This is the text from the Google Doc notes the group wrote notes on during the Future of Text Symposium 2020 and launch of The Future of Text book. The original document, which is now locked is at https://docs.google.com/document/d/183YIr7R4v1tqYi6WQAJtbn1epkLu6P2t6BfgsXRt-mE/edit?usp=sharing
Notes for a Letter to Future Contributors
Please feel free to enter your thoughts below, making sure to add your name should you wish.
Breakout Room 1
- Liquid as proof of concept: UI’s that fluidly integrate all functionality around texts
- Quality samples (summaries), not exhaustive explorations
- Physical intuition leading to new platforms
- Massive, centralized systems with creepy algorithmic aggregation and shaping of data are drivers toward unchecked consumerism and resource exhaustion. It could lead to a reversion to low-tech or we could move in a more decentralized and localized / small-is-beautiful direction.
- We noticed that there is a lot of potential for good debate in the subjects but we need a platform and regular meetings maybe to support / further the debateW
- Fuzzy matching between competing ontologies vs. collaborative development of shared ontologies
- Systems that promote contextualization as default can perhaps help us overcome certain evolutionary relic cognitive biases.
- Messages were dystopian, is this a 2020 artefact?
- It seems that people tend to believe what’s written, even the doubtful words on the Internet, maybe because it is confirmation of some deep beliefs?
- We need to create areas where people can find middle ground instead of just finding their ‘filter bubble’ amplified.
- Anything written should have a clear indication that it is a fabrication.
- Part of the future of text should be the ability of people to establish the truth value of texts; there’s a role for wikipedia and the like here: could there be a text that estimates its own truth value/level?
- There are parallel universes of disinformation that confirm people’s false beliefs
- Suppressing bad info doesn’t work. But we can contextualize it so people see the falsehoods in contrast with reality.
- We need technologies to disrupt our prejudices and fixed ideas
Ismail: Can machine learning understand concepts? Through context?
Denotative and connotative cloud of every concept; connections within text can constrain (or expand!) ambiguity. Training data as machine’s context. How rich is that context compared to human context? (Which is also changing) We all translate the text’s concepts into our personal cultural context. Machines as having a much lower-dimensional context thanks to limited sensorium, but similar process otherwise.
Jeremy: AI capital and global conflict. (based in share humanity)
Conversational bridges as counteracting that. (Shared reference to Society Library)
Breakout Room 2
Notes: time was too short …
Sustainability – how to preserve over time.
Some of the issues are social, legal rather than technical
Emulation & Migration forwards
What to keep: code; representation; experience
How to teach that for the future – eg awareness of code
How to discern disinformation?
Trust in information – how to foster it?
Filtering internet information –
Design aspects of platforms to facilitate sense-making and rough consensus rather than outrage and chaos (example of how Taiwan deals with disinformation and polls citizens on controversial topics – digital democracy)
Breakout Room 3
Notes: We discussed, briefly, what is kept for tomorrow? Well, this is what is kept from group 3. DDeR. Also:
- Importance of Markdown
Room 3 round 2…
Observations hastily noted (please improve this everyone!):
Wave of science, math, tech, algorithms
Anglo centric. Unicode centric (does not encode all scripts well). Content addressing ignores encoding – focusses on bits rather than meaning.
Can’t make all references work in a moving world, doesn’t make sense to try, all communication will be local(?)
Is language evolving more quickly now? Generational balkanisation of the online world (in time rather than space|)?
At same time the world is more connected.
Some countries have two major groups…
Secondary meaning of emojis, in different territories
When people try to stop us communicating we find a way
Will the future be more voice rather than text?
Influence of free video chats… biggest change in recent years
Pandemic/lockdown has led to more youtube usage and everything that goes with it
Video can be good at conveying emotion. Is it so good for looking at an argument, describing systemic behaviour?
Video not so beneficial if visually impaired. Video to text features? Beyond audio description.
A slow reader can skip ahead for a better part. Navigation.
Death of languages… will we end up with one? Or 6 or 4?
“No!” Text is an imperfect encoding of spoken language.
…identity. Jargon. Subversion.
Individuals have more than one language anyway
If we could read semiotics of videos, we will subvert the subversions, …
Will dominant languages change?
Wendy: History of hypertext conferences where technologists and poets would meet. The Web conferences were narrower. Welcoming the presence of creators here. (A tale of three conferences) (semantics as 3rd conference: logicians)
Sonja: similar divergence history in design. Offline vs online, writing vs visual, etc. and subsequent convergence.
Asking about algorithmic identities: the simplified identity as understood and targeted by marketing algorithms…
Seemed to be a theme: the text is a consensus object, around which a tribe congregates, but also tries to reach out to other people. Hypertext as an attempt to create wormholes between worldviews. The pulsating rhythm of divergence and convergence around meaning.
