COMPETITION IS RAGING AFTER CHATGPT GEMINI WINS BUT HOW FAR ?

The first wave of enthusiasm generated by generative artificial intelligence (AI) was unprecedented. Just two months after its launch in November 2022, ChatGPT already had 100 million users. Internet searches for “artificial intelligence” have exploded; more than $40 billion of venture capital was invested in AI companies in the first half of this year alone.

Towards a second phase of development

However, a second, more serious phase is currently underway. A whole new industry, based on super-powerful AI models, is starting to take shape. Three strategic elements will determine its future shape and indicate whether OpenAI will maintain its dominance or whether other players will take the lead.

The first element is computing power, which represents a high cost, incentivizing creators of algorithmic models to be more efficient. Due to the exorbitant costs of training and running more powerful models, OpenAI is currently focusing on GPT-4.5 rather than the next big model, GPT-5. This factor could allow financially strong competitors, such as Google, to catch up. Google's state-of-the-art model, Gemini, appears more powerful than the current version of OpenAI's model.

High technology costs have also led to the proliferation of much smaller models, specialized for specific tasks. Startups such as Replite focus on computer coding to help developers write programs. Open source models also allow new players to enter the field of generative AI. According to Hugging Face, an AI company, around 1,500 versions of these highly targeted models exist.
In search of data… and profitability
The second force shaping the generative AI market is the quest for data. Large models, such as those from OpenAI and Google, are voracious, having been trained on more than 1 trillion words. , the equivalent of more than 250 Wikipedias in English. As they become more important, they will need to be fed even more. However, the limits of the Internet are almost being reached, prompting model creators to sign contracts with news and photo agencies, create synthetic training data, and explore new forms of data, such as videos.

The third crucial force is money. Many creators of algorithmic models are already turning to more profitable applications than consumer bots like ChatGPT. OpenAI has evolved as a non-profit company towards more profitable models, bespoke services. OpenAI, which was founded in 2015 as a non-profit company, has not only licensed its models to Microsoft, but also builds tailor-made tools for companies. such as Morgan Stanley and Salesforce. Abu Dhabi plans to create a company to help commercialize the services of Falcon, its open source AI model.

Other approaches include attracting software developers by offering tools to create products based on their models. OpenAI offers tools to help them create products using its models; Meta, Facebook's company, hopes that LlamA, its open source model, will help create a loyal community of programmers.

Although companies like OpenAI and Google, with their large user bases and considerable financial resources, have an initial advantage, finding effective solutions remains a major challenge. The model designer who adopts the most effective approach, the most ingenious method of synthesizing data or the most convincing argument for potential customers will have every chance of success.. .

Wishing that the AI ​​does not slip up very quickly !!....