The retained GPTOnline cover uses a circuit-brain and hand illustration for this historical story; it is not a product-interface screenshot.
The world of artificial intelligence has been shaken to its core with the release and subsequent rapid evolution of OpenAI’s ChatGPT. In just a year since its unveiling, this technology has made strides that have left both experts and enthusiasts in awe.
When ChatGPT was first launched on November 30, 2022, it was a text-only platform that could answer questions based on training data that only extended up until September 2021. It was an impressive feat for sure, but it was not without its limitations. For one, it had a notable propensity for “hallucination” – a term coined to describe the AI’s tendency to fabricate facts when it found itself in uncertain territory. This occasional lapse in accuracy did little to deter users, however. Within the first five days of its release, over a million people had already logged on to interact with ChatGPT. The tech community watched in wonder as this new AI tool quickly became a sensation.
By November 2023, ChatGPT had gained several capabilities, depending on the plan and model. OpenAI’s GPT-4 Turbo announcement specified knowledge through April 2023, while web browsing could retrieve newer information. Retrieval did not change the model’s training cutoff or ensure that every answer was current. Furthermore, it has evolved into a multimodal AI, capable of processing images or documents in addition to text, and even communicating via spoken language.
The recent DevDay event hosted by OpenAI introduced another exciting development in the realm of generative AI: the ability to create customized GPTs. These GPTs combined instructions, supplied knowledge and selected capabilities for different tasks; creating one did not necessarily mean training a new underlying model. From automating website creation to generating other custom GPTs, the applications seem endless.
But the potential of generative AI extends far beyond ChatGPT. The technology is being applied in more specific domains, including law, medicine, and climate change adaptation. While a generic tool like ChatGPT might not be the best fit for complex tasks such as drafting a legal brief or diagnosing medical conditions, domain-specific systems can be designed for specialized workflows, but their accuracy and reliability still require evaluation. Customization alone does not make a system suitable for consequential professional decisions.
Generative AI has also made a significant impact on the world of computer coding. It can assist novices and experienced programmers with code, explanations and debugging, but generated code still needs testing and the user still needs to understand its behavior. Large language models serve as natural language interfaces to complex programs, enabling users to tap into the full potential of comprehensive tools like Photoshop, Word, and Excel.
Image generation was already part of this first-year story: OpenAI announced DALL·E 3 availability in ChatGPT Plus and Enterprise on October 19, 2023. Users could describe images conversationally, although generated results still needed review and were not guaranteed to match every detail. Further improvements remained a possibility.
However, with all these advancements, it’s essential to exercise caution. Human oversight and permission boundaries are important when AI tools take actions; the safeguards depend on the particular product and configuration. The power of AI is immense, and unchecked autonomy could lead to unintended consequences.
As we look to the future, it’s clear that the evolutionary cycle of ChatGPT and similar technologies is set to accelerate. As AI engines begin to interact with each other and take independent actions, we’re likely to see an explosion of opportunities – and challenges. However, with careful management and responsible use, the potential benefits of these advancements in AI could be truly revolutionary.
Historical context: this article discusses ChatGPT’s first year, ending in November 2023, rather than today’s plans or feature availability. Sources: November 2023 GPTs announcement; GPT-4 Turbo and DevDay announcements; October 2023 DALL·E 3 availability.
