OpenAI Was the Biggest Disrupter. Now, That Could Change
Few people have upended the technology world in as dramatic a fashion as Sam Altman, CEO of OpenAI, which in November 2022 launched the first real form of generative artificial intelligence and large language models with ChatGPT. Speech Technology readers even named Altman as the most apt for induction into a hypothetical Hall of Fame for speech technologies, as Jim Larson highlights in this issue’s Forward Thinking column, “ChatGPT: A Generative AI Revolution.”
Generative AI tools like ChatGPT mark a significant shift in how users interact with computers, Larson asserts, noting that unlike traditional methods requiring complex commands, ChatGPT allows users to ask questions and receive answers in natural language. “Generative AI, pioneered by Sam Altman and others, is transforming how we interact with computers and promises to revolutionize various industries,” Larson concludes.
He goes on to say, though, that “as technology evolves, we can expect even more innovative applications and a future where AI seamlessly integrates into our daily lives.”
Nothing could have been more prophetic. Just days after Larson finished penning his column, Chinese AI lab DeepSeek launched an open-source large language model that immediately sent ripples through the tech world and raised questions about the future of AI. President Donald Trump called DeepSeek a “wake-up call” for U.S. tech companies who must now focus on “competing to win.”
So why all the attention? After all, DeepSeek’s AI-powered chatbot, dubbed R1, reportedly looks, feels, and works very much like ChatGPT, Anthropic’s Claude, Meta’s LLAMA, Google Gemini, and similar technologies; will be used for many of the same tasks; and is about as powerful.
Quite simply, DeepSeek claims that its model uses less memory than its rivals and delivers results at a much, much lower cost than the other models. And the DeepSeek R1 model’s open-source roots put it in a great position to democratize AI chatbot development and access to AI.
“The hubbub surrounding DeepSeek, in a nutshell, is that the company proved a point that you don’t need a gazillion dollars to train an AI model,” Constellation Research analyst Andy Thurai said in a research note.
Shailesh Nalawadi, head of product at Sendbird, a provider of a messaging-as-a-service API for chat, voice, and video messaging, shares that belief. “The launch of DeepSeek R1 highlights a fundamental shift in the AI industry: Our previous assumptions about the cost of training and inference for large language models were wrong. Costs are declining faster than anticipated, reshaping the competitive landscape and accelerating AI adoption,” he says in an email.
DeepSeek’s impact on the speech technology world could be huge, according to Nalawadi. “As inference costs drop, we can make more liberal use of LLMs to develop increasingly sophisticated AI agents,” he states. “This allows us to enhance conversational experiences, improve automation, and drive deeper personalization for our customers. The ability to experiment more aggressively and iterate faster will push the boundaries of what’s possible in AI-powered chat and engagement solutions.”
Others maintain that DeepSeek’s model could move applications away from the cloud and into more private environments, even onto devices like phones and smart speakers.
Nalawadi further expects DeepSeek’s model to lead to “more powerful, accurate, and scalable AI-driven applications.”
That trend is already under way. In its first week, DeepSeek’s AI assistant became the most downloaded free app in the Apple App Store (seeing more than 1.9 million downloads) and the Google Play Store (having been downloaded more than 1.2 million times).
DeepSeek has clearly disrupted the market, but it still should be treated with care. Stemming from DeepSeek’s Chinese origins, there might be very legitimate concerns about data privacy and security.
While businesses might be inclined to jump on board the DeepSeek train quickly to gain a competitive edge, they will be better served by going slow, thinking through the use cases, the problems they’d like to solve with it, and the data that it will be able to access.
Leonard Klie is the editor of Speech Technology magazine. He can be reached at lklie@infotoday.com.
Related Articles
The next waves of innovation in the speech technology space promise to be more transformative than ever before.
06 Mar 2025