ChatGPT and Lived Experience
Where now for writing?
Everything everywhere is always moving. Forever.
Now, everything is moving faster and farther in the world of writing. The dust has yet to settle on the latest vigorous movement in the life of LLMs. But the signs point toward substantial, lasting change.
Almost certainly, a tidal wave of written content and difficulty in detecting what is true and what is truly worth reading.
Will LLMs like ChatGPT occupy the middle zone of the written world, pushing human writing to the tails?
Poetic high peaks and one-pass, unpolished, punk and crafty valleys. A sanitised, homogenised, mediocre zone of mass produced automation in the middle.
Perhaps the act of writing will be increasingly a tool for thinking and preserved as an almost sacred thing like Taleb in tribute to early Christian scholars.
I asked ChatGPT directly. It suggested, rather unconvincingly, that LLMs would not replace human writing.
Although LLMs (Language Learning Machines) may be able to generate written content with high levels of accuracy and fluency, they are not capable of capturing the unique perspective and creativity of human writers.
Writing is an important form of communication and expression that allows people to share their ideas, experiences, and emotions with others.
The emphasise on writing to capture human experience makes me think of the most extreme version of this. The Piraha of the Amazon practice an Immediacy of Experience Principle in their culture that is firmly embedded in their language:
Pirahas avoid formulaic encodings of values and instead transmit values and information via actions and words that are original in composition with the person acting or speaking, that have been witnessed by this person, or that have been told to this person by a witness.
The Piraha people do not talk about unexperienced events, generalisations or broad comparisons. Their language is grounded in the local. The act of witnessing is central to their communication.
This does not appear to diminish the role of storytelling. Everyday stories play a vital binding role. They are enjoyed and enjoyable.
What relevance for the Future of Writing?
With powerful, paradigm-shifting technological changes it is worth considering what exactly the technology cannot do. What can they never do?
An LLM cannot witness.
Witnessing and describing experience is a protected niche.
Prepare for a flourishing of first-person accounts: of storytelling and the human perspective. Little islands of human-ness amongst the rising tide of automated mediocrity.
— — —
Sources & Inspiration
ChatGPT
Don't Sleep There are Snakes, Daniel Everett
The Dawn of Mediocre Computing, Venkatesh Rao
How I write, Nassim Nicholas Taleb
Invest early in Ad Integrity
Platforms do not exercise full control over ad integrity.
While a platform can set its policies, and develop strong policing mechanisms to moderate content, any ad integrity system will be driven by what content is submitted in the first place.
Platforms may not have control over what gets submitted, but they can influence it. They can create the conditions for 'good ads' to flourish and where fewer violating ads are submitted.
This is an exercise in signalling.
Typically platforms get away with sub-par moderation systems in their early days as it is comparatively less lucrative for bad actors to exploit users.
Not only is it less lucrative - but there is friction. A bad actor needs to learn the platform’s rules, tools and ways of working. This creates a natural disincentive for bad actors early on.
So it is tempting for Platforms to under-invest in ad integrity until bad actors show up.
This is a mistake.
The existence of fraudulent ads on a platform signals to other bad actors that it is possible to circumvent the moderation systems. It also signals that it may be worth doing so.
In other words, a bad actor may conclude that the presence of fraudulent ads indicates that it is lucrative to do so.
Bad ads attract more bad ads. In so doing, the moderation systems can be severely stressed.
But it can get much worse. Bad actors, like weeds, can lie dormant for extended periods before blooming when conditions for growth are suddenly favourable.
Once they take hold, it is serious work to get rid of them.
Prevention is better than cure.
— — —
This post is part of the Trust, Safety & Integrity Sequence a series of short thoughts inspired by my time working on Monetisation Integrity with TikTok. The posts focus mostly on the experience of trust and safety and ad integrity online, with particular emphasis on the larger digital platforms.
Norman Foster on the Future of Design
"The future will be a fusion of learning from the past, learning from and working with nature, and using the most advanced technology of our age.
The story of civilization is the story of technology."
From a conversation on the Future of Design with Neri Oxman.
The Great Hall of Possible Inventions
Imagine a cavernous hall stretching before you as far as the eye can see.
Sitting on towering oak shelves is every invention conceivable. Billions upon billions of artifacts and documents spanning wall to wall, floor to ceiling.
This incredible cathedral is the Great Hall of all Possible Inventions.
