IT vs AI: will cars be taken away from their creators?
Conversations that artificial intelligence systems will sooner or later force out people from a number of professions have been underway for more than a decade. Robots penetrate medicine, heavy industry, solve complex analytical and creative tasks. And The Guardian recently reported that one of the Japanese insurance companies has cut some of its employees, replacing them with IBM's Watson Explorer, which, according to management, should be 30% more productive than people.
According to Andrew Un, in the short run, AI systems can take many away from work — all because the volume of tasks that artificial intelligence can automate is now as great as ever. But what is the future of IT professionals in this regard?
We decided to consider this issue on the example of the three areas of the IT sector and discuss the likelihood that in the near future you will be replaced by a machine.
')
In this material:
Photo Duncan Hull CC-BY
What we automate: answers to frequently asked questions, basic customer training. Many questions that interest customers, especially in the B2C sphere, can really be trusted to the machine: instead of reading a long FAQ or hanging on the phone waiting for an answer, the user sends his question to the chat and gets an immediate answer.
How to automate: chat bots based on machine learning. Existing solutions often use machine learning algorithms and analyze both chat logs with the technical support department and the knowledge base and FAQ.
Pros: quick response. One of the indisputable advantages of the machine compared with the person in this area. It is important for users to get an answer right now - which is what the bot can do, ready to answer the question regardless of what time it is and how many “living” employees work in the office. At the same time, AI systems allow you to make the answer quite "human". Not to mention the fact that the bot does not spoil the mood, and it does not suffer from many hours of routine.
Cons: limited chat bot to recognize complex patterns of natural language. If the situation is non-standard, and the user cannot give an unequivocal answer to the question, it will be practically impossible to manage without a support service specialist. By the way, Jonathan Mugan, co-founder of the DeepGrammar project specializing in the analysis of natural languages, cites an interesting analysis of the problems facing modern chat bots and their possible solutions.
Forecasts and predictions: chat bots based on AI systems have every chance to take root in the field of technical support. Of course, the bot will not answer all the questions - but it will be able to determine when the user really needs the help of a person - and thus reduces the number of employees to the most qualified ones. And this is despite the fact that, in general, users do not particularly need human communication - 40% of HubSpot respondents said that they do not care from whom to receive advice and recommendations. It turns out that to press the "living" experts in this field, the Turing test will not have to be taken by bots.
What we automate: increasing work efficiency, searching for “black swans”, monitoring and diagnostics in complex distributed IT systems.
System administrators are already using a large number of products to help them automate their work. In thematic threads, for example, on Reddit, there are enough examples of how a team of 3-5 system administrators successfully serves the infrastructure necessary for the operation of thousands of client machines.
Nevertheless, solutions based on artificial intelligence appear in this, and so well automated, sphere. Mark Shuttleworth, founder of Canonical (and the second space tourist in the world), says that AI systems allow, for example, to compare and study the logs of multiple servers (both inside and outside the organization) and to compare the statistics of millions service and system deployments. Training based on such a volume of data allows AI systems to find so-called “black swans” - seemingly insignificant events that precede serious problems and in retrospect have a completely adequate rational explanation.
How to automate: an interesting example of building such a system is given by scientists who have worked on a project on the use of machine learning and AI-based systems for monitoring and diagnosing the state of the IT infrastructure involved in the Large Hadron Collider at CERN. The authors described the principles of the system in an article for the Journal of Physics, the mechanism implementation is described in more detail in a doctoral dissertation of one of the scientists.
Pros: judging by the findings of IT specialists from CERN, AI-based solutions may prove to be a good idea for monitoring complex systems. In the end, administrators working on projects like the Large Hadron Collider are unlikely to refuse additional assistance - including artificial intelligence. And an additional opportunity to look for “black swans” in general takes the work of a system administrator to a new level - this is not just the deployment and support of the current infrastructure, but advanced analytics.
Disadvantages: in addition to the difficulties in deploying and implementing such tools, their main drawback (for artificial intelligence itself, but not for the system administrator) is that such solutions are unlikely to be able to completely oust people from the profession - at least in the foreseeable future.
