinese . She looks like the most .
- In China, electric vehicle license plates are green. The authorities promise that soon, during the days of the “red” level of smog, they will stop letting all cars into the center of Beijing, except for electric ones. Taxi companies buy electric cars by the thousands. The author rode such a taxi, it looks impressive.
What's going on in IT?
Moore's Law has become well known as it has endured with a huge CAGR of around 41% for almost 40 years. What other examples of good CAGRs are there in IT? There are many of them, for example, Google's revenue growth with a CAGR of 43% over 16 years:
Source:
Looking at this graph, some of the people (especially those whose applications were banned from the Google Play Store) felt uncomfortable. There is something to think about here. Last week, while driving a car, the smartphone began to persistently suggest switching to Google navigation, despite the fact that I was already driving with Yandex.Navigator. Probably, they already lack the size of the market, but they need to raise incomes, I thought. And I thought too.
However, there are also more optimistic purely technical graphs, for example, given on a logarithmic scale, the decrease in the price of disk space and the increase in the speed of Internet connections by 2019:
Source:
It is easy to see that there is a tendency to reach a plateau there; growth rate is reduced. Nevertheless, they grew well for decades. If you look at the hard drives in more detail, you can see that the next return to the exhibitor is usually provided by the following technology:
Source:
So we are waiting for SSDs to catch up with HDDs and leave them far behind.
Also, with an excellent CAGR of 59%, the cost of digital camera pixels fell at one time (Handy's law):
Source:
In the last 10 years, the camera pixel size has also been exponentially decreasing.
Also, with a good CAGR of about 25% (10 times in 10 years), for about 40 years, the cost of a conventional display pixel has been falling, while the brightness and contrast of pixels are also growing (i.e., higher quality is offered at a lower price). By and large, manufacturers no longer know what to do with pixels. 8K TVs are already quite affordable, but what to show on them is a good question. Any number of pixels could be absorbed by autostereoscopy, but there are unresolved issues. However, this is a different story. In any case, the enchanting reduction in the cost of a pixel brings autostereoscopy closer.
In addition, the exponential spread of many software services:
Source:
For example, AppleTV or Facebook. And, as mentioned above, thanks to social networks, among other things, the speed of dissemination of innovations is increasing.
Total:
- Largely due to the exponential processes over the past 20 years, IT companies have greatly pushed others into the list of the largest companies in the world. And they're not going to stop (whatever that means).
- Improvements in most technologies in IT are exponential. Moreover, the classics are S-shaped curves, when in the same area one technology replaces another, each time causing another return to the exponent.
Neural networks
Neural networks have been extremely popular lately. Let's look at the number of patents on them in recent years:
Damn... Again, it looks like an exhibitor (although the term is too short). However, if we look at startups over a longer period, the picture is approximately the same (14 times in 15 years - this is a CAGR of 19% - very good):
Source: (yes, yes, I know what's in there in the next 3 years)
At the same time, neural networks in many areas unanimously demonstrate a better result than the average person:
Source:
And okay, when the result is on ImageNet (although the direct consequence is a new generation of industrial robots), but the picture is the same in speech recognition:
Source:
In fact, neural networks have just overtaken the average person in speech recognition and are well on their way to overtaking them in recognition in all common languages. Wherein, .
How they joke about this topic - until recently, we thought: yes, soon robots will be able to show tricks at the level of monkeys, and it was assumed that it was very far from the level of a stupid person, and even more so Einstein:
Source:
But suddenly it turned out that the level of an ordinary person had already been reached (and continues to be reached). ), and to the level of a rare genius (as shown by competitions with a person in chess and Go) the distance was suddenly less than expected:
Source:
In chess, outstanding people were overtaken about 15 years ago, in Go - , and the trend is:
Source:
As the legendary CEO of General Electric, Jack Welch, once said: “If the rate of change on the outside is greater than the rate of change on the inside, the end is near.” Those. if a company does not change faster than the situation around it changes, it takes a very big risk. Alas, he left the post 18 years ago, and since then, GE's business has deteriorated markedly. GE does not have time to change.
Recalling the predictions about the phone of Western Union specialists, Lord Kelvin's forecasts, estimates of the market for home computers Digital Equipment and smartphones from Microsoft, against the background of Sechin's forecasts, I have reasonable fears. Because history repeats itself. And again. And again. And again.
Many specialists stop developing further after studying their field at the institute/university. And forecasts are made using the technologies of the last century (in every sense). In recent years, I have been tormented by the question - how quickly will neural networks replace experts who do not know how to apply CAGR? And just like that I want to make a forecast, and I'm afraid to make a mistake. In the direction of undershoot, as you understand.
But seriously, the rapid pace of change is like the wind. If you know how to set sails correctly (and the sailboat matches), then even an oncoming one will not prevent you from moving forward, and if it is a passing one, and even with a large CAGR !!!
A passing CAGR to all who have read it in the back!
UPD
Habraeffect still works! On the day this material was published, an article about ! The example has not yet been translated, but a start has already been made. Additionally, you can see or
AcknowledgementsI would like to sincerely thank:
- Laboratory of Computer Graphics VMK Moscow State University. M.V. Lomonosov for his contribution to the development of computer graphics in Russia and beyond,
- personally Konstantin Kozhemyakov, who did a lot to make this article better and clearer,
- and, finally, many thanks to Kirill Malyshev, Egor Sklyarov, Ivan Molodetsky, Nikolai Oplachko, Evgeny Lyapustin, Alexander Ploshkin, Andrey Moskalenko, Aidar Khatiullin, Dmitry Klepikov, Dmitry Konovalchuk, Maxim Velikanov, Alexander Yakovenko and Evgeny Kuptsov for a large number of useful comments and edits that made this text much better!
Source: habr.com