OpenAI’s AI-Powered Assistants Set to Hit Mainstream in 2025
OpenAI has announced that it anticipates its artificial intelligence (AI)-powered assistants will become widely adopted by next year, amidst a burgeoning race among tech giants, including Google and Apple, to introduce AI agents to the public.
AI agents are advanced systems capable of reasoning and executing intricate tasks, representing the latest battlefield for tech corporations vying to capitalise on this quickly advancing technology. Kevin Weil, OpenAI's Chief Product Officer, expressed his vision during the company's Developer Day event in San Francisco. "We want to make it possible to interact with AI in all of the ways that you interact with another human being," Weil stated. "These more agentic systems are going to become possible, and it is why I think 2025 is gonna be the year that agentic systems finally hit the mainstream."
During the Developer Day, OpenAI unveiled enhanced access to its new model series, termed o1, which boasts improved reasoning capabilities alongside GPT-4o's advanced voice functionalities. This will allow developers real-time access to technology that comprehends voice commands and facilitates live conversational interactions similar to a call.
The effort to mainstream AI agents is part of OpenAI’s strategy to drive future profits while transitioning to a for-profit structure. The startup is on the verge of completing a $6.5 billion funding round, valuing the company at $150 billion, by convincing investors of its potential to surpass rivals and lead in technological milestones. High-profile investors reportedly engaged in these discussions include Microsoft, Nvidia, SoftBank, and venture capital entities such as Thrive Capital and Tiger Global.
Concurrently, tech companies such as Microsoft, Salesforce, Workday, Google, and Meta have highlighted AI agents as central to their AI strategies. Although AI-enabled assistants have been in development for nearly a decade, recent advancements pave the way for smoother, natural voice interactions and superior comprehension powered by large language models (LLMs).
In the previous year, OpenAI launched the "assistants application programming interface" to let developers create AI agents using its technology. However, progress was initially restricted due to the capabilities of earlier models. OpenAI now claims that its latest models’ enhanced reasoning abilities will be evident in its offerings, including ChatGPT and tools utilised by startups and developers through its API. Nevertheless, OpenAI has not disclosed plans for its own AI agent development.
A specific example of these capabilities demonstrated during Developer Day included an AI system organising product purchases locally, following user instructions to place an order with a business based on parameters such as quantity and budget.
Although these interactions will not conceal the AI's non-human nature, the technology is limited to six preset voices for now to prevent use of unauthorised or customised voices. "If we do it right, it takes us to a world where we actually get to spend more time on the things that matter and a little less time staring at our phones," Weil noted.
Despite setbacks such as the recent resignations of Chief Technology Officer Mira Murati and Chief Research Officer Bob McGrew, OpenAI remains undeterred in its progress. Kevin Weil reassured that these departures will not hinder the company’s advancements: “Bob and Mira have been awesome leaders. I’ve learned a lot from them, and they are a huge part of getting us to where we are today,” he said. “And also, we’re not going to slow down.”
At the 2024 DevDay, OpenAI introduced several new tools including a public beta of its “Realtime API” for developers to build applications featuring low-latency, AI-generated voice responses. Demonstrations showcased applications such as a trip planning app that allows users to verbally interact with an AI assistant and receive prompt responses.
OpenAI’s Realtime API, which can integrate with calling APIs like Twilio, enables AI to engage in telephone interactions, though it does not automatically disclose its non-human identity. Developers are responsible for ensuring such disclosures, potentially in compliance with new laws like the one recently enacted in California.
Additionally, the company introduced vision fine-tuning within its API, empowering developers to enhance GPT-4o’s performance for visually oriented tasks by using images alongside text. This feature ensures adherence to OpenAI’s safety policies by restricting uploads of copyrighted or violent imagery.
Further, a model distillation feature was unveiled to allow developers to use larger models, such as o1-preview and GPT-4o, to fine-tune smaller, more cost-effective models. The beta evaluation tool included in this feature assists developers in measuring the performance of their fine-tuned models.
Missing from this year’s Developer Day announcements were updates on the GPT Store, first revealed at last year’s event. While OpenAI has been piloting a revenue-sharing program with prominent GPT creators, no substantial developments were disclosed.
Despite these omissions, OpenAI remains at the forefront of the AI landscape, with expectations high for further advancements and new model releases anticipated in the future.
Source: Noah Wire Services