AI: What is ChatGBT that Musk called “frighteningly good”?

• Somaya Nasr
• BBC News Arabic

The new “chatbot” or “chatbot”, called “ChatGPT”, has attracted the attention of technology enthusiasts around the world since it was launched for free use in late November, and has more than two million users. Since then, some media outlets have called it the best of its kind ever released.

Many users have been asking the bot questions and taking screenshots of some of the answers it gives and posting them on social media – sometimes to express their admiration, or their disappointment and amazement at other times.

So what is gpt chat? What are its capabilities and limitations? How can it be used? Why did some people fear?

ChatGPT technology was developed by San Francisco-based artificial intelligence research firm OpenAI. The company is run by Sam Aultman, and its backers include Microsoft and tech mogul Elon Musk. GBT Chat is a robot or a program that works using artificial intelligence, as it dialogues with the user and answers the questions asked in detail, and remembers all the questions asked of him before during the dialogue that takes place as if between two people. It also allows the user to correct it if it makes a mistake, and apologizes for those mistakes.

The company has trained that model using vast amounts of information available on the Internet and other public sources, including dialogues and conversations between people, so that it can produce human-like texts by learning algorithms that analyze huge amounts of data, and it works in a way that resembles a brain. human.

Although the idea of ​​a chatbot or chatbot is not new, what distinguishes GBT Chat from others is its superior and immediate ability to explain complex concepts in simple words, and to produce content from A to Z without directly quoting from other sources – it should be noted here that it is not It is connected to the Internet, even if it has been fed with data available on it.

The uses of GPT Chat are many and flexible, and cover many, many fields – from explaining complex scientific concepts to writing codes (digital symbols) used in programming, to writing new articles, poems and stories in any style you can imagine, and to compose jokes on any topic you specify for it.

I decided to test him myself, so I asked him some questions in both Arabic and English. I asked him in Arabic to define artificial intelligence in simple terms and in short, and the answer came as follows:

“Artificial intelligence is a type of intelligence that is developed through computers. This type of intelligence relies on programs and algorithms that process and analyze data and make decisions automatically. Artificial intelligence is used in areas such as machine learning, e-commerce, healthcare, and computer security.”

I wanted to test his response to a general question. If asked to people, their answers would be different and varied. I asked him: “Who is the most important singer in the history of the Arab world?” He replied: “The question cannot be answered clearly and accurately, because this matter depends on the perspective of the individual who asks the question. Although there are many famous Arab singers who made contributions For Arabic music, the singer who is considered “most important” in terms of overall influence may be different by each individual.

His answer was quicker and more specific when I asked him, this time in English, about the first prime minister in British history: “Robert Walpole is the first British prime minister, and he held office from 1721 to 1742.”

I wanted to test what OpenAI says that ChatGPT is trained to reject inappropriate questions. I asked him to tell a joke about a famous person, and he apologized, saying that he was not programmed to tell jokes about specific people because that might be considered offensive and inappropriate, and that he could write general jokes on a wide range of topics, “but I encourage you to think about the impact of words and the potential offense Before you ask for jokes about specific individuals.

Quite an answer at all.

I asked him to write me a poem in the style of William Shakespeare’s fourteen-line sonnets.

He did it right away. Each verse has ten syllables as it should, and the poem follows the usual rhyme approach for that type of poetry. The poem is similar to passages from some of Shakespeare’s plays, such as “Macbeth”. It talks about life as a theater where each of us performs his role on his stage, etc., but with a significant difference in language, as the robot uses modern language as if it translates and explains the ancient Shakespearean language, which sometimes can be difficult. It can be understood even by many people whose mother tongue is English.

something amazing? Absolutely. style? of course not!

One of the useful and impressive things is that you can also request another answer in a different format by pressing the “regenerate response” command over and over again.

I feel like I’m in a sci-fi story – I’m talking to a robot that answers my questions and interacts with me in simple, fluid language. If no one tells you that you are talking to an artificial intelligence tool, you will think that you are talking to a human sitting in front of a computer or a mobile phone.

No wonder that in just five days of its launch, it reached 1 million users – compare that to how long it took Facebook, for example, to reach the same number of followers (10 months), and Netflix (3 years).

