Picture a ChatGPT that remembers every conversation you've ever had with it, complete with in-depth knowledge about your life, work, preferences, and more. One that provides personalized responses tailored specifically to you, each and every time.
Recently, researchers published a groundbreaking study detailing their efforts to develop an AI model capable of handling 1-2M+ tokens. This is the kind of research that could once again be a fundamental disruption in the world of AI.
To put things in perspective, ChatGPT's current memory is limited. When utilizing GPT-4, it has a context window of around 8,000 tokens, which corresponds to approximately 12 pages of text. This means the AI's capacity to remember your past exchanges is bound by this limitation.
Now, envision a ChatGPT with a context window encompassing 1 to 2 million tokens. Such a model could have the capacity to comprehend entire codebases, absorb your full business knowledge base, or even possess an intimate understanding of you on a personal basis. ChatGPT could potentially equal your closest friend in terms of understanding you, or possess as much insight into your company as the CEO.
The key to the success of AI lies in the quality of data and inputs you’re able to feed it. The more context we provide, the better its performance. If a ChatGPT response is off, it's rarely due to a deficiency in reasoning ability, but rather the limitations of the given context and the difficulty in identifying the optimal solution.
This area of AI is moving fast, and this research paper offers a glimpse of what the future might hold. As the context window expands, the output generated by AI models will become increasingly powerful. This is definitely a development to keep a close eye on.