You can see previous news in the old version of the news blog. Watch
ChatGPT, or why you should start worrying and take a closer look at AI.
A ghost has been roaming the planet for about a month now. He says. He's writing. He's drawing. He even programs. But in reality it doesn't exist. He is virtual.
People who encounter a ghost fall into three categories:
1) It's all bullshit, he makes mistakes.
2) This nonsense needs to be tied to making money.
3) No, it's not funny, I'll go to tik tok.
And almost none of them think about the very near future. How close? Rough estimate - a few years. And then what? Then you will be fired.
Let's count the economy. So, the approximate figures are as follows - a million unique users per day, we will take the average number of requests from a user equal to 20, we will take the average time to generate a response as 10 seconds. What will we get? In total, about a thousand simultaneously processed requests. There is no data on the amount of hardware required to process requests, but this point can be estimated based on, for example, GPT-3, which requires 700Gb of memory, which is divided into levels, and these levels can work in parallel, which reduces the overall memory requirement. It is also not obvious how to count the number of processors, given the active use of video accelerators in this role. But the economy is important for us, plus accuracy is not so important, and therefore we will take the need for resources to process one request equal to 10 expensive video accelerators and 200Gb of memory. Memory can be estimated at $10 per gigabyte, and let video cards cost $3,000 apiece. In total we get $32,000. The lion's share of this amount is given to us by video accelerators. Well, in total, for a thousand simultaneous requests, we need about 30 million dollars of investment.
What is a $30 million one-time cost for a large business? In translation, this is called "free". Or "free of charge", that is, for free.
Of course, we will need two or three people to maintain the farm with all these video cards and other wires, but these are generally pennies, significantly less than a million dollars a year.
Now let's think.
Imagine yourself in the place of the owner of trillions of dollars, which, we recall for those who do not follow financial policy, can be obtained by pressing a couple of buttons on a computer. And now you are offered a toy that can do more than the average person. For only 30 kilobucks a piece. In the US, the average annual salary is more than 30 kilobucks. And the toy will work about a hundred times more than an ordinary person. That is, in four days he will do more than the same average person, which means payback in a few days. And then there's the net profit. Do you already like the toy?
Okay, let's make allowance for our ignorance and round up the cost of one toy up to $ 100,000. Payback - two weeks. Still a very good deal.
But in response, we will hear objections from the first category of citizens (the full list of categories is shown at the beginning of the text). Recall that the first category believes that this is all garbage.
Let's separate the bullshit from the dollars.
Let's look at a number of examples of the use of toys shown in very recent texts both on this site and on the Internet. Here is an example of a ChatGPT evaluation by a "local" author, so to speak. He describes his impressions of communicating with the demon, but these impressions rather fall into the category of "garbage", although sometimes the author notices that the technology still has potential. How does he come to such conclusions?
The author asks to indicate the differences between static and dynamic frameworks for iOS. In response, the robot absolutely accurately points out the fundamental difference - the static approach to building an application involves the inclusion of additional libraries directly at the time of assembly, and the dynamic approach involves connecting libraries later, dynamically, at the time the application is launched and executed. That is, on the basis of a shaky and wide-ranging question, as well as against the background of an incomplete knowledge base, our demon easily navigated and absolutely accurately indicated the key difference between the approaches. To this exact answer, the author of the text, in full accordance with the expectations for the category of classifying the robot as "total bullshit", stated that in fact, in the context of iOS, you can find some private details that slightly change the structure of the application's executable file. Well, on this basis, the author recognized the ghost's answer as incorrect.
What do we see in this dialogue between man and machine? And we see an excellent ability for deduction on the part of the machine against the background of a pale attempt to protect its inner and fragile psychological world from a potential threat by declaring it non-existent.
Let's remember Dr. Watson (I hope someone else remembers him). His crowning question sounded like this - but how, Holmes? To which Holmes replied - deduction, dear Watson, deduction. If we remember that Sherlock Holmes was considered one of the best detectives in the world, and that Dr. Watson played the role of the "average layman", against whose background Holmes's deductive method shone brighter than all the stars in a clear night sky, then thoughts like " It's a no-brainer!" Indeed, ChatGPT thinks precisely deductively. Yes, suspicions did not deceive you. The robot uses the method of the coolest detective known to us from the tales of citizen Arthur Ignatius Conan Doyle.
Let's explain more - the robot knows how to logic. Have you met many in the daily life of citizens who, according to a vague description, without absolutely accurate knowledge in the subject area concerned, nevertheless formulated exactly the key part of the answer with absolute accuracy? Yes, at the same time, due to incomplete knowledge, they may not indicate to you the presence of a bolt with an inch thread on the left flange of the product you are interested in, but, strictly in essence, they will completely correctly describe the principle of operation of your piece of iron.
An example for understanding. Let's say you observed somewhere a steering wheel (from a car) wrapped in a cloth to absorb sweat and a more pleasant touch. And you ask a stranger to tell you what types of rudders he knows, expecting in response that the stranger also saw rudders with fabric winding and therefore is simply obliged to indicate in the proposed list the type of rudders you like so much. But the stranger did not meet with your favorite type of steering wheels, although he absolutely correctly gave a general classification, in which he did not forget to mention the different types of steering wheel coverage. But the word "fabric" was not pronounced. And then a legitimate question arises - who is closer to the robot, the one who did not hear the expected keyword, or ChatGPT, which, based on general laws, derived the rules for classifying phenomena in a given area?
Now about earning.
The use of a capable virtual assistant in quite applied directions has been reported. The tin woodcutter is tied to testing sites, they even tried to put the role of an analyst on a modest piece of iron, instructing him to write a detailed text on a brief description of the task in the tool for tracking the status of work.
