How ChatGPT’s new search capabilities will pave the way for AI agents

The future of AI assistants: OpenAI executives Olivier Godement and Romain Huet outline their vision for AI agents, which are advanced AI assistants capable of completing complex tasks and acting as personal chief of staff.

• Godement envisions a future where every person and business has an AI agent that knows their preferences intimately and can access various digital tools to complete tasks.
• OpenAI’s strategy involves both developing its own agents and providing tools for developers to create their own, with voice interaction playing a crucial role in the user experience.
• The company is working on making these AI assistants more reliable, faster, and cost-effective in the short term, while expanding their capabilities to fields such as law, accounting, and economics in the long term.

Key developments paving the way: Recent announcements from OpenAI demonstrate progress towards more capable AI agents, including updates to their voice technology and the introduction of ChatGPT search.

• The Realtime API platform now offers new voices and a prompt generation function, enabling developers to build more sophisticated apps and voice assistants.
• ChatGPT search allows users to access internet information directly through the chatbot, enhancing its ability to provide up-to-date and relevant information.
• These advancements are seen as stepping stones towards creating AI agents that can interact with various digital tools and complete complex tasks autonomously.

Challenges to overcome: Godement identifies two major hurdles that need to be addressed before AI agents can become a reality: reasoning capabilities and the ability to connect different tools.

• OpenAI is working on improving AI reasoning through their “chain of thought” approach, which allows models to recognize mistakes, break down problems, and try different solutions.
• However, experts like Chirag Shah caution that current AI models may be mimicking reasoning rather than truly understanding, highlighting the need for further development.
• The ability to connect and use various digital tools is crucial for AI agents to perform complex tasks, with OpenAI’s o1 model showing some progress in this area, though reliability remains a challenge.

Expanding AI capabilities: OpenAI is focusing on broadening the range of tasks AI can perform and improving its ability to interact with the real world.

• The company aims to apply its chain-of-thought technique to fields beyond science, coding, and math, such as law and economics.
• Features like ChatGPT search enable AI models to access up-to-date information beyond their training data.
• OpenAI is also working on enabling AI to interact with computer interfaces, similar to Anthropic’s Claude chatbot, which can control a computer by clicking on interface elements.

Near-term expectations and surprises: While Godement anticipates growth in AI adoption for customer support and assistant-based tasks, he acknowledges the unpredictability of how people will use the technology.

• The executive expects significant adoption of AI in customer support roles within the next year.
• However, Godement also notes that past developments have often surprised the company, with unexpected use cases emerging.
• This uncertainty highlights the potential for AI to be applied in novel and unforeseen ways as the technology continues to evolve.

Analyzing the implications: As AI agents become more sophisticated, they have the potential to dramatically reshape how individuals and businesses interact with technology and manage their daily tasks.

• The development of AI agents raises important questions about privacy, as these assistants would have access to vast amounts of personal data.
• The ability of AI to reason and make decisions autonomously also brings ethical considerations to the forefront, particularly in sensitive fields like law and finance.
• While the promise of AI agents is exciting, their widespread adoption will likely require significant advancements in both technology and public trust, as well as careful consideration of their societal impact.