
OpenAI has made an announcement that ChatGPT is currently in the process of rolling out to Android users this week, and this comes two months after the artificial intelligence (AI) chatbot launched on iOS.
Android users are able to pre-register for the app on the Google Play Store, and while the app is free, pre-registration will guarantee that it will be installed on your device as soon as it goes live.
Taking to Twitter, OpenAI wrote: “Announcing ChatGPT for Android! The app will be rolling out to users next week, and you can pre-order in the Google Play Store starting today.”
The Google Play Store listing makes sure to emphasize that the app is free and will sync your history across your devices and keep up to date with OpenAI’s latest model improvements.
At the moment, Android users are able to gain access ChatGPT through a browser or Microsoft’s ChatGPT-4-powered Bing app. With that being said, an independent app will likely be beneficial for users.
TechCrunch reported that the app had accumulated over half a million downloads within six days of its App Store launch, so it proved popular for iOS users.
Powered by by OpenAI’s GPT-4 large language model, researchers have raised concerns over a deterioration in ChatGPT’s response to prompts. It was a team of three researchers from Stanford and UC Berkeley that measured the qualitative aspects of GPT-4 and its predecessor, GPT-3.5, across four main categories:
- Solving maths problems
- Generating code
- Answering sensitive questions
- Visual reasoning
- The team only noticed improvements in the chatbot’s visual reasoning, with GPT-4 showing noticeable declines in the other three categories.
Matei Zaharia who is one of the researchers, said it was “very hard” to determine the cause of the deterioration. “It could definitely be that RLHF [reinforcement learning from human feedback] and fine-tuning are hitting a wall, but might also be bugs. Definitely seems tricky to manage quality,” Zaharia added.
The team has plans in place for running a longer follow-up study, and Zaharia asked his Twitter followers for feedback on which questions they should use to measure GPT’s performance.