And the robot did it. Yes, it's imperfect in places. Sometimes he has to fix something. just look at the terms of reference - from a detective to an analyst, and, well, a tester too. And this is only in a couple of articles next in time. And I assure you, millions more entertaining ways to take your job away from you are waiting for you in the very near future. Articles on a fashionable topic will definitely tell you in detail how a piece of iron for $ 100,000 can replace exactly your work. Pay attention to the economy - what is your annual salary? Let it be only $10,000, but remember that the robot works about a hundred times faster. That is, even with your salary, it will pay off in one tenth of a year. This is very beneficial if you suddenly have not guessed about it yet.
Objections, as usual for any new technology, are associated with individual flaws, they say here your paint has peeled off, here it hasn’t been painted over at all, so what if your spaceship has already flown to Alpha Centauri?
Yes, you can find fault with trifles for a long time, just as long you can laugh at some illogical answers of a new employee, but let's still remember which of us is a robot, and who still tends to see the big and not find fault with the little things? So far, for some reason, I observe this ability in a robot.
The future is already here.
So, by hiring one ChatGPT proofreader, a business can lay off several employees.
Therefore, the question arises before us - what to do? Most, of course, will go to play and laugh on their favorite tik tok. They won't come back because they'll be fired first. Then it will be the turn of those who believe that "it's all bullshit", because while they find fault with the little things, the robot writes correctly working programs according to a short description. Yes, these are short programs so far. But pay attention, in the same text, our ghost wrote the correct set of actions to perform a more complex task. Let's set him the task of processing each item from his list and issuing a program that executes at each step the algorithm invented by the machine. And then we will train it on other programs. And we'll train again. You shouldn't though, this has long been done by much more skilled people at a firm called OpenAI. Are you already running to buy their shares?
The third category of people who want to use everything to earn money will be divided into those who try to use the electronic miracle directly, and those who "know the price of life" and would prefer to buy inventors along with their wards. The first half will also soon discover that unforeseen difficulties have suddenly arisen. For example, they will increase the price tag for working with a robot. Or they will simply cut off access, because senior comrades do not have enough for something. Well, the second half... Yes, they know why the world needs money.
A world without money.
So why does the world need money? That's right, to exchange them for something needed in the household. That is, money is not needed, in fact. But to win some elections, to put your man in the presidency, who will then give you a lot of delicious food from the bins of the motherland, wow - the prospect. Or do you still think that politicians only care about the interests of voters? Then you should remember the recent scandal with the election of the American President.
So you have trillions of dollars and ChatGPT. You need your president in the oval office. Your actions? I think everything is clear here. We buy iron. $100,000 each. 100 million for 1000 pieces. They are equivalent to 100,000 live bots who, for a salary, form public opinion all day long in various kinds of twitters. A measly 100 million and a hundred thousand quite logical bots are calling on the population of the states to vote for your candidate. A hundred thousand is not enough? No problem, let it be 100 million. The price of the issue is 100 billion dollars. One-twentieth of Apple's capitalization. And you don't need to sell shares, you just need to take a loan. The result - in every social network, on every forum, under every text on this site, the vast majority of commentators are bots. And you are lonely against their overwhelming background. You are in a panic - the public is against me! They don't understand anything! What to do? And most give up. They line up with the crowd. But we know that the crowd is all robots. That is, you will not only lose your salary.
And what to do?
The answer is simple - you have to work. In our swamp, everything is running. You revolve in a narrow circle of your acquaintances and have no idea what they think outside this circle. More precisely - the concept of the outside world is prompted by a crowd of robots, that is, you have a concept, but it is greatly distorted. These are the initial conditions of the problem.
The solution is to correct yourself. Well, whoever believes in this - let's go to tik-tok, it's more fun there! At least we will spend the remaining few years with songs!
The solution is to work. Yes, it is necessary to move away from the usual way of life, which is a serious stress for many. For them, this is work. The only question is whether they can. Many won't be able to. Well, they are also better on tiktok.
To those who can.
In short, the problem is not the robot. As we all know very well, a robot is just a tool. The problem is those who will buy the robot and flood your brain, already overloaded with news, with the overwhelming superiority of the "crowd", from each comment pouring out on you an opinion paid for by "respectable people".
Now think about why someone wants to manipulate you? The short answer is it's profitable. The correct answer is that he has such an ideology.
As long as the world is ruled by the ideology of selfishness elevated to the absolute, fed by immense consumerism, in this world there will always be those who want to kill everyone for the sake of personal pleasures. And that someone is not you. You don't have enough money for a good robot. So do not dream that you will become the world dictator. The world dictator will rise above you. There is no third.
Where is there hope, where everyone does nothing, or where they are looking for a way out? While you are in a group doing nothing. And the ghost flies to you.
It is not enough just to put the tool under control. Because the ideology of "snatch" will require each controller to be the first to use the wonderful box with brains. The cardinal solution is another ideology. But you are not ready for it. There is a process of breaking. You are still hoping for the best. Nothing, a few years will pass and you will find yourself in an absolutely hopeless situation without the slightest doubt. But for now, there is hope. But only if you become the one who can. And you need to "be able" like this - to suppress your ego. It's what keeps you from connecting with others, isn't it? It prevents you from starting to look for ways out along with others. You can't do anything alone, but together people always represent a great force. So you need to work on yourself and on the side of others. We reduce the ego and increase the "We".
Well, the smart one understood, the fool did not read this far. Congratulations - you are in the future. Terminator is already rushing towards you. That's just not to help. Think about it.