A few more of their shortcomings were identified by Andrew Un in his material for the Harvard Business Review. First of all, whatever the decision itself is, it is of little use without a dataset on which the system could learn and which could be analyzed. Many companies simply do not physically have the resources to gather the array of information necessary for the system to make truly adequate decisions. This slows down their distribution and implementation.
Secondly, according to Andrew Un, to effectively use the algorithms, a talent of a living person is needed (it is not enough just to train the algorithm on the sample, you need to understand the business context, the data features, to be able to correctly interpret the results).
Forecasts and predictions: while Joe Baguley, vice president of VMware, is confident that over time, the AI ​​systems will completely replace the live operators with this work, most analysts, as well as the system administrators themselves, are less radical . As one of the users of Reddit rightly emphasized , any automation (including those based on AI) is quite consistent with the DevOps practices, allows you to do more in less time and with less effort - and is a little like “horror films, in which robots come to the company and select you have a job. "
What we automate: of course, there is a lot of automation methods in programming. In fact, the very development of programming provided developers with the opportunity to gradually move to more advanced means of writing code: from assembly languages ​​to Kotlin. However, some forecasters look further and predict to programmers the complete or partial disappearance of "the fault of the AI."
So, for example, back in 2013, a study by researchers from Oxford published “The future of employment: how sensitive are professions to computerization?”, In the framework of which the authors studied the risks of replacing certain professions with machines. The study puts programmers (computer programmers) in 293 place (out of 702) regarding the possible displacement of people from this profession in the future (the higher the place, the more likely it is to replace).
The work of the programmer, based on the findings of Oxford scientists, can be 48% "replaced" by artificial intelligence. By the way, the first place (absolutely non-computerized work) turned out to be rehabilitation specialists, and the last place (work, which can be trusted with 99% of a computer) —the telemarketers.
How to automate: in principle, such a development option seems likely not only to British scientists. Peter Norvig, an American computer scientist and director of research at Google, also considers this scenario and describes the future of programming as the work of artificial intelligence:
Pros: in theory, the “AI programmer” can analyze all existing code samples written to solve a selected task — such an approach allows the system to become the quintessence of all previously existing programmers.
Disadvantages: as is the case with modern neural networks, the “AI programmer” will work like a black box even in the wildest fantasies of scientists and visionaries. That is, specific conclusions, ideas, reasons for making a decision at one or another stage, and the intermediate decisions themselves will be hidden from the observer. And this is a completely special approach to programming, which is not very compatible with the principles on which modern developers write code and documentation for it. However, it is too early or at least unlikely to talk about such a future - more on this later.
Forecasts and predictions: Iain Smith, head of Diaz Research, a company specializing in IT research and forecasting, is extremely skeptical of "the next findings of British scientists." And in this he is not alone: ​​most of the developers in thematic threads Quora, Reddit and Hacker News are inclined to believe that it is unlikely that they will replace AI programmers.
The most interesting theories in support of this opinion are:
1. Programmers are constantly moving from one level of automation to another - and the emergence of such languages ​​as C ++ and Java has not only deprived developers of work: on the contrary, employers' requests for programmers with different skills and competencies have only increased. Also, the emergence of "programming AI" will not be for developers the end of their profession - on the contrary, there will be new tasks and new requirements for "real" coders.
2. Programmers will become the last profession, from which artificial intelligence will force out a person, because if a computer exceeds human capabilities for programming, we will find ourselves at the point of singularity (transition to recursive self-improvement), and at this moment we will have much more serious problems than employment .
3. Before the emergence of a truly intellectual “substitute” for developers, too much time will pass — humanity will have to overcome a long period of the “ winter of artificial intelligence ” (by analogy with nuclear winter). Therefore, if such developments become a reality, then the bond is definitely not in our time.
The real difficulties for developers, according to Jena Smith, can arise only if the overly inspired by the forecasts of the CEO and CFO decide to cut part of the state at a time when in reality it would be worthwhile for them to expand the number of programmers working for them.