- - -
Walking through the hall, the inventions are not distributed randomly. Instead, they are placed with a particular organisation, a defined structure. In one area sits all possible types of axe - from rudimentary stone age axes of hunter gatherers to today's alloyed steel blades. To axes yet to be discovered - axes woven from nano-materials, and grown from diamonds.
Beside this area is a huge set of chainsaws, lazers and other evolutions and off-shoots of the axe.
And so it is for every other possible invention.
- - -
Now, three things are noteworthy about the Great Hall.
Most items near other items are almost identical with only very minor modifications. Our alloyed steel axe blade could have slight variations in size or shape or composition.
A lot of items are fairly poor. Weird variations that don't do much or are very suboptimal for the problem they intend to solve. The good stuff is rare.
Most of the items have never existed. They have yet to be invented. Even near artifacts that you are familiar with, most items have never been brought out of the Great Hall and into the world. On top of this, whole swathes of the Great Hall have never even been explored. Nobody has ever walked past those shelves - or even near them. The overwhelming majority of items have yet to be invented.
- - -
Invention is a search of the Great Hall.
This search began deep into our sapien past. Exploration has accelerated with the birth of civilisation and again into the Internet Age. Each invention shedding light on the shelves around it.
Drawing our attention to what is possible to create.
The more we explore the room, the more possibilities open before us.
Spend enough time wandering between the shelves of the Great Hall and you may begin to see where the most interesting possibilities sit.
- - -
Sources & Inspiration
Why Greatness Cannot be Planned by Kenneth Stanley and Joel Lehman
In Praise of Possibilities
Some sentences make better futures possible.
These sentences should be cherished. These sentences should be learned. For they have the ability to unlock seemingly impossible futures. To leap the motivation gap and spur innovators and inventors onwards.
The best sentences reframe ideas. They are, of course, gross simplifications.
But useful ones.
———
“There are only two real obstacles to any problem: physics and knowledge. If it’s physically possible, then the fundamental obstacle is always knowledge.”
- David Deutsch
“Painting a masterpiece as essentially discovering it within the set of all possible images. It’s as if we are searching through all the possibilities for the one we want.”
- Kenneth Stanley and Joel Lehman
“The ideas are there, you just have to think of them.”
-Wille Motherway
“Live in the future, then build what’s missing.”
-Paul Graham
“This is the moment that folks in the future will look back at and say, “Oh, to have been alive and well back then!" ... You are not late.”
-Kevin Kelly
Navigating the FOMOsphere
We are exposed to more. And so there's more to miss.
And there is more we imagine we're missing.
Increasingly our online experience is one of navigating the FOMOsphere. Interacting with an endless stream of possibilities.
This existential overwhelm of the modern world is that it provides an inexhaustible supply of things that seem worth doing. We're only beginning to develop cultural adaptations to confronting the FOMOsphere. These adaptations take one of two forms:
Limit the FOMO-space.
Build internal FOMO resilience.
Limiting the FOMO-space is a question of limiting time exposed to the FOMOsphere or re-engineering what constitutes that FOMOsphere. Time-limiting is perhaps most straightforward. At its most extreme, it's simply renouncing social media and unplugging entirely. More practically, it is limiting the amount of time one uses social media - and indeed when.
Re-engineering the FOMOsphere is more involved. Typically it’s fine-tuning one's curation. At the extreme end it means following fewer accounts (perhaps only close friends), or it could mean cutting off entire domains that one is interested in but not actively engaged in. It could be geographic filters where you don't see concerts or events outside your own hometown. It means ruthlessly culling exposure to accounts that cause more FOMO than they bring opportunity to engage and create.
Building FOMO resilience is a parallel path. Practicing mindfulness and reinforcing messages of understanding that FOMO is a real phenomena. On its own, this habit of willpower is not enough.
The opportunity for designers then is how to productise and amplify these desires to limit FOMO.
HMW constrain the FOMOsphere for users?
Designing limitation via user controls that are front-and-centre and intuitive. Controls that learn based on user habits; that reduces the heavy lifting of policing ones own FOMOsphere.
Or introducing truncated recommendations. Recommendations that temporarily stop once a user has engaged. Once they've chosen - we don't re-open that possibility for them.
More interesting - is the possibility of injecting the FOMOsphere with JOMO-interactions.
Interactions that allow the user to actively close off paths - actively disentangle from optionality.
Make this joy-of-missing-out an active part within the experience, without putting the willpower-onus of FOMO-limiting and resilience on users themselves.