Poetry book w/ performances in it
Speaking to camera wearing headphones
Video on a page back in 1979… not faked BUT…?
Lesia: a giant wardrobe on the side
Reading about old dead computers
Is the future populated with junk from the past…?
Windows is a bit of a zombie technology
I remember when WindowsXP came out, it was reckoned to be a big advance… but it’s hardly changed since then, that’s 20 years ago! And Microsoft doesn’t like ditching old stuff. While still trying to remain compatibility with the 90s. Programs like Office… and Microsoft basically updates the user interface.
Huge compatibility issues/costs to Microsoft adding new things.
It seems like they have put a lot of money and time from being “corporate” to being “we do open source now”. But I don’t see a future in which Word gets updated; maybe trying to fit into contemporary developer culture… but
Now everyone is using their shit (but not like their services like Microsoft Azure). Indeed, the money has moved on: it doesn’t have to do with things like selling copies of office or copies of Windows for that matter; they want to make money in different ways. I don’t know if bleeding edge tech people use Azure, but the people who use that aren’t going to be thrilled to sign up for a 3 year contract for Outlook 365.
“Why can’t we use Google” — students/others say. Industry put a lot of money into staying in an ossified ecosystem.
Why do people beat up on COBOL?
When it came out, the “Programmer” education didn’t exist… people just signed up informally and no one was trained properly.
So this is why you got spaghetti code.
Is COBOL the big zombie?
I managed to talk with an old banking exec… is it true that everything is written in COBOL. Actually it’s written in PL1 — this was supposed to take the best from COBOL, ALGOL and FORTRAN.
But people don’t read code for fun / aesthetics … do they?
Knuth has this idea of Literate Programming that is much quoted, though not clear how much this has been taken up. Does he mean reading code as an aesthetic thing?
Yeah, he meant that code could be more expressive.
E.g., Knuth is for “expressive” rather than “short” variable namse.
At the same time this is where he felt he could express himself as a programmer and with a human readable text.
Similar to “self documenting code”… it should be apparent what it does! This makes it more likely to work, apart from anything else.
But if it didn’t work with COBOL…
When you’re talking about reading text it sounds like you’re talking
At the start of reading and writing, a lot of it didn’t make that much sense. People didn’t know how to tell stories properly. Maybe we are at that point in coding too. We don’t know how to express it properly
Breakout Room 4
— The application that Frode built is excellent, it would be nice to have it on the web (not tied in to the Mac platform)
— This tool is the essence of Engelbart: the quest to make that ‘online’ is important. Gyuri is trying to do this (and he’s here)
— Roam… now many people are coming out of the woodwork.
— Stephanie: I endorse the Englebart vision & the importance of the individual but I am also more wary of the power of these decisions not in the hands of individuals but rather in the hands of the large institutions who are controlling these decisions.
- maximize ambiguity options for survival
Those more involve with the computer field — vs someone like me who deals w/ the classics. We need the ability to work with the text. E.g., 40 to 60K words. Complex realities require a certain amount of space. The word is not the unit, the sentence is the unit. This is what allows you to have explanatory text. These are not descriptors of concepts; they require the ability to manipulate text. People would then have the ability to manipulate these things. Wondering about how “that kind of communication” would be significantly affected by automation or otherwise. The stuff that Frode did facilitates the process for someone who has the ideas to pull them together.
Polygraph altogether one thing joseph corneli
Can analyze these structures
Same technology used for advertising can be used as liberating devices
Legacy of humanity wrapped up in these discussions
Manipulation is not random
— But I don’t think we should accept that this is a foregone conclusion that we WILL SIMPLIFY OUR THINKING to match computer structures. We may continue on with how things have been doing things WITH COMPUTERS ALSO adding more information.
Chris: content addressing is something that may save us. IPFS: addressibility. What is it that we need to save out of all this? Lots of the information & bits maybe don’t matter so much. “At least we should save the idea of the atom.”
Internet Archive: who will select what deserves to be saved? The bill from the last dinner w/ my wife or grandson? The latest edition of Aristotle? Now we are adding 430zetabytes per year; the amount of information is becoming huge. Vint is worrying that we aren’t saving the codes when the programs will have disappeared: but, how far do we keep adding to the volume of information without a capacity to filter it.
- Sufficiently self-describing information doesn’t depend on particular code.. Or if the code is itself content-addressed (and open and platform independent), the information can at least reliably reference it.
Chris: People connect, verifiably; what matters will emerge from the structure. If you make the conversation map itself, that’s one way to sort out what’s relevant and not. The significance of conversations can self-signify.
