A flurry of AI releases now: GPT-4 and new educational projects

Bryan Alexander 2023-03-15

The generative AI world continues to develop at a very fast pace.  I’m buried in project work right now, but wanted to quickly note some developments.

The big announcement today was OpenAI’s launch of GPT-4, the next version of its large learning model.  4 draws on a larger training set and can also write responses to user-uploaded images.

Note how OpenAI uses higher education tests to demonstrate the software’s improvement, version by version:

GPT3 35 and 4 on tests_OpenAI 2023 March 14

Currently it’s only available to ChatGPT+ customers.  I became one for this purpose, and asked the bot a question: “How are you different from ChatGPT’s previous version?”

The reply:

As an AI language model, I am a constantly evolving and improving entity based on the advancements in natural language processing and machine learning. Although I’m presented as ChatGPT, I am based on the GPT-4 architecture, which is an advanced version of the previous GPT-3 model. While specific differences may not be mentioned explicitly, the improvements between the versions typically include:

  1. Larger training dataset: As newer versions are developed, they usually benefit from an increased amount of data used in the training process. This enables me to have a broader understanding of various topics and generate more accurate and contextually relevant responses.
  2. Improved language understanding: With advancements in the underlying architecture, I can better understand and process complex sentences, nuances, and contextual information. This allows me to generate more coherent and contextually accurate responses.
  3. Enhanced conversational skills: As the model evolves, my ability to maintain longer and more meaningful conversations improves. This includes better context retention and the capacity to address more complex queries effectively.
  4. Reduced bias and harmful outputs: The development process aims to minimize the biases and harmful outputs that might arise from the model. This involves refining the training data and improving the fine-tuning process to create a more responsible and safe AI.

Please note that while I am an improvement over previous models, I am still not perfect. There may be instances where I generate incorrect or nonsensical answers. If you encounter such issues, please feel free to correct me or ask for clarification.

I’m exploring it.

Meanwhile, a bunch of educational projects using the ChatGPT API have rolled out.  Khan Academy announced Khanmigo, a putative tutor:

At almost the same time Grammarly announced GrammarlyGo, a ChatGPT-powered tool to help you write.  It’s somewhere between a tutor and coauthor:

What other educational applications have you seen which use ChatGPT’s API?

(thanks to T.H.E. for one link)