According to Andrew Un, in the short run, AI systems can take many away from work — all because the volume of tasks that artificial intelligence can automate is now as great as ever. But what is the future of IT professionals in this regard?
We decided to consider this issue on the example of the three areas of the IT sector and discuss the likelihood that in the near future you will be replaced by a machine.
')
In this material:
- How realistic is it to replace the system administrator, technical support service and developer with an AI system?
- What are the advantages, disadvantages and unsolved problems of examples of such automation?
- what the leading players of the IT industry, analysts and specialists themselves say about this
- here the black swans, the Large Hadron Collider and British scientists
Photo Duncan Hull CC-BYSupport specialist: only the best will survive
What we automate: answers to frequently asked questions, basic customer training. Many questions that interest customers, especially in the B2C sphere, can really be trusted to the machine: instead of reading a long FAQ or hanging on the phone waiting for an answer, the user sends his question to the chat and gets an immediate answer.
How to automate: chat bots based on machine learning. Existing solutions often use machine learning algorithms and analyze both chat logs with the technical support department and the knowledge base and FAQ.
Pros: quick response. One of the indisputable advantages of the machine compared with the person in this area. It is important for users to get an answer right now - which is what the bot can do, ready to answer the question regardless of what time it is and how many “living” employees work in the office. At the same time, AI systems allow you to make the answer quite "human". Not to mention the fact that the bot does not spoil the mood, and it does not suffer from many hours of routine.
Cons: limited chat bot to recognize complex patterns of natural language. If the situation is non-standard, and the user cannot give an unequivocal answer to the question, it will be practically impossible to manage without a support service specialist. By the way, Jonathan Mugan, co-founder of the DeepGrammar project specializing in the analysis of natural languages, cites an interesting analysis of the problems facing modern chat bots and their possible solutions.
Forecasts and predictions: chat bots based on AI systems have every chance to take root in the field of technical support. Of course, the bot will not answer all the questions - but it will be able to determine when the user really needs the help of a person - and thus reduces the number of employees to the most qualified ones. And this is despite the fact that, in general, users do not particularly need human communication - 40% of HubSpot respondents said that they do not care from whom to receive advice and recommendations. It turns out that to press the "living" experts in this field, the Turing test will not have to be taken by bots.
System administrator: AI to help
What we automate: increasing work efficiency, searching for “black swans”, monitoring and diagnostics in complex distributed IT systems.
System administrators are already using a large number of products to help them automate their work. In thematic threads, for example, on Reddit, there are enough examples of how a team of 3-5 system administrators successfully serves the infrastructure necessary for the operation of thousands of client machines.
Nevertheless, solutions based on artificial intelligence appear in this, and so well automated, sphere. Mark Shuttleworth, founder of Canonical (and the second space tourist in the world), says that AI systems allow, for example, to compare and study the logs of multiple servers (both inside and outside the organization) and to compare the statistics of millions service and system deployments. Training based on such a volume of data allows AI systems to find so-called “black swans” - seemingly insignificant events that precede serious problems and in retrospect have a completely adequate rational explanation.
How to automate: an interesting example of building such a system is given by scientists who have worked on a project on the use of machine learning and AI-based systems for monitoring and diagnosing the state of the IT infrastructure involved in the Large Hadron Collider at CERN. The authors described the principles of the system in an article for the Journal of Physics, the mechanism implementation is described in more detail in a doctoral dissertation of one of the scientists.
Pros: judging by the findings of IT specialists from CERN, AI-based solutions may prove to be a good idea for monitoring complex systems. In the end, administrators working on projects like the Large Hadron Collider are unlikely to refuse additional assistance - including artificial intelligence. And an additional opportunity to look for “black swans” in general takes the work of a system administrator to a new level - this is not just the deployment and support of the current infrastructure, but advanced analytics.
Disadvantages: in addition to the difficulties in deploying and implementing such tools, their main drawback (for artificial intelligence itself, but not for the system administrator) is that such solutions are unlikely to be able to completely oust people from the profession - at least in the foreseeable future.