The FOMOsphere
The FOMOsphere is the space of all possible experiences that you can imagine having.
Real or imagined, the FOMOsphere includes those experiences presented on digital platforms. But also the much larger set of experiences and possibilities hinted at by these images and videos you have seen.
It is not just that Instagram post about the The National playing The Marquee. But other bands you like who might be playing somewhere out there.
Increasingly our online experience is one of navigating the FOMOsphere.
Interacting with and being aware of an endless stream of activities that an individual is not involved in, but could be.
Trust, Safety and Organic’s Blurring Lines
Creativity in the 2020's is no longer centred on pictures in a feed with clearly outlined areas for ads to appear.
It’s video. It’s immersive media. It’s generative AI.
It’s new ways of creating content and it’s new business models interwoven with it.
Amplified by these macro-trends, the traditional borders between organic and monetized content are blurring. On top of which, users rightly demand authenticity and safety from brands.
This is creating an interesting shift in Trust and Safety. A shift that’s under-appreciated. Bluntly stated: How do users know what to trust?
Ad integrity sits within this emerging space of new challenges and unanswered questions.
And this shift is especially important for younger users. Individuals are entering this new internet at increasingly earlier ages. All platforms have a duty to protect them. But more than that - to raise standards in how industries speak to children and teenagers when they are promoting their brands and products.
Despite the latest noise about moderation at Twitter. The platforms need more moderation. Not less.
— — —
This post is part of the Trust, Safety & Integrity Sequence a series of short thoughts inspired by my time working on Monetisation Integrity with TikTok. The posts focus mostly on the experience of trust and safety and ad integrity online, with particular emphasis on the larger digital platforms.
Why I joined TikTok
I love the app, and what the app promises.
It's fun, it's creative, it's silly. It's a guaranteed laugh.
From a Product perspective, it's immersive, AI-first and truly designed for mobile in way that other platforms are not.
On top of that, the challenge of ad integrity is a problem-worth-solving. Making sure that hundreds of millions of users are safe is both important and incredibly difficult.
The convergence of these four macro-trends in a single role was incredibly alluring. The fact that I had been independently monitoring these trends with the Digital Futures team at Mastercard, strengthened my conviction.
Super-ceding these reasons, however, was the fact that I wanted to get back into the startup game.
To test myself at that hyper-pace. To take a more central role that immediately influenced company direction. Rather than the inevitably advisory role that longer-term innovation teams take.
— — —
This post is part of the TikTok Sequence, a series of short thoughts inspired by my time working with TikTok. The posts focus mostly on the experience of work culture at TikTok and companies like it.
Futurism and Procrastination
“The idea of the future, pregnant with an infinity of possibilities, is more fruitful than the future itself”.
The possibility space of the futures funnel is one dripping with hope. Not simply the hope of a great future. But the ecstatic hope of many great futures.
One path into the future amongst infinite possible futures.
I've felt the heavy weight of procrastination when confronted with career decisions. Of course, this is a strategy of emotional avoidance - an attempt to avoid the psychological distress that accompanies the acknowledgement that my own future is a finite thing.
Futurism offers a reprieve of sorts.
Exploring and documenting the possibility space of the future, keeps close contact with all of these many great futures.
And delays commitment. For now.
Experience Design and existential overwhelm
The existential overwhelm of the modern world is that it provides an inexhaustible supply of things that seem worth doing.
Experience design exists within this context of infinite possible experiences available to the user. Mostly, users are vaguely aware of this lingering in the background.
But some product experiences make this explicit.
Facebook, for example, is an efficient way to stay informed about events you might like to attend. It’s also a guaranteed way to hear about more events you’d like to attend than anyone possibly could attend. For even the least superficial, most functional user of Facebook, there is no escape from FOMO.
Perhaps this is an inevitability of connecting users to others and offering those users choices.
But it strikes me as fertile ground for design innovation.
How might we design experiences that allow users to take joy in missing out?
To revel in the experience of having chosen one option and determinedly cut off all others.
Sources
Four Thousand Weeks, Oliver Burkeman
Progress and the accumulation of Knowledge
If progress is real … it is not because we are born any healthier, better, or wiser than infants were in the past, but because we are born to a richer heritage, born on a higher level of that pedestal which the accumulation of knowledge and art raises as the ground and support of our being.
From the Lessons of History by Will and Ariel Durant.
Here then is a heuristic through which we can measure the impact of a technology and its contribution to Progress.
