A growing concern in the AI domain is the lack of collaborative efforts, which may exacerbate inequalities. Ashish Vaswani, CEO and co-founder of Essential AI, a San Francisco-based startup, warns that the pressures to monetize AI models—“difficult to ignore”—could lead companies to deprioritize long-term research initiatives.

Vaswani is known for his pioneering contributions to deep learning, notably co-authoring the transformative 2017 paper “Attention is All You Need.” This seminal work introduced the transformer architecture, laying the foundation for generative AI applications like ChatGPT and its successors.

After his time at Google Brain, Vaswani co-founded Adept AI Labs and later established Essential AI alongside Niki Parmar in 2023. The startup is backed by industry giants such as Google, Nvidia, and AMD, and aims to create full-stack AI solutions designed to enhance efficiency in automating labor-intensive tasks.

Despite billions of dollars invested in scaling AI models in recent years, many companies find themselves with minimal return on investment. During a recent interview with ET in India, Vaswani expressed concern that this reality could negatively impact AI research.

“Companies are increasingly shuttering long-term R&D efforts in favor of focusing resources on immediate profit-generating initiatives, especially during challenging times,” he explained.

Vaswani emphasized that leadership must prioritize sustainable business models alongside healthy long-term investments in research and development.

Further complicating the AI landscape, companies are at risk of sidelining important innovative ideas in the race for leadership. “In any race, existing ideas are often prioritized, while riskier, alternative paths may be overlooked,” he said.

The Importance of Transparency

Vaswani highlighted a troubling trend: the decline in openness, which is stifling idea advancement. “In 2017, numerous labs produced revolutionary ideas. Innovations were often collaborative, leading to nonlinear improvements. Today, this chance of shared progress is considerably diminished,” he noted.

The aspiration to achieve artificial general intelligence (AGI) is influencing current perspectives in AI development. “I don’t perceive AGI as a definitive endpoint but rather as an ongoing progression,” Vaswani elaborated.

Advocating for Open-Source Models

At Essential AI, Vaswani’s team is committed to pioneering an open science model. “We aim to build a frontier model that is openly accessible, allowing others to innovate on our work,” he stated. This includes sharing their AI models and training data to foster collaboration.

This open-source strategy is viewed as a strategic imperative. “When companies can observe our process, their trust in our models will naturally increase,” he said.

Regarding financial viability, he acknowledged, “Generating a return on investment with open-source models isn’t trivial, but keeping data closed doesn’t guarantee success either.”

Exploring Market Opportunities

Vaswani noted a favorable trend: potential collaborators and customers are enthusiastic about utilizing open-source models. “We see a myriad of business opportunities and aspire to be recognized like DeepSeek in the West—similar to a renowned Chinese AI startup known for innovative large language models,” he remarked.