A few more of their shortcomings were identified by Andrew Un in his material for the Harvard Business Review. First of all, whatever the decision itself is, it is of little use without a dataset on which the system could learn and which could be analyzed. Many companies simply do not physically have the resources to gather the array of information necessary for the system to make truly adequate decisions. This slows down their distribution and implementation.
Secondly, according to Andrew Un, to effectively use the algorithms, a talent of a living person is needed (it is not enough just to train the algorithm on the sample, you need to understand the business context, the data features, to be able to correctly interpret the results).
Forecasts and predictions: while Joe Baguley, vice president of VMware, is confident that over time, the AI ​​systems will completely replace the live operators with this work, most analysts, as well as the system administrators themselves, are less radical . As one of the users of Reddit rightly emphasized , any automation (including those based on AI) is quite consistent with the DevOps practices, allows you to do more in less time and with less effort - and is a little like “horror films, in which robots come to the company and select you have a job. "
Developer: prospects for the distant future
What we automate: of course, there is a lot of automation methods in programming. In fact, the very development of programming provided developers with the opportunity to gradually move to more advanced means of writing code: from assembly languages ​​to Kotlin. However, some forecasters look further and predict to programmers the complete or partial disappearance of "the fault of the AI."
So, for example, back in 2013, a study by researchers from Oxford published “The future of employment: how sensitive are professions to computerization?”, In the framework of which the authors studied the risks of replacing certain professions with machines. The study puts programmers (computer programmers) in 293 place (out of 702) regarding the possible displacement of people from this profession in the future (the higher the place, the more likely it is to replace).
The work of the programmer, based on the findings of Oxford scientists, can be 48% "replaced" by artificial intelligence. By the way, the first place (absolutely non-computerized work) turned out to be rehabilitation specialists, and the last place (work, which can be trusted with 99% of a computer) —the telemarketers.
How to automate: in principle, such a development option seems likely not only to British scientists. Peter Norvig, an American computer scientist and director of research at Google, also considers this scenario and describes the future of programming as the work of artificial intelligence:
Pros: in theory, the “AI programmer” can analyze all existing code samples written to solve a selected task — such an approach allows the system to become the quintessence of all previously existing programmers.
Disadvantages: as is the case with modern neural networks, the “AI programmer” will work like a black box even in the wildest fantasies of scientists and visionaries. That is, specific conclusions, ideas, reasons for making a decision at one or another stage, and the intermediate decisions themselves will be hidden from the observer. And this is a completely special approach to programming, which is not very compatible with the principles on which modern developers write code and documentation for it. However, it is too early or at least unlikely to talk about such a future - more on this later.
Forecasts and predictions: Iain Smith, head of Diaz Research, a company specializing in IT research and forecasting, is extremely skeptical of "the next findings of British scientists." And in this he is not alone: ​​most of the developers in thematic threads Quora, Reddit and Hacker News are inclined to believe that it is unlikely that they will replace AI programmers.
The most interesting theories in support of this opinion are:
1. Programmers are constantly moving from one level of automation to another - and the emergence of such languages ​​as C ++ and Java has not only deprived developers of work: on the contrary, employers' requests for programmers with different skills and competencies have only increased. Also, the emergence of "programming AI" will not be for developers the end of their profession - on the contrary, there will be new tasks and new requirements for "real" coders.
2. Programmers will become the last profession, from which artificial intelligence will force out a person, because if a computer exceeds human capabilities for programming, we will find ourselves at the point of singularity (transition to recursive self-improvement), and at this moment we will have much more serious problems than employment .
3. Before the emergence of a truly intellectual “substitute” for developers, too much time will pass — humanity will have to overcome a long period of the “ winter of artificial intelligence ” (by analogy with nuclear winter). Therefore, if such developments become a reality, then the bond is definitely not in our time.
The real difficulties for developers, according to Jena Smith, can arise only if the overly inspired by the forecasts of the CEO and CFO decide to cut part of the state at a time when in reality it would be worthwhile for them to expand the number of programmers working for them.
Source: https://habr.com/ru/post/338410/
All Articles