Human-Centred Design's Christian roots
Human-Centered design aims to make systems usable and useful by focusing on the users, their needs and requirements.
At its core, HCD is humanist design - it puts the experience of the individual at its centre. With this basis, HCD aims to improve human well-being, user satisfaction, accessibility and sustainability. Equally, HCD aims to counteract possible adverse effects of use on human health, safety and performance.
Humanism derives ultimately from claims made in the Bible. That humans are made in God’s image; that his Son died equally for everyone. That everyone is equal - that there is neither Jew nor Greek, slave nor free, male nor female. From those seeds planted 2000 years ago, Tom Holland argues in Dominion, that secularism, the enlightenment and humanism emerged from these fundamentally Christian assumptions and moral values.
It follows then that Human-Centred Design, the pre-eminent design movement of the past 40 years, shares these Christian origins and its underlying assumptions.
With Christianity as the ultimate seed of HCD - two questions spring to mind:
Does making explicit these origins help us to better design from a HCD-perspective?
Does making explicit these origins shine light on possible short-comings of HCD?
Why Ad Integrity is important
For better or for worse, advertising powers the Internet.
In fact, it powers the information economy more broadly. Until we find practical ways to reliably implement micro-payments for content, advertising is here to stay.*
Ensuring that advertising is working effectively then is to ensure that digital capitalism is working.
A digital ad is a guarantee that a message will be seen. It is a little space on the internet where individuals or organisations can deliver a message to thousands or millions of people.
In fact, ads are one of the only ways you can guarantee that your message will be seen and heard.
Bad actors know this. And bad actors like to spread malicious messages. Those messages can be propaganda and disinformation or more typically those messages can be scams or fraud.
Because ads guarantee Reach** and if you agree that Freedom of Speech should not mean Freedom of Reach, then ads most certainly must be moderated strictly.
Defending ads as cultural attack vectors and degraders of trust in commerce is a problem worth solving. Without it, commerce could not take place online.
Notes
*Even the on-going shift toward ecommerce on social platforms will include most of the elements that power today's digital advertising.
** This is not entirely true for all ad products but in all cases is more true than for organic content.
— — —
This post is part of the Trust, Safety & Integrity Sequence a series of short thoughts inspired by my time working on Monetisation Integrity with TikTok. The posts focus mostly on the experience of trust and safety and ad integrity online, with particular emphasis on the larger digital platforms.
Who should join TikTok
Tik Tok is a bad fit for most people working in Tech.
It is an unusual company, doing unusual things in unusual ways. And doing them unusually fast.
But there are plenty of reasons to join. And reasons why joining may be a truly great decision for you. Tik Tok may be a great decision for you if:
you are undervalued and looking for an opportunity to prove yourself at the highest level.
want to undertake something truly unique with all of the cultural challenges that come with that.
operate at scale, at a mix of levels above and below what your are used to.
The people who do best at Tik Tok have three things in common.
They have a fire in their belly, a chip on the shoulder or some other intrinsic motivation spurring them on through challenges, frustrations and setbacks.
They are comfortable moving up and down the org stack. One minute getting their hands dirty on a detailed solution, the next creating a future-facing location strategy for a medium-sized org.
They excel in ambiguity, are adaptable and can adjust ways of working between the stable 'best practice' of western big tech to rapid, unstructured and occasionally opaque 'ready, fire, aim' of Chinese tech development.
If any two of these describes you, then you will likely be unnaturally suited to working at Tik Tok and so benefit from getting a front seat at the world's fastest growing tech company.
— — —
This post is part of the TikTok Sequence, a series of short thoughts inspired by my time working with TikTok. The posts focus mostly on the experience of work culture at TikTok and companies like it.
FOMO in the Exponential Age
We fear missing out, because missing out could mean (social) death. In an exponential age, we fear missing out exponentially.
We are children of a world that is complex, dynamic and mostly imperceptible. We are sensing organisms and our survival requires that we know what are threats and what are opportunities. Processing information from the environment is vital. We have evolved to filter and triage information from the physical environment of grasslands and forests in social environments of small groups.
Our environment is increasingly mediated by social platforms.
So processing information increasingly means processing social signals from people well-outside of our immediate physical and cultural environments. In social space, we don't have the same fast-trigger cues that would be expected in the real world of trees and tribe. Cues that allow us to filter out the relevant from the irrelevant.
On social media, we tend to look at everything.
Every signal is a potential opportunity to tap into exponential growth. Deciding what to ignore then becomes life and death. Becomes a toss up between growth and stagnation.
