Roughly Half Of Nations Investing In AI, Develop Generative Models In-House: Report
Of the 62 nations, roughly half (35) developed their gen AI solutions in-house, with around 90% focusing on one model type.
Roughly half of the nation develop their Gen AI solution in-house with nine in 10 focusing on one model type, according to a report.
AIport, an online community of artificial intelligence writers, researchers and data scientists, has released the first edition of its Global Generative AI Landscape 2024.
The analysis emphasised regional attributes and showed which players are developing gen AI solutions, their locations and the nature of their contributions. It contains a total of 128 generative AI models from 107 companies.
The report examined all 62 countries that invest in the AI market, as featured in the Global AI Index by Tortoise. Of the 62 nations, roughly half (35) developed their gen AI solutions in-house, with around 90% focusing on one model type.
Key Highlights
Among the top five nations in the Index, South Korea and Singapore were absent from the list of top gen AI developers, suggesting that these nations prioritise investment in other AI domains.
Regional leaders by the number of active gen AI companies are North America: United States; South America: Argentina; Europe: UK and France; Asia: China and Israel; Oceania: Australia and New Zealand; and Africa: South Africa.
The average number of gen AI models per company was the highest in North America, being the only region to have at least one model from each of the 10 model categories.
Approximately 10% of all companies covered in the study have implemented multimodality in their gen AI models, with a majority of these developers located in the US. This indicates that while multimodality represents an emerging trend, its adoption outside North America still remains in the nascent stages.
Eleven companies globally have developed more than one type of gen AI model. Stability AI leads with five distinct gen AI model types (image, video, audio, 3D and code), followed closely by OpenAI (chatbot, audio, video and multimodal) and Google (text, image, audio and multimodal)—both with four model types.
Microsoft, Meta, Tencent, Baidu and Yandex are among those companies that developed between two to three types of distinct gen AI models.
Thirteen companies have developed multiple models within a single gen AI category. AssemblyAI has two speech-to-text models, MosaicML offers two iterations of its MPT for code generation, while IPOXCap has introduced two chatbots designed for business intelligence applications.
“Many generative AI landscapes tend to focus either on the Silicon Valley giants or the tech powerhouses of Europe, covering no more than 10 countries on average," said Avi Chawla, data scientist and community manager at AIport. "We believe Volume 1 of our Global Generative AI Landscape 2024 provides an objectively international outlook.”
