Synthesia, an artificial intelligence firm, presented a video on the social media site X showing lifelike avatars that can portray human emotions using text inputs. The London-based startup displayed "expressive avatars," showing pleasure, sadness, and irritation.

This comes amid worries over deepfake videos, where avatars impersonate people to commit fraud or propagate disinformation. According to the MIT Technology Review, Synthesia favors "synthetic media" over deepfake to avoid negative connotations.

The Future of Video Production

Nvidia-backed Synthesia aims to simplify video creation by eliminating performers, editing, cameras, and costs. Real actors perform scripts in front of a green screen to train AI-powered avatars. Notably, Amazon Web Services used the technology to create an avatar of managing director Tanuja Randery for a Wednesday event, per Euronews.

According to Synthesia, 55,000 businesses use their technology to create corporate avatars. The organization uses Know Your Client to verify identification and prevent false information. It also promotes watermarking in the Coalition for Content Provenance and Authenticity.

Moreover, the company noted that the AI videos generated provide agile and consistent internal messaging and training. The AI avatars engaging style can be readily updated, translated, and personalized, fostering a more informed, competent, and motivated workforce that drives the organization's learning efforts and productivity.

In 2020, Synthesia launched its first business AI avatars after focusing on dubbing and lip-synching technologies since 2017. It became a unicorn last year, topping €1 billion ($1.07 billion).

Read Also: UK Competition Regulators to Investigate Microsoft and Amazon AI Deals 

Business Industry Adopting to AI Tech

Amid the continued advancements in the AI field, a recent study noted how the business industry is adopting the technology.

TechTimes previously reported that most CEOs use generative AI (GenAI), according to INSEAD research, dispelling concerns about job displacement from the technology. Over 1,200 INSEAD graduates from various businesses and locations participated in the poll, which revealed AI adoption opinions.

Despite fears about AI undermining traditional job arrangements, the study shows that employees and business executives are excited about GenAI's revolutionary potential.

Two-thirds of respondents use GenAI in their personal and professional lives, demonstrating its widespread acceptance. In particular, the study shows a shift from job displacement to AI abuse fears.

INSEAD Associate Professor Jason P. Davis emphasizes the survey's usefulness in understanding executives' views on AI technology and its trajectory across sectors and locations.

The worldwide INSEAD alumni sample gives a complete picture of GenAI adoption attitudes. Over half of respondents use GenAI in their businesses, demonstrating the widespread use of generative technologies.

A significant majority revealed plans to include GenAI soon, indicating an increasing AI adoption trend. However, Europe was more skeptical of new technologies and concerned about digital privacy when it came to AI adoption.

In another development in business AI, Microsoft unveiled Phi-3-mini, a "significantly cheaper" artificial intelligence model for smaller tasks, targeting enterprises with few resources. As the IT giant bets its future on a technology that will transform global operations, Phi-3-mini is its first small language model (SLM).
Microsoft's vice president of GenAI research, Sébastien Bubeck, said the Phi-3-mini is more affordable than previous models.

According to Microsoft, the Phi-3 family of open models is the most efficient and cost-effective little language model. Large-language models remain the dominant AI efforts. The tech giant noted that Phi-3 models surpass similar-sized counterparts in language, coding, and arithmetic benchmarks.

The Phi-3-mini model is available on Hugging Face, Ollama, and the Microsoft Azure AI Model Catalog. It will also be available as a standard API-accessible NVIDIA NIM microservice in a variety of contexts.

Microsoft also announced the sale of more Phi-3 family products, extending the cost and quality possibilities. Phi-3-medium (14 billion parameters) and Phi-3-small (7 billion parameters) will soon be available in model repositories and the Azure AI Model Catalog.

Related Article: AI Companies Join Forces to Ensure Child Safety Principles in Technologies