At the Qatar Web Summit, we tried to cover the events as much as possible with the help of artificial intelligence...but!!

(Associated Press)

The Web Summit, which was held in the Qatari capital, Doha, ended, leaving behind many successes, whether for companies, government agencies, or even the private sector and individuals.

Everyone was a winner at this summit except artificial intelligence.

Before the summit, I was very optimistic that artificial intelligence technology was currently capable of helping me, as a technology editor at a site like Al Jazeera, cover the summit well, and that I could save time, effort, and even resources by using this technology in preparing, organizing, and even helping to produce content.

I was fully prepared before the summit, and by prepared I do not mean that I subscribed to paid GPT chat, but rather I meant that I made a complete plan to rely heavily on the assistance of artificial intelligence in many tasks that require effort and time to perform.

A plan that included artificial intelligence tools, linking different tools, and a special methodology in editing and research.

In my optimism, I even started naming my report for the next day after the Web Summit, which was completely different from the title of this report.

But unfortunately, during the real test, all illusions fell away, and the bitter truth emerged that I thought should be documented in a report so that no one would fall under the illusion that these tools can help the editor work in a pressure environment such as conferences and field coverage.

Preparing the coverage plan

The plan to use artificial intelligence in covering the web conference was not random. I was careful to plan in detail for this experiment, which I knew was just an experiment that might succeed or fail, but the results were extremely bad.

I have set three main goals for using artificial intelligence to help me cover this summit:

  • Covering the largest number of sessions at the summit.

  • Assist in preparing for scheduled interviews with guests at the Summit.

  • Assist in filtering and classifying materials according to the platforms on which they will be published (website and social media platforms).

It was never one of the goals to ask artificial intelligence to produce complete materials for the sessions, and I advise everyone against producing complete content using artificial intelligence.

Although it is very important, the deterrent to not doing so is not only the moral aspect, but the technical aspect and the consequences resulting from it.

According to statistics, it was found that the best current artificial intelligence systems cannot write a good article that encourages the audience to interact.

Digital marketing expert Neil Patel says that more than 70% of the articles produced by these systems received the lowest engagement rate versus the highest number of views.

Therefore, I did not waste time relying on artificial intelligence to produce complete content.

Artificial intelligence technology was not able to help cover the summit well (Al Jazeera)

Methodology for covering summit sessions with artificial intelligence

In the plan that I developed, the number of sessions that I decided to cover with the help of artificial intelligence per day was five sessions, making a total of 15 sessions in the three days of the conference.

Although the normal rate for a reporter is two to three sessions, the number I set was a great challenge, as the average reporter cannot enter these sessions and leave to write about five different topics and publish them in the same day.

To do this, I relied on a specific methodology and tools that I had previously used in similar operations. The methodology consisted of the following steps:

  • I first linked the GPT chat program to an Excel file divided into three portals: the full article, a classification table within the article, and materials for social media platforms.

  • The “full material” was the audio session material converted into text by an audio recording application on the phone that uses artificial intelligence to convert the audio material into written text.

  • I set a command in the GPT chat that materials that are transcribed as text are immediately translated into Arabic if the text is in English. In this way, I guarantee that all material texts are in Arabic.

  • After that, another command passes these materials to the second stage of work, which is classifying the content of the material. I placed a command in the GBT chat to create a table that reads the downloaded material and filters it into several sections:


    Facts: where it places sentences that are considered facts in the material or numbers in Special column.


    Information: It places sentences that contain information in another column in the table.


    Style: He places sentences that contain the speaker's style, simile, or instruction in a third column in the table.


    Miscellaneous: The rest of the sentences that are not related to the topic, such as incomprehensible speech, comments, or questions from speakers.

As for the third stage, I wrote a command that takes the two columns of information and facts in the second stage and turns them into materials suitable for publication on social media platforms.

A screenshot showing the process of converting an audio recording into text and translating it on GPT chat (Al Jazeera)

Mechanism of Action

The goal of this methodology was to help me record the session, transcribe its content into several sections, and then create specific templates based on which the content could be categorized.

The working mechanism was as follows:

  • The first session of the audio recording application is recorded and converted into text, and the text is transcribed into the table using GPT chat commands, and then filling the columns with the appropriate content according to the classification using artificial intelligence, and then taking the appropriate facts and information from the columns and producing publications that are appropriate for communication platforms to send. For the communication platforms team.

  • In the second session, while the audio recording application is recording the second session, I work on writing a report on the material of the first session and uploading it to the system, after reading the transcribed and classified content, and so on in each session.

Of course, I knew that there were problems with estimating the appropriate time and the ability to work at the time of the session, or even the ability to enter the scheduled session at the appropriate time.

I took all these logistical things into consideration and I wasn't trying to break a record as much as I wanted to prove that the system worked well even if the result was only three items or less.

On conference day, technology is honored or disgraced

Yes, I took everything into consideration, and I tested the system before the conference in a short article, and it largely succeeded in doing 80% of what was required.

I made amendments to the material, but the surprise came at the beginning of the conference.

Firstly, the application that I used many times to convert audio to text;

By recording the entire session, it was less than 45 minutes, but when trying to convert the session to text, the application only converted the first 20 minutes, even though the audio recording was complete.

This led me to try several applications that convert audio to text (I will mention them later), and in the end the Kkato website was able to transcribe the audio content and convert it to text.

Now came the second step, which was transcribing the textual content into the table, and here some problems with classification appeared. The material was long, and the classification was accurate, so there was a number of information that was not classified from the original material, and I had to read the original material to find out what it was, and sometimes there was information. Incomplete and not posted to the information column.

The step that artificial intelligence can implement well is to transform the information and facts that it was able to transfer from the original text into materials that can be published on communication platforms.

Artificial intelligence was also able to help me prepare for meetings very well, as I would ask for a summary of the person and his experiences, and I would ask to prepare questions from several axes.

GBT Chat was able to accomplish this task very well and easily, and I do not think that meeting planners need more than this application.

Tools and applications you used

  • Recording audio and converting it to text (I tried several applications and websites, including


    the Recorder Pro application and


    the Vid application).

    IO


    Kkato website

  • gbt chat

  • Link Excel files to the GPT Chat application

An enjoyable experience despite the results

After the conference ended and I knew that artificial intelligence technology so far is insufficient in dealing with pressure and helping when you have long materials, and that analyzing and classifying them accurately is not an easy matter, the experience was useful in knowing the limits of artificial intelligence.

Perhaps in this edition of the 2024 Web Summit, artificial intelligence was not lucky enough to defeat humans in covering the summit’s activities well, but it is hoped that the 2028 edition, which will also be in Doha;

Artificial intelligence will play a major and important role in covering events in a distinguished manner.

Source: Al Jazeera