ChatGPT just became more than a Bicycle for the Mind

Remember when computers (and software) were just bicycles for the mind? Attributed to Steve Jobs, he recognized the potential of computers to augment human intellect and capabilities, much like how bicycles increase the efficiency of human locomotion. This metaphor has since been widely referenced to describe the transformative impact of computers and software on human cognition and productivity.

In the domain of artificial intelligence, OpenAI has made a significant advancement with the introduction of a plugin system for its AI assistant, ChatGPT. This innovative development has the potential to revolutionize AI interactions with the global network by facilitating access to various network services, including flight booking, grocery ordering, and web browsing. As a result, the scope of AI applications has experienced a substantial expansion, paving the way for new opportunities and use cases in the field of artificial intelligence.

Traditionally, large language models like ChatGPT have been confined to their own bubble, capable only of producing text based on their training set and user input. Despite advancements in Bing Chat, which allows it to search the web for more recent information, ChatGPT has remained limited in its scope. The inherent drawback of this isolation is that the AI can often make factual errors or "hallucinations."

Now, OpenAI has developed a ChatGPT plugin interface, functioning as the AI's primary means of interaction with external services. This interface enables developers to design plugins that facilitate interaction between ChatGPT and a wide array of internet services, effectively enhancing its capabilities while mitigating the occurrence of hallucinations. The presentation of these plugins resembles that of an app store, featuring distinct icons and descriptions for each plugin, streamlining user selection and implementation.

The initial set of plugins have been created by a diverse group of companies, including Expedia, Instacart, Kayak, OpenTable, Shopify, and Wolfram, among others. Among the various plugins enhancing ChatGPT's capabilities, the Zapier plugin stands out as a significant addition, providing the AI with a robust interface to an extensive software automation system. As stated by Zapier, this integration enables ChatGPT to execute any of their 50,000 actions across more than 5,000 supported applications, effectively converting text-based interactions into actionable tasks.

The synergy between ChatGPT and Zapier's automation features allows for a wide range of tasks to be conducted efficiently. This includes the AI's ability to write and send emails, locate and modify contacts within a CRM system, or even insert data rows into a spreadsheet and forward them as Slack messages. The scope of this collaboration appears limitless, paving the way for innovative AI-driven solutions.

In the spirit of innovation and collaboration, OpenAI has chosen to refer to this selection process as a "plugin store," although no plans to sell individual plugins have been announced. Developers with early access to ChatGPT plugins have been quickly experimenting and creating new plugin prototypes. Notably, the process of developing ChatGPT plugins diverges from traditional approaches. Rather than relying on complex "glue code" to establish a connection between an API and ChatGPT, developers simply provide instructions in natural language, explaining how their service should be utilized by ChatGPT. The AI then takes these instructions and carries out the specified tasks, streamlining the plugin development process.

In addition to the streamlined development process, developers are leveraging ChatGPT and GPT-4 to create ChatGPT plugin manifests, which are machine-readable descriptions outlining the plugin's capabilities and invocation methods, as per OpenAI's definition. This further simplifies the development process, enhancing the efficiency of creating new plugins.

This new frontier of AI development has left some programmers wondering if they have reached the pinnacle of software innovation. A Twitter user by the name of Rohit expressed apprehension regarding the potential consequences of AI advancements, asking,

Guys. Existential crisis. Did OpenAI just finish software? What's there left to do but clean-up and sweep?

This inquiry underscores the possible ramifications of accelerated advancements in AI on the wider discipline of software engineering, as it brings forth debates about the future roles and responsibilities of developers in an increasingly AI-driven landscape.

Safety, of course, is a concern that accompanies any groundbreaking technology. OpenAI acknowledges the potential risks and has been transparent about the safety measures it is implementing. By gradually deploying access to plugins and applying rules to plugin usage, the company aims to mitigate the dangers posed by AI-powered harms, such as phishing rings, disinformation campaigns, and personal attacks.

As with any groundbreaking innovation, the introduction of ChatGPT plugins has raised questions and concerns that will need to be addressed as the technology continues to evolve. Currently available on an alpha basis to select developers and approved users, OpenAI plans to gradually roll out larger-scale access to the plugin system. As this artistic vision unfolds, the future of AI will undoubtedly be reshaped in ways we can only begin to imagine.