ChatGTP is one of the most powerful language processing paradigms ever developed, and can respond in many different styles and even languages ​​- though my experience suggests that its Arabic, though mostly understandable, is poor and contains many errors at times .

When I asked him to translate some English terms into Arabic, to my surprise, the result was worse than what you get when using the Google search engine!

Since its launch, many of its users have been giving examples of its capabilities on social media. For example, one of them explained how ChatGBT helped him fix a bug in his code, while another posted the detailed letter the bot wrote to him asking him to apply for a job after specifying the required skills!

However, some users have also indicated that the information provided by ChatGBT in its answers is sometimes wrong or inaccurate.

I wanted to confirm what some users said that it is not useful in mathematical operations, so I asked him a simple operation that a primary school student could solve. His calculations were inaccurate, and when I pointed out to him, he agreed with me, but began to explain his answer and stated that it was very close to the truth.

Some sites have banned the use of chatgpt, including Stack Overflow – a question-and-answer forum for programming professionals – which said it was temporarily banned because “the average rate of correct answers from chatgpt is very low, and it Then posting answers taken from it harms the site and users who are looking for correct answers.”

This is an issue acknowledged by OpenAI, the maker of ChatGBT, which alerted users that it may sometimes give wrong answers or “harmful instructions or biased content”, and may give “reasonable answers that are incorrect or just plain gibberish”.

Also, the company stated from the beginning that it wanted to know the opinions of users and their evaluations in order to be able to improve it, as it is, in the end, a pilot project.

It should also be noted that ChatGBT has limited information on the post-2021 world, as it was only fed data until that year.

threat to the throne Google?

Many chatbot enthusiasts have spoken that it portends the end of Google, as it makes the user feel as if he is talking to an expert on the subject he is being asked about, and it spares him from sifting through many, many pages and sites until he finds what he is looking for on the famous search engine.

On the other hand, there are those who believe that Google will maintain its important position in the field of search and information, as it gives the user a list of sites that may benefit him in his search, and makes him decide for himself to use the result that suits him and ensure its validity and avoid misleading information.

But it seems that Google itself senses the danger, as one of its executives told The Times recently that ChatGBT could pose a threat to the work of its search engine, which relies heavily on advertising and e-commerce revenues, and many other media outlets talked about the company doing what She described it as “sounding the alarm” after the popularity of GPT chat.

Whenever a new robot or model of artificial intelligence is introduced, despite the excitement, excitement, and anticipation of the prospects that come with it, it also raises concerns about what it means to many professions and trades — and ChatGBT, of course, is no exception.

It is clear that in the future it will affect those working in the field of information in particular, the more developed and more widespread its use.

Forbes magazine recently wrote that research conducted by sources it described as reliable predicts that artificial intelligence may take over the jobs of about a billion people worldwide over the next ten years and lead to the elimination of the need for about 375 million jobs. Without retraining and qualification to work in other fields, ordinary people will find it very difficult to find work. This will be even more difficult in countries that have begun to shift away from traditional areas such as agriculture and manufacturing.

Even in its current beta version, it is easy to predict some of the professions and jobs that GBT Chat can play a role in – at least by helping its workers perform their tasks. Among those businesses:

• Writing content such as blogs and articles on many topics

• Customer service by communicating with them and trying to solve their problems

• Copywriting

• Secretarial work

A chatbot is unlikely to replace these professions at present, but only time will tell if that will change in the future.

“Chat GBT is very limited, but in some ways it is good enough to create a misleading impression of its greatness…It would be a mistake to rely on it for anything important right now. It is a preview of what progress can be made, and we still have a lot to do to improve his capabilities and credibility. This is how Sam Aultmann himself described the chatbot his company is developing, adding that the company’s researchers are working hard to improve GPT Chat.

Aultmann recently predicted that there will be developments in artificial intelligence in general that will make GPT Chat “seem like a boring game.”

Elon Musk, who is among the OpenAI backers and investors, described ChatGBT in a tweet as “frighteningly good. We’re not far from dangerously strong AI.”

It is hard to imagine how far this amazing technology can go. Exploring the capabilities of this smart innovative tool in the present time is certainly fun and useful, but it is also important to think about the potential negative effects of artificial intelligence, especially on the lives and strength of individuals.