Banshee is designed to supersede the functionality offered by Apple's existing voice technology.
In theory, right now iPhone users can use the built-in keyboard-dictation to enter text via voice. In practise, the experience is broken, the device usually fails to dictate at all and the designated UX is confusing. It does convert speech-to-text pretty fast (on the rare occasion when it works). In terms of its accuracy, it definitely seems to be at par with state-of-the-art.
iPhone also ships with Siri, a powerful voice-assistant. In practise, Siri hasn't worked for me for months upon months. It currently seems to be able to capture my request, but fails to process simple tasks, such as 'turn down the volume'.
iPhone and MacOS users can also enjoy the Voice Memo app. I use this frequently, throughout the week, to note down ideas. It has never crashed. However, it offers no transcription, so users are left to search through notes by listening to them in their entirety, or by reference to the note's title, which can be set by the user.
To summarise, the existing voice AI offered by Apple is unfortunate for the user, given the phenomenal power of speech-to-text (STT) as well as existing UI technology.
There is currently no 'Apple way' to transcribe notes, preserving the audio whilst simultaneously generating a transcript.
Banshee: Audio-Notes Pro. The app will offer the functionality of Apple Voice-Memos, with added speech-to-text. We're also considering injecting a super-performant GPT-3 search facility that should make searching a total breeze.
At first, Banshee will be offered purely over web, but there are definitely plans to shift the experience to native devices, too.
Banshee users want and expect a build quality as high as what they've come to expect from Apple, who have definitely set a high bar for quality. With additional state-of-the-art speech-to-text, I really think that Banshee could offer pro users and personnel a totally new way of 'writing'. Why do anything through text or type, when you could do so via speech?
What if you could 'write' notes collaboratively, and ping them throughout your organisation? Would you prefer it? Would it entail quicker ideation because you're practically sharing your thoughts and elucidating them coherently as you speak?
It's an experiment. 'Writing' notes, through voice is not something that most of the economy does naturally. Most prefer typing. But were the interface good enough, would we consider speech as our first recourse? Let's find out…
We envision a proto-Martian economy that promotes voice interfaces. Not only are they less susceptible to wear and tear, as you don't have to touch anything, a good voice-interface, coupled with intelligent processing, should be capable of understanding people with ever-increasing sophistication. That means that voice AI should be able to deliver thoughtful intelligence at the speed of thought.
As technological automation accelerates on Earth, humans will be freer to spend more time away from GUIs. We envision a world where humans rely on accessing technology, including LLMs such as GPT-3 etc., through voice because intelligence is delivered in a way that's instinctively human.
The more power we imbue into voice interfaces, the more analytical insight they can offer.
What if you could ask Banshee, 'what should I prioritise today, this week, this quarter?' Such questions are far more natural to ask, than to type out.
On Mars, an intelligent voice-assistant could highlight urgent issues and as a last resort, point the user to a GUI to assess. We aim for a voice-assistant that supersedes the capacity for a human to assess colossal volumes of signal and make decisions, or at least offer focus.
What if NASA was an AI? That would empower new generations of space-farers, who wouldn't need to rely on a Houston to mission control. Were Mission Control operational intelligence to be delivered over voice and focused on nominal performance, the cost of accessing space could fall dramatically.
All of this points to a Banshee instance that needs to reliably deliver success on Mars for private and public entities alike. We're focusing on the AI so that there is dependable infra already available for the intrepid explorer.