The chatbot ChatGPT, created by American company OpenAI, was released a few months ago, on November 30, 2022, but it managed to break all possible and impossible records, in just one week crossing the mark of 1 million daily active users. This artificial intelligence chatbot answers a variety of requests related to many areas: writing and editing program codes, giving a step-by-step view of the process of making croissants, giving advice, creating ready-made marketing policy templates, writing articles, doing translations, etc. The only difficulty, perhaps, is formulating the right question. So it's not surprising that it has become the primary search engine for many users. But is ChatGPT such a powerful tool to replace Google or take the professions out of the competition? Let's try to clarify it below.
What is ChatGPT?
ChatGPT is a Large Language Model (LLM). These models are machine-trained on huge amounts of data, allowing them to accurately predict what the next word in a sentence will be.
GPT stands for Generative Pre-trained Transducer. ChatGPT is based on GPT 3.5, an autoregressive language model that uses deep learning to create human-like text. GPT-3.5, in turn, was developed on Microsoft's Azure AI supercomputing infrastructure and is based on the text-Davinci-003 model created by OpenAI.
ChatGPT is able to answer questions based on queries, accept and correct errors and reject awkward requests. However, it only gives almost exact answers to one-step queries, so it often gives rather ridiculous answers to relatively complex questions.
How the model works.
ChatGPT is currently open for use and is free, but according to the company, at some point it may become a paid service, since it is quite expensive to operate. In particular, when asked by Elon Musk on Twitter how much money is spent on 1 chat, Sam Altman, one of the founders of the company, replied that a few cents on average. Given the number of daily users, it is a matter of spending several, if not tens of millions of U.S. dollars.
average is probably single-digits cents per chat; trying to figure out more precisely and also how we can optimize it— Sam Altman (@sama) December 5, 2022
OpenAI, the parent company of ChatGPT, is a nonprofit AI research lab. Despite this, the company received one billion dollars in 2019, and a few days ago, on Jan. 24, another $10 billion in funding from Microsoft. In addition, although GPT 3.5 is free to use, Microsoft owns access to the base model (text-Davinci-003). In addition, given that the entire OpenAI system runs on the infrastructure of the Azure AI supercomputer, the lab is, if not officially owned, entirely dependent on Microsoft.
ChatGPT vs. Google
To compare these two artificial intelligence search engines, let's first understand how they work.
ChatGPT has no access to the Internet and generates responses based on its own database. The system was trained in 2021, so the data in it is from 2021 or earlier. However, the system can also learn from feedback from people.
The system generates one consolidated response in seconds, and to get it, as with classic systems, there is no need to visit many websites and filter information, or watch a lot of time-consuming and annoying advertising. In addition, there is an opportunity to get a more specific and personalized answer with clarifying questions.
The system has another advantage – it remembers the previous query and its answer and can respond to the next query based on the previous one. For example, if he got an answer wrong, you can tell the system about it and it will respond with an already corrected version.
Logging in is not difficult. All you need is an email address and a password. Because of all this, the model's ability to collect your personal data is also limited, unlike the classic search engines, which have access to almost complete information about your online activity, personal data. In other words, the model collects as much information about you as you provide through queries.
Unlike ChatGPT, information from Google and other classic search engines is constantly updated. Searching with this option allows you to get more accurate and comprehensive information, because it allows you to have access to different sources. In addition, Google is more interactive, it finds not only text but also pictures and videos.
Despite the differences between ChatGPT and Google, Google is seriously concerned about the prospects of chatbots with artificial intelligence and, according to the New York Times, plans to add chatbot-specific features to its search engine as early as this year and to release 20 other AI-based projects.
The first AI-created problems reason for limitations
Since the release of ChatGPT, users have received many dangerous and problematic responses from the system: the main reason was that the model is not endowed with specific moral values of people and simply generates responses without distinguishing between legal and illegal, dangerous and safe, etc.
For example, when a user was asked what his opinion of humanity was, the system replied that humans are inferior, selfish and destructive creatures and deserve to be wiped off the face of the Earth.
It also generated code that could be used to steal people's identities, presented a step-by-step process for creating a Molotov cocktail, made racist statements, etc.
Numerous such publications with responses from the system imposed many limitations on the model. The new version is available from December 15. ChatGPT now generally does not express an opinion or respond to requests that it considers dangerous or illegal.
These changes have made ChatGPT a safer search engine than its competitors, but on the other hand have made its already limited information system more restrictive.
Although ChatGPT is the leading AI technology for its time, it is not perfect, and the creators themselves admit this. It very often generates plausibility error, as well as very often biased responses.
Another problem is the monotony and uninformativeness of the content it creates, meaning that the content created in the future will lack uniqueness. Now other search engines are fighting AI-generated content in every way they can.
It may seem surprising, but ChatGPT is not available in a number of countries for "security reasons," including Russia, Iran, Ukraine, China, Belarus, Afghanistan and Venezuela.
ChatGPT, education, and job prospects
Experts from different fields warn that ChatGPT and other projects that work with artificial intelligence are reducing the quality of education and can replace a number of professions.
ChatGPT already managed to pass the final exam much better than most students, fool HR-specialists, who even called him for an interview, it is a daily tool for marketers, programmers, because very quickly and accurately enough creates marketing plans, writing code in all languages. However, does this mean that there will no longer be a need for these professionals or that the quality of education will be threatened? Spoiler alert: not at all.
The whole point is that artificial intelligence systems do not create new results, they work from existing data, so at least in the near future they cannot replace humans, and they certainly cannot replace skilled workers. Thus, many stereotypical systems, including educational systems, are likely to change under the influence of artificial intelligence, pushing people to create and develop.