Alan Kay was able to run his program written in 1972 was able to run the thing in the browser. It’s possible to create bootstrapped things that live forever. Englebart had a 10K line kernel which they could co-evolve. ‘PERSONAL AGENCY’, ‘CONTENT ADDRESSIBLE’, and ‘SLAY LAYERS’ and have a tiny kernel.
Jack: Building open source machine reading platform turning loose on pub med
Parsers → sentences structured to the degree that professionals have their own sentence structures, get a parse tree, this is the beginning of the graph; write processors to pull meaning structures out of these graphs; the meaning structures are like [A causes B] and from there you can draw more structure — topic structures & Conceptual Graphs to map a large area. Build a map of a large territory.
“Machine Reading comprehension”
6th Extinctions; online storehouse of content; social machines may serve as a prosthetic…
The extent to which when we deal with text: want to understand, multiple levels of context – few words can communicate profound ideas (for the people, by the people, … )
What is the meaning – have to explain a lot of the context – history slavery, lincoln – using sentences to provide context, need to context to understand the sentence
- Requires a bigger interaction between machines
- Likely interaction and interpretive ability
- Important concepts like liberty and limitations of civic engagement
- Ideas like this only can be stored in text
Scientists don’t want to stake their reputation on a causal claim. Leave it to you to determine their intention.
See the relationship as a partnership, not a confrontation – robots will not take all our jobs, and we will have a 3-day weekend!
Both humans and machines can annotate inferred meaning, emotion, etc. within text, to help build context.
Breakout Room 5
Chatting about executable text, which we find inspiring.
“The design challenge is to make URLs human readable, writable (natural language) and explorable, machine readable (failsafe), space-efficient (short), portable to different contexts (digital ←→ analog) and democratic.” https://thinkingtools.space/t/proposal-a-common-spec/33/7
NEW BREAKOUT ROOM 5! Please improve…
“More creative” set of videos(?) How to distinguish what we’ve seen:
– Are people thinking of text a writers, the power of language
– For others it’s technical, code, information, transformation – these people speak from external view
Can text be multimedia friendly? Yes. And it remediates.
The textual, live environment
Social machines – reconceptualisation needed, video was good
Transformed relationship with music. But not discussing music today!
Strong apocalyptic thread running through, not a moment of optimism
Is AI driving the conversations (in public sphere) in certain directions? Yes.
Is Ai driving us towards smaller communities and static beliefs – small communities get bigger voice, amplifications, distortions
Who is in charge of the AIs, Companies. Governments. Not individuals.
“Does AI live in rabbitholes?”
Text as a code for mediation and remediation
Frode: lots of struggles to see text
Text + picture is better?
Computational text, produced through code, popping up
AI as tool. Shallow learning. The new copy and paste.
ANOTHER NEW ROOM 5
Is there any evidence that DDeR is human?
Niko, Cornell CS, ML in robotics meets people from original rooms 3 and 8!
AI has pioneered in text, volumes available for training. Can language models be trainable on much smaller datasets.
Failure modes of AI – need to add common sense intuitions? Can that emerge or do we need to inject ontological prior?
Will the tech enable us to bring people back together – aligned world models (again)?
Neurosymbolic approaches needed
“Computation always wins”, you’ll lose if you bet against it
Discussion of gender diversity and the book
Something in the framing of approaches? Listening?
Breakout Room 6
Breakout Room 7
- Preservation. When you write, think about sustaining that text – or its encoded meaning alive over time. Embracing Open formats is non-trivial to aid long term survival
- Issue of modern html being server-dependent, hence more fragile. (Compare with serverless distributed web, but also introduces code state dependency)
- Tension between the vision (InfoCentral and HyperKnowledge and earlier Microcosm and Memex) of text as always embedded in and connected to a broader context of contrasting texts; vs the dystopic vision of personalized text (and the importance of privacy)
- Incremental formalization in interaction with fuzzy or ineffable ideas. Semantic web did not fulfill that promise yet. Limits of disciplined thinking, difficulty in approaching abstraction. (The UX of semantics has not been worked out yet)
- This is all in Ted Nelson’s work also. Quixotic quality of this conference. The technology may finally catch up with the dreams and ideas of long ago.
- Difficulty of maintaining semantic links in current HTML, so they become sheer marketing devices.
When you see a triangle that isn’t there — you see tables one that is much longer than the other… been triggered by this social machine idea. We’re doomed: there was a big world championship of Chess, and the guys who won used bog standard Android chess programs and combined humans & machines. The trust in the machine… I regularly write word combination that Google’s never seen. If I write “Indieverse” you know what I’m talking about.
At this point, we’re all basically cyborgs, combinations of us and the surrounding machinery.