But it's worse. If others are growing exponentially, and you're growing but only growing linearly then you are not only stagnating. You are moving comparatively backwards. This is the Red Queen Effect on steroids. Fear of missing out, becomes fear of being left entirely behind.
An understandable fear. But an unsustainable one.
Filtering then is the skill of the 21st century.
Some of this can be designed up-front. We can curate our information sources and take an active role in training the platforms to our preferences.
Our attention is too valuable to passively rely entirely on the algorithms of others.
Nor is it sufficient to solve this through curation alone. Most of what even esteemed others say online is not relevant to us. It is mere distraction.
We must be able to quickly gauge if there's something in any given social signal worth paying attention to. Fast-filtering that will immediately allow us to determine if this is worth paying attention to. That requires deliberate practice.
It is a skill worth cultivating.
Lewis Mumford On Technology
“We are not suffering from too much machinery, but from too little.
For let there be enough iron servants and more of us shall be able to sit on the tops of mountains and stare into the blue sky and waste valuable hours, imagining the things that ought to be.”
And so it is with digital servants. The promise of automation once again presents us with this opportunity to sit and imagine.
Should we choose to take it.
Principles of SMS Design
Designing for SMS is an exercise in constraint.
This sparse, bare-bones medium offers a spartan challenge for designers and product managers to hone their skills and return to design fundamentals. SMS design makes explicit the wicked trade-offs between minimalism and establishing context, between clarity and natural language, between usefulness and brand development.
Two primary experiences designing for the medium of SMS have sharpened my design practice in other media. The first was a series of payments and commerce proofs-of-concept that combined SMS and voice for bottom-of-the-pyramid customers in Africa and Latin America. The second was the design of a system used at the outbreak of Covid to monitor out-patients at St. James' Hospital in Dublin.
SMS is linear, asynchronous, slow.
Design constraints mean that interactions must be well-thought-out and clear. As a medium, SMS offers two broad design families with which to create the user experience: conversation design or menu design. The use-case and user expectations should point to which of these families of design best fits the problem.
Conversation design mirrors human to human interaction with SMS. This more natural approach, when done correctly, makes the interaction and brand association more human, more personable. It can be seamlessly integrated with chatbots when users have internet access. These benefits come at the cost of complexity. Human conversations can and do wander. As chatbot designers know, humans say the most unexpected things. In an online setting, there is scope to clarify users’ texts - it is faster, and cheaper than SMS. In SMS, the guard-rails for conversation must be tighter.
Menu design is perhaps the ultimate limitation of these guardrails. It is a premeditated calcification of where these text interactions can lead. As such, it lacks the flexibility of dealing with edge-cases. It forces the user quite explicitly into using our mental model of navigating the service. These limitations are strict. The menu should be limited to a very few options; ideally 2-5, up to 9 at a stretch. And of course, menus are anything but human. They are faceless, artificial, bureaucratic tedious things. No brand has ever been improved by an SMS list of menu-options.
Some principles then:
Less is way more. Only the most essential product features should be included. Each additional menu item beyond 5, often has negative marginal utility. It degrades the user experience.
Spell it out. Unlike most mediums of communication, SMS offers few signifiers of what's possible. Potential actions need to be specified explicitly from one text to the next. Expecting users to remember common commands is unreasonable.
Expect spelling mstakes. Responses are prone to error and abbreviations. Numbering entries reduces the chances of error. However, long lists introduce their own possibility of error.
Expect delays. SMSs do not always appear when they should, or indeed in the order specified. Roaming in particular messes this up. Design for this frustration.
Say it again. SMS must allow for the fact that coverage issues may mean that texts do not get delivered. A resend option should be included.
Where do the edge-cases go? There's no cheap hyperlink to sweep all the details under the carpet. Or is there. User journeys that demand a switch in medium should be considered non-essential.
Cost and trust. Either you shoulder the cost of the SMS within your business model or assume that users will quickly learn to curb their use. Users are generally mistrustful of corporate texts and worry that they are being scammed. Even reasonable clarification texts become a major frustration.
Decision-boundaries and super-labellers
Super-labellers are the untapped resource of algorithmic improvement.
For a given AI problem there are two fundamental sources of possible improvement: the algorithm or the data.*
In data, more is better.
Quantity has a quality all of its own. This unreasonable effectiveness of data was true in 2011 when Peter Norvig pointed it out; it's truer today.
