Artificial intelligence (AI) has become an increasingly important topic in everyday life. As technology has evolved, we have seen the creation of various forms of AI, each with its own functionality. Among these is generative AI. In this article, we will explore what generative AI is, its history, its limitations and the impact of this technology.

Generative AI is a form of AI that uses artificial neural networks to generate original content from existing data. It is capable of producing a wide variety of content such as images, text, music, video, and even computer programs. Unlike other forms of AI that need a massive training dataset to function, generative AI is able to create original content with very little data.

The technical foundations

It all started in 1952 with the invention of Machine Learning, followed by the introduction of AI in 1956. Over the decades, the computing power and amount of data increased, leading to the emergence of Deep Learning in 2012.

Deep Learning allows a machine to learn from data without being explicitly programmed to perform a specific task. In other words, Deep Learning allows machines to learn from large amounts of data, using neural networks that simulate the functioning of the human brain.

Generative AI is a branch of deep learning and makes it possible to create original content. This technology has revolutionized AI by allowing computers to learn from raw data.

Next, Transformers were introduced in 2017, offering a new method for natural language understanding - leading to significant advances in machine translation and text generation. They use natural language processing techniques commonly known as NLP(Natural Language Processing in English), including the attention mechanism, to understand meaning. For example, GPT (Generative Pre-trained Transformer) is the generative AI model developed by OpenAI using Transformers.

ChatGPT and the AI market

ChatGPT

ChatGPT is a natural language model based on the GPT-3.5 architecture developed by OpenAI and will switch to the GPT-4 architecture, which is already available on their interface. It was launched in November 2022 and is considered one of the most successful language models currently available with currently over 100 million users, setting a record for the fastest and most massive growth of its user base in just 2 months after its official launch!

ChatGPT has been trained on a huge amount of text data from various sources, allowing it to generate text continuously and consistently. It can be used for a variety of applications, including virtual assistance, text generation, machine translation and more. ChatGPT is also able to continuously learn and improve its performance over time. Here's why ChatGPT was a highlight: 

• AI is now usable by the general public.
• Costs go down, AI becomes a commodity.
• The real revolution: the user experience.

Current biases and limitations of ChatGPT

Like any generative AI, ChatGPT has limitations: inconsistency, repetition or preference for frequent data. It may have difficulty producing consistent, long-term text. In addition, it can be easily fooled by malicious or poorly formatted input data.

It is important to note that ChatGPT was trained on data prior to 2021 and does not have access to the internet, which may limit its ability to produce relevant and timely content.

Currently, ChatGPT has implemented in Alpha for some users a plug-in that allows the artificial intelligence to work with current data.

However, for those who do not have access to this plug-in, there is WebChatGPT, a chrome extension that allows you to do the same thing. This allows to have safe and sourced information.

The different types of generative AI and ChatGPT competitors

Faced with Microsoft's ChatGPT integration into Bing and Teams, competitors are forced to react.

Google announces the launch of its own AI, called Bard, in February 2023. Bard is designed to be able to generate and explain code, which sets it apart from other chatbots on the market. Bard is based on an autonomous language model, which uses Machine Learning to understand and produce natural language responses.

Meta (formerly Facebook) launched its own autonomous language model called LLaMa also in February 2023. Compared to Bard, LLaMa was not launched as a public chatbot, but rather as an open source package.

China and other Nations, such as Israel, have also invested in AI and chatbots. Baidu, in particular, has developed several chatbots for different applications, including healthcare and customer support. Tencent, another Chinese company, has created a chatbot called Xiaowei for reservations and ticket purchases, while Israel has developed a military chatbot called Tzayad.

Secondly, there are generative AIs capable of creating images from prompts (texts entered by the user). For example, Midjourney, the direct competitor of DALL-E generates high quality images through the Discord platform.

However, OpenAI does not cover all types of generative AI, none of the firm's products can transform text into audio or video as you can see in the table:

‍

Impact of Artificial Intelligence

Impact on jobs

Generative AI presents both opportunities for improving employee efficiency as well as the creation of new jobs, but also threats of job destruction.

They enable large-scale automation of repetitive tasks, improved efficiency and personalization of the customer experience, which can lead to better customer satisfaction, employee satisfaction and business growth.

On the flip side, that's about 300 million jobs worldwide that could be impacted by AI and automation according to a recent report by Goldman Sachs in a CNBC article.

Impact on the environment

The impact of AI on the environment is a complex topic that is still being studied and debated by researchers and experts. They point out that the energy consumption and carbon footprint associated with AI development and deployment can contribute to environmental degradation.

According to a recent MIT Technology report in a Forbes article, training a single AI model can emit more than 626,000 pounds of carbon dioxide equivalent, nearly five times the greenhouse gas emissions produced by an average U.S. car over its lifetime.

Impact on the protection of personal data

First of all, there is a lack of information about data processing. Although OpenAI specifies the nature of the data processed in their legal notice, they remain unclear about the purpose of their service and the applicable legal basis.

Recently, Italy asked OpenAI to suspend access to ChatGPT on its territory due to a lack of respect for personal data protection. 

After Italy, Samsung Electronics also banned the use of generative AI such as ChatGPT within its company because generating responses on ChatGPT could result in the disclosure of sensitive information.

Impact on copyright

ChatGPT users face the problem of not being able to prohibit copying of their AI-generated content by a third party on copyright grounds like any other generative AI.

The issue of copyright is therefore very complex and is the subject of much debate.

Over time, the use of generative AI can limit creativity and encourage conformity, which can lead to a standardization of the content produced.

Conclusion

Generative AI is a technology that has grown exponentially over the years. With applications in various domains such as text, image, music and video generation, it offers incredible opportunities to improve efficiency and customer experience. However, it also presents challenges, including bias, technological limitations and security issues.