The Future of Generative AI: Slowing Performance and Increasing Competition

Advances in generative artificial intelligence may be hitting a plateau as large language models face challenges and competition intensifies.

Generative artificial intelligence (AI) has captured the corporate world’s imagination with its ability to produce and summarize text. However, the rapid progress in large language models (LLMs) that power generative AI may be slowing down. Companies like OpenAI, Google, Cohere, and Anthropic, which have been at the forefront of this technology, are now facing increasing competition in the field. Moreover, the lack of uniqueness in the technology and the commoditization of LLMs are raising questions about the future of generative AI.

The Rise and Challenges of LLMs:

San Francisco-based OpenAI made waves in November 2022 when it released ChatGPT, powered by a large language model. Earlier versions of LLMs produced incoherent text, but today’s models are impressively fluent. Google’s release of its Gemini suite of LLMs in December showcased the challenges in making further progress. While Google’s Gemini model outperformed OpenAI’s GPT-4 in some measures, the margin was minimal. This suggests that LLMs may become commoditized, lacking differentiation from their competitors.

However, challenges persist in the field of generative AI. LLMs have a tendency to hallucinate and make things up, and companies using generative AI face legal issues related to training on copyrighted material. Striking licensing deals with content providers can mitigate this problem but may impact profit margins. As a result, some experts believe that companies in this space may be overvalued and anticipate a recalibration in the coming years.

The Plateau of Performance:

The advancement of LLMs has largely relied on scale, combining vast amounts of training data with substantial computing power to build complex models with billions of parameters. However, increasing the number of nodes in these models is yielding diminishing returns. The efficiency and performance of LLMs are still improving, but at a slower pace than before. This slowdown in progress raises questions about the future trajectory of generative AI.

See also  The Rise of Electric Cars: From Machines to Gadgets

Increasing Competition and Open-Source Models:

Despite the potential plateau in performance, the number of LLMs available to corporate customers continues to grow. In addition to proprietary developers like OpenAI, there is a growing ecosystem of open-source LLMs that can be used for commercial purposes. New players, such as Mistral AI and Meta Platforms Inc., are entering the market, further intensifying competition.

Companies using generative AI are no longer tied to a single provider, as switching from one LLM to another is relatively straightforward. Meta’s open-source push aims to prevent tech giants from monopolizing the market with proprietary technology, emphasizing the importance of commoditization. As a result, developers of LLMs will have to compete based on different attributes, such as privacy, security, and cost.

The Importance of Cost and Open-Source Models:

Cost is emerging as a crucial factor in the adoption of generative AI. Open-source models have the advantage of being more cost-effective, allowing companies to pass on savings to their customers. Many Canadian companies are opting for open-source models, leveraging the scale and speed to market they provide. However, a hybrid approach that combines different technologies is also expected to be adopted by companies.

Conclusion:

The future of generative AI is at a crossroads. While the performance gains in large language models may be slowing down, competition in the field is intensifying. The commoditization of LLMs and the availability of open-source models present challenges and opportunities for companies in the space. As the market evolves, organizations are advised to work with multiple players and consider factors such as privacy, security, and cost when adopting generative AI. The recalibration of the industry in the coming years will determine the true potential and impact of this transformative technology.

See also  California Explores AI Technology for Disaster Response: Experts Weigh In