Extra update - version 14/11/2023 released
Summary
OpenAI are having reliability issues, which means that Assessment Bot was having reliability issues.
We've added an extra AI provider to address this.
We've made a few under-the-hood changes to make it easier and quicker to update things like this in the future.
Reliability issues fixed
Ordinarily, we wouldn't release another version so soon after the last. OpenAI, the company who provide the AI backend for Assessment Bot, have had issues with cyberattacks and warned that their service may be unreliable over the next few weeks while they deal with the demand created with all the new technologies they announced recently.
This has meant that Assessment Bot has not been as reliable as it should be.
To improve things, we've signed up with an additional AI provider to allow Assessment Bot to work as it should. Unfortunately, they don't support the latest model we got so excited about in the previous release which means that you wont benefit from the improved marking accuracy. You will benefit from much faster response times, however. This will be the default model because we think it's more important that Assessment Bot works than you getting more accurate results, some of the time.
You can still use the latest ChatGPT model (gpt-3.5-turbo-1106) but it may not work reliably. We've set it to run much slower than the default model (gpt-3.5-turbo) to minimise the chance of errors.
Under the hood changes
We've made a number of under-the-hood changes that won't impact you at all, but will mean that we can make certain changes much faster in the future.
Bug fixes
The dropdown menu now works when you choose the model in the 'Options' sheet.