I tried using Rewind recently which records the contents of your screen (securely and locally) in order to search for things you saw at a later date. There’s automatic OCR on text it captured on screen which means it can be remarkably effective in responding to keywords. You can then ‘rewind’ through the recordings in order to find what you’re looking for. It’s an interesting proposal but I’ve not found it enormously useful in practice, beyond helping me reconstruct some work that Scrivener lost. It’s capacity to record a Zoom meeting in this way (you can set it to ask you at the start of each meeting) is promising but it just feels weird to have that conversation every time at the start of the meeting. The transcription summaries are good, though these involve sending the (locally produced) transcript to the company so the privacy proposition changes here. There’s an iPhone version but presumably it’s not going to be possible to synch across devices while claiming to ensure privacy.
I’ve just got access to that the GPT-infused version of the software. It doesn’t work very well yet but it’s easy to see how this could become extremely powerful, if you were happy to let it record everything. This is its summary of the 10 things I spent the most time on this week:
1. Working on your project “Generative AI for Academics”: You were near to completing various sections of different chapters, indicating substantial progress in your work [1 ↗].
2. Managing your website: You were actively managing your website, with attention to your posts and stats [2 ↗].
3. Exploring the use of the Rewind app: You were looking into its pricing and features, and you signed up for the app and verified your email for it [3 ↗][4 ↗][5 ↗][6 ↗].
4. Working on a project titled “Building my system”: You were organizing your system, with references to various tools and platforms [7 ↗].
5. Reading and responding to emails: You were actively communicating via email, with topics ranging from work-related matters to verifying your email for the Rewind app [5 ↗][8 ↗].
6. Researching and writing about AI and automation: You were engaged in reading and writing about these topics, as evidenced by your website posts and the articles you were reading [9 ↗][10 ↗].
7. Organizing your files: You were managing your files, as evidenced by the screen captures of your system [11 ↗][12 ↗].
8. Exploring the ‘Ask Rewind’ feature: You were considering the potential usefulness of the app’s ‘Ask Rewind’ feature, which is powered by GPT-4 [7 ↗][13 ↗].
9. Planning for a course: You had a document titled “First Course Reading University of Manchester.pdf” in your system, suggesting that you were preparing for a course [11 ↗].
10. Engaging with academic literature: You had a document titled “Literature review and …” in your email, suggesting that you were engaged in academic reading or writing [8 ↗].
With the exception of (1) this just isn’t an accurate summary of what I’ve done on my laptop this week. But it’s interesting to be able to ask that question. I suspect it will be more useful in recalling fringe thoughts that got lost through the week, as well as enabling more detail factual queries or analysis. At present though I’m not sure it adds a great to the non-AI version, which is extremely quick, as much as I’m intrigued by it.
It’s hard not to worry about privacy though, as the material for the queries is passed to OpenAI, raising questions about data ethics and security in transit. I accidentally recorded a few meetings when I left it on (now deleted) which is extremely worrying, as its use of the screen bypasses privacy features within Zoom. Someone can easily be recording and transcribing a Zoom meeting with your knowledge. The more dystopian prospect though is that how this might pivot towards being an organisational surveillance product. Imagine how a company could require its staff to use this on their work computers, transmit it to company servers and then have the capacity to rapidly search through their work.
Mark Carrigan 