Transcribing all our conversations 24/7 will be weird and also useful maybe
13.50, Wednesday 14 Dec 2022 Link to this post
Sooner or later, every single conversation I have will be recorded and transcribed and I’ll be able to look back at it later – details from a phone call with the bank, in the hardware store asking a question, someone mentions a book at the pub, an idea in a workshop. Ignoring the societal consequences for a sec lol ahem… how should the app to manage all that chatter work?
What can you do when you record everything?
I record myself on audio 24x7 and use an AI to process the information.
I bought a couple of Chinese microphones, I wear them and turn them on all day recording everything I speak, at the end of the day the files are processed with OpenAi’s Whisper and transformed into text files from which the information is extracted.
(Whisper is OpenAI’s new, open source automatic speech recognition neural net.)
Here’s a neat feature: he has a built-in activation keyword and stop word to indicate when the AI should pass a phrase off for additional processing.
For example to register my weight for the day I simply say out loud
Robert WEIGHT 60.1 end Robert
And another, to record expenses:
Every expense I make during the day I repeat it out loud to record it.
All of these appear on a personal dashboard at the end of the day.
Another, an unnerving concept:
According to studies on couple relationships, it is possible to predict with an accuracy of up to 90% if the couple is going to divorce by studying the interactions, specifically the relationship between positive and negative interactions between the couple … The magic ratio is 5 to 1.
He hasn’t built this (yet).
Dam’s system is a huge jump… into something, I’m not sure. It’s a memory prosthetic, partly? I wonder what I would stop doing. Would I stop adding items to the household shared shopping list because I would know that I could just search my conversations later? So in a way it’s replacing app interactions; not just for memory but a kind of very slow voice assistant.
All of this without even being realtime.
(OpenAI’s Whisper isn’t perfect. But, as a Brit, I have never had good experiences with voice recognition. Machines don’t understand my voice. I get the feeling that, for people with a North American accent, Whisper is only an incremental important. But let me tell you: for me, it’s night and day.)
How about if transcription were realtime?
Transcription basically means that conversations because machine-readable, or rather AI-consumable.
Here’s a tweet from the CEO of DoNotPay which is an app that gives legal advice on, e.g., how to fight speeding tickets.
Anyone with a speeding ticket hearing coming up, please DM me.
We want to build a @donotpay bot that listens to the court hearing via your AirPods and whispers what to say with GPT-3 and LLMs.
We just want to experiment and will pay the ticket, even if you lose!
AirPods + AI = a cyborg prosthesis of the centaur type.
Anyways, that’s for the future.
All I mean to say is that always-on transcription + realtime is enabling in all kinds of ways. Exploration required.
So let’s scope this back. Let’s imagine we record all conversations but maybe only in a work context and only in meetings.
Is there an app, 50% email client and 50% note-taking “tool for thought” that stores all of these conversations, letting me process them when I need to, and automagically surfacing tasks and relevant information?
- Realtime seems vital – or at least close to realtime. Behaviour change occurs when you close the feedback loop from conversation to action. How can insights from the transcript start influencing the conversation that you’re in?
- Search and also automation – there’s going to be a huge volume of words stored. I want to be able to say “oh yeah I need to add a line to slide 12 crediting X, Y, and Z” and find that later when I need it. But then also live tagging with trigger words: I’m inspired by Dam’s approach of simply saying his expenses out loud and then turning that into a spreadsheet each evening.
- Multiplayer – there is something troublingly asymmetric about me recording and the rest of the group not see it. It feels like, for a small group, we could all be on the same level playing field somehow, like we all have the same access, and it’s private beyond that?
You need all of this because otherwise it’s just like the meeting transcripts you get out the back of Zoom and nobody, like nobody looks at those.
Taken as given: reliable speaker ID, timestamps and geography.
Future: so how about using the transcripts as training data to make AI colleagues. Instead of searching for what my teammate X has said, why not make a bot that knows everything that X knows, and then have a conversation with them? “Hey X what was it I said I was going to email you today?” – that kind of thing.
Realtime instrumented conversation will be disruptively weird and maybe also positive.
Here’s us+ (2013) by artist Lauren McCarthy with Kyle McDonald, a plugin for Google video chats that encourages equal voices in meetings.
us+ is a Google Hangout video chat app that uses audio, facial expression, and linguistic analysis to optimize conversations based on the Linguistic Inquiry Word Count (LIWC) database, and the concept of Linguistic Style Matching (LSM). The app displays a visualization, provides pop up notifications to each participant, and takes actions (like auto-muting) when the conversation gets out of balance.
(Thanks Daniel Goddemeyer for the pointer.)
It’s a great provocation, right? It’s pretty punchy.
And this is what a good chair already does, isn’t it. Makes sure all voices are heard, coaches people into good team behaviour.
So why not build that into the software? It means that meetings can be smaller (tighter, more effective) because the “chairing” capability doesn’t need to be in the room. Some teams are naturally good at this, some need help.
It’s definitely the feedback loop that matters here.
Anyway. Good to see some experiments before this all kicks off for real.