The phrase itself is deceptive, however. Most use-cases do not lend themselves to increasing the input data by orders of magnitude. Data must be labelled. And labelling is expensive. Humans have to do the work and humans must be paid a living wage.
In practice, ‘Just add more data’ is rarely the answer. We must decide where to spend on labelling. While data, in aggregate, is a commodity - it is not a uniform commodity. Some data is more equal than others.
Do we emphasise the most important use-cases, the most used use-cases, those use-cases prone to catastrophic error? We can have all three if we are particular about which data points in feature space we choose to label.
Within each labelling task, there exists a subset of labels that are most valuable. The most valuable data is data close to the decision-boundary. Unique examples that help to define the line between one class and another.**
Invest in decision-boundary data.
To invest proportionately more in the decision-boundary data; we need to find examples that we would expect to be ambiguous. Examples where labellers disagree. Examples where we know models have gotten them wrong in the past. Examples where user feedback disagrees with our own decision-making or indicates there’s nuance we haven’t captured.
This allows us to identify the population of data from which to label. We have immediately saved time and effort. And cost.
But there's more.
If these areas are difficult, ambiguous and contested, then we may simply be feeding the algorithms with noise. This is especially true if we consider that most large-scale labelling is carried out in developing countries by non-expert labellers without full cultural context of the data and use-case.
So we need labellers who are themselves more accurate; more discerning. Who have a sixth sense for this specific labelling problem.
These super-labellers exist. As in most domains, there are individuals who are substantially better than their peers. Labellers that intuitively know which side of the boundary this example should sit. Regardless of whether the labelling instructions deem them to be so or not. ***
We want super-labellers.
How do we get them? We seek them out. We nurture them. We train them.
A finely tuned training program that interrogates labelling decisions in detail to find out exactly what constitutes the boundary. This goes well beyond traditional Quality Reviews. Training programs where trainers help super-labellers to build their intuition. To listen to it. To feed it back to rules-writing and Policy.
At the organisational level, encouraging labellers to hyper-specialise and become niche subject matter experts. Far below typical SME focus area, we are looking for these labellers to inhabit the feature space and understand it at a deep, fundamental level.
A super-labeller interrogates the decision boundary. Probing and refining, hungrily seeking out examples close to it and feeding that back to our policy teams and to other labellers.
And ultimately to the algorithms.
Notes
* Here, we’re focusing purely on the AI domain. We could also improve via problem framing, UI/UX.
** “Line” used here very loosely.
*** Labelling tasks are outlined in an SOP, where the designer will need to identify edge-cases and rules that govern those edge-cases well in advance. These are not always accurate at the boundary.
The Bullshit-Industrial Complex just got Bullshitier
Generative AI poses an acute short-term risk to information quality on the internet.
2022 has been the year that Generative AI has truly come into its own - it has evolved from 'hmm, that's interesting' to 'useful for some narrow use-cases'. Unfortunately, the incentives to adoption align perfectly with those who create low quality content and lots of it.
Prediction: the scaled output from LLMs will soon pour rocket fuel on the furnaces of the Bullshit-Industrial Complex.
Deconstructing, this implies that LLMs will be adopted and adopted quickly. And that on average LLM-augmented writing will be worse (in style, usefulness or truthfulness) than those illustrious few writing outside the Bullshit-Industrial Complex.
What has me confident in this pessimistic outlook are two core assertions:
LLMs do not need widespread adoption to flood the internet. They are fast and cheap. With the means of production automated, a cohort of early adopters can quickly outproduce trad writers.
The most-incentivised users create the lowest quality content. The major tech platforms reward the regular and frequent creation of content. Search rewards (amongst other things) hyper-specific content (with affiliate marketing piggy-backing on that).
There are of course counter-arguments to this prediction.
Perhaps some second order effect will now drive readers to more thoughtful curation of sources, or search and recommendation will be forced to learn better ways of surfacing the most relevant content. Or we may learn that humans are naturally better editors than blank slate writers - and so, article quality may instead improve.
I hope this is true.
At a 3-5 year time horizon, I expect that users and the algorithms that feed us will adapt to the coming tidal wave of machine-generated content. Going further, it may even be a good thing and hasten the demise of the Bullshit-Industrial Complex.
And there are, of course, other use-cases where the next generation of LLMs in particular are expected to excel.
But in 2023 and 2024, I wear my pessimism hat proudly and expect that the average quality of written content online will nose-dive.