I BELIEVE in the man-machine future but not the machine future. “They think they can rely on machines” but they are just wrong, that’s the real problem. The whole epistemology is wrong. They really are clueless.
You see that in some of the discussions here, knowledge is monotonic, deterministic, analysable… I was a physicist but I’m mostly a humanist now and you know that’s not true. The knowledge exists as opposed to data. Data is just people in disguise. The dignity of existence… oh yes we can do it! No we can’t. Knowledge is evolving context. Also w/ lived experience.
Human 20%——80% machine; let’s be clear where we sit on this spectrum.
Maria Karam: I work in captioning for professional sports. If we were able to capture this text in real time, there’s an accessibility angle.
But also: text is an imperfect serialisation of language. It doesnt’ capture a lot.
One written language for many different spoken “languages” in Chinese.
Ambiguity maximises opportunities for survival — that changes your view. It’s important! I want these. What is the bit that you can pass on w/ time banding.
Maria: most of what I work with is audio-visual. We started with text on the internet and in a way maybe we are coming full circle. As someone coming from outside this world, it looks like text is going to come full circle. We need to integrate the different studies: AI, augmented reality, throw text everywhere! Having labels pop up on everything. I’m overwhelmed with excitement… how to use text to link together the different sensory dimensions to explain & build human understanding? That’s a summary of my excitement!
A lot of the texts now are basically text fragments w/o captions, context, information about where they come from. They don’t have a lot of meaning. Capturing the text for posterity relies on additional material in some form. _Nigel Shadbolt… prosidy?_
Context… this is a great merging of many different fields.
In the videos there were so many different viewpoints, and ways to apply it to all the ways of thinking.
INFRADEUS? Takes any text and works like the i-ching. You can view the context from different angles.
You should be able to map some of the context, intertwingularity… as you look at things from different points of view, any node in the graph should give you a different worldview. This is my vision of text “going live”. If you create one connection, the entire world rearranges.
Prezi as a presentation — you can take your own route through a series of slides / mindmap.
All of these games (e.g. single-shooter games). Sensory substitution, text into other modalities…
Breakout Room 8
- It’s cool to see so many awesome people here. (+1)
- Future of Text, book is inspirational .
- The author app is awesome. I ‘ve stopped using all other word processor apps
The capture of BibTeX information in copying and pasting is brilliant. We will quickly see however, how poor the management of metadata is. Hopefully, this will get people to focus more attention onto the management of metadata. We need to get people to create beautifully crafted metadata
Versioning in citations should be important.
Versioning should exist in a few domains: both for individual citations and for global comparison of citations within the text. -Alex H
A dynamic view for citations could be useful. This could help with investigating how information from cited papers is distributed spatially within a text. -Alex H
Dynamic view could be similar to the website ‘connected papers’ that shows the references in a graph view, as well as differentiates between influential papers (ones that inspired the paper) and derivative papers (ones that reference the paper)? (https://www.connectedpapers.com/)
High tech vs low tech
Gtp 3, chiral network, vs low tech web
Important dichotomy of the future of the internet, two visions
What about the commercial element? Everything is driven by commerce. Comment in one of the videos, how the formats we are using will not be accessible in 20-30 years from now?
How to make it more multi-media friendly – consuming standards – reading a document – how to make text itself more multi-media
What about us a we as consumers – passive consumption vs active consumption
How can the media facilitate more active engagement with the content, rather than maximize engagement and time spent.
Or, commercial of intellectual labor? Customers can recognize brands – impressive how companies are able to unconsciously able to transfer knowledge and information almost without awareness or effort.
Viewing the users emotion while consuming the content.
We need better ways to hold ambiguity without imposing structure upon it. Incremental formalisation.
Are we producing too *much* information (text)? A problem of information triage.
Blending of editorial actual news is problematic for
“I spoke with a media producer from LA this week who said he is replacing human actors with animated ones—COVID has shown him and others in Hollywood that this approach will allow for them to survive financially”
Well, that was inspirational. (Abstract Video)
(Not sure what break out room I was in): Is computation competent, in any speculative future, to reconfigure/establish/integrate meaningful relations between text and language-as-such? (recalling that (text !== language) as far as anyone can show) – John C
History, archaeology, forensics and even scientific reproducibility (etc.) are all exercises in the reconstruction of meaning, context and provenance from surviving (often textual) artefacts. I’m not sure this will ever be simple. There are additional dangers with the survival of digital material – it is far more in the control of a few (cloudy, technocratic) organisations than was previously the case, and the cost of production is minimal so the signal to noise ratio (or truth to falsehood if you wish) may be poor enough to render to reliable extraction of meaning almost impossible. (Neil J, post-script)