Expanding language-model capabilities beyond English

The Explosive Rise of Generative Artificial Intelligence

The year 2022 is marked in history as the period when society first encountered Generative Artificial Intelligence (GenAI): systems capable of producing human-quality content such as text, images, and computer code. These systems can also interact with the world in unprecedented ways, using natural human language across interfaces such as text messages, voice, and video. Among other capabilities, GenAI can read, process, and summarize content thousands of times faster than humans.

GenAI has the potential to radically transform the anatomy of work, enhancing the capabilities of professionals at all levels through large-scale automation of many of their activities. Currently, GenAI systems are produced by small groups of AI researchers and professionals around the world, especially in Silicon Valley, with most of them being employees of major technology companies (1).

Perhaps the main limitation to the emergence of new players in the field is the scarcity of researchers and professionals. This context may limit the impact of this technology on the market.

Challenges of Organizations and the Generative AI Market

The challenge we aim to address is the increasing inability of human organizations to cope with the speed of change, whether they are threats or opportunities.

We believe that GenAI applied to all layers of decision-making and operation of an organization can provide efficiency and productivity gains never seen before. New business models will emerge, and established value chain structures may undergo radical changes. A radical transfer of value should occur from established industries to players who master GenAI and create new business models from use cases of this technology.

The latest expectations suggest that GenAI could add around USD 4 trillion a year to the global economy(2). This is greater than the GDP of the United Kingdom and twice the GDP of Brazil.

If Generative AI fulfills its promises, the global labor market could face significant disruption, exposing the equivalent of 300 million full-time jobs to automation(3).

The global Generative AI market is expected to reach USD 109.37 billion by 2030, according to a new report from Grand View Research, Inc. The market is expected to expand at an annual rate of 35.6% from 2023 to 2030(4). The growing demand to modernize workflows across all industries is expected to drive the demand for Generative AI applications.

The Emergence of a New Oligopoly?

Generative AIs require a high level of technical sophistication based on years of academic and private sector research. Only in the year 2022, with the escalation of training these models with massive amounts of data and the refinement promoted by human feedback, have we reached the models now popularly known, such as OpenAI's ChatGPT and Alphabet's BARD. The central issue is that the construction of these models is still restricted to a few companies, due to the scarcity of qualified professionals and mainly to the closed model adopted by these companies. Microsoft, an investor and strategic partner of OpenAI, is adopting a mixed approach, offering an Open Source model in partnership with Meta.

Other players, like Hugging Face, offer Open Source alternatives, positioning themselves as an alternative to closed architecture models.

The large number of startups and other companies that have recently emerged to explore this market operate in building interfaces for the generative models of these few companies, with the majority functioning on top of "black-box" models and not in the creation and training of new models.

We believe that the greatest value to be captured in the emerging Generative AI market lies precisely in the creation and training of new generative models, especially with proprietary data from the end-users of these models.

With this vision, we are creating and training new models initially with the help of closed models and progressively migrating to specialized models with the aid of large-scale processing and a large amount of data with proven origin and quality. For this, an extremely qualified and experienced team is necessary, something that infusion ai is positioned to deliver.

With most actors currently based in the United States and some initiatives being launched in Europe, Brazil still does not have any serious candidate to occupy the national market.

‍

References

1. Two seminal references in the field: (i) "Generative Adversarial Networks" - Department of Computer Science and Operations Research at the University of Montreal [Link]and (ii) "Attention is All You Need" - Google/Alphabet [Link].

2. McKinsey Consulting estimates an impact of between $2.6 to $4.4 trillion per year, based solely on a group of 63 use cases studied [Link].

3. The Potentially Large Effects of Artificial Intelligence on Economic Growth (Briggs/Kodnani), Goldman Sachs [Link]

4. Generative AI Market Size To Reach $109.37 Billion By 2030 [Link]

‍