Is Generative AI a Feature or a Platform?
Is Generative AI a Feature or a Platform?
...or an app or a service, or all of these and more?
We’re coming up to a year since OpenAI unleashed ChatGPT on the world and in some ways, we’ve seen a lot of progress on a weekly, and sometimes daily basis, since then. In other ways we haven’t seen a huge amount of progress either. Enterprise applications of generative AI technologies are few and far between and there are still plenty of ethical, usability and operational challenges to overcome. I expect 2024 to be a bigger year for generative AI than 2023 has been and whilst thinking ahead to next year, I started wondering how pervasive generative AI will be and what different forms we should expect it to take.
To help with this exploration, it’s useful to look to the past for guidance, but also not to draw too many parallels as generative AI will no doubt evolve and impact society in ways we can’t even imagine yet. Arguably, social media is the closest technology that we’ve seen to generative AI. It certainly shares many of the ethical challenges social media has, and social media has had a huge effect on society over the last 20 years which is a path I expect generative AI to follow, albeit with much great impact across all industries.
When you look back at the way social media evolved, it started as a just feature in existing application and services (i.e. the ‘share button’) before maturing over time into the huge social media platforms we see today. I am convinced that generative AI will take a similar path but will do so at a much faster pace. Because of this, I think it’s worth, even at this early stage of the technology, making some predictions of how this could play out.
The Anatomy of Digital Ecosystems
Before we get into our predictions, however, let’s define some terms and set some ground rules to help guide us:
Features
Features are specific functionalities embedded within an existing application or service to perform a particular task.
For instance, the 'share button' in early social media was a feature integrated into various websites, allowing users to share content but not necessarily to interact in a social networking context.
Apps
Apps are a standalone piece of software designed to perform a range of related tasks. Apps can embody one or multiple features, but they exist independently, often having their own user interface and operating logic.
It’s a bit 2020, but Houseparty is (was?!) a good example of a social media app - it had a singular purpose to facilitate group video calls and doesn’t have all the trappings of a typical social media platform.
Services
A Service is an offering that performs specific tasks on behalf of the user, often operating in the background. Services may be part of an app or platform, but they can also stand alone, often requiring some form of subscription or payment model.
Buffer is a useful social media service that allows users to schedule posts, track social media engagement, and manage multiple accounts across various social platforms. It operates in the background, fulfilling a very specific set of tasks related to social media, typically without providing a platform for social interaction itself.
Platforms
A Platform is a comprehensive ecosystem that combines various features, apps, and services into a unified experience. Platforms offer a broad range of functionalities, often integrated in such a way that they augment each other, providing a cohesive user experience. Some of the key features that you see on platforms are:
Multiple entry points (web, app, widgets).
User profiles.
Customisation, inc. privacy controls.
Notifications.
Search functionality.
Activity feeds.
Content curation.
Collaboration tools & sharing features.
Integrations with other apps/services.
Facebook, Instagram, TikTok, LinkedIn, X (Twitter), Pinterest, Snapchat, Slack - there are far too many social media platforms to name them all, but you get the idea!
Generative AI as a…
It’s worth stating before we get into each of the different parts outlined above that unlike social media, we’ve already seen generative AI manifest as features, applications, and services. Generative AI is not going to see the same linear progression that we saw with the development of social media - it’s moving much too fast for that.
Side note: if you plot how generative AI has evolved so far, despite it’s rapid pace I think it’s probably progressed (and will continue to progress) mostly in this order: Service → App → Feature → Platform, whereas social media mostly progressed as follows: Feature → App → Service → Platform.
Generative AI as a… Feature
Over the last few months, we’ve started to see many digital apps and platforms announce generative AI features. One of the first digital platforms out of the gates was SnapChat with My AI. Since then we’ve seen announcements from Microsoft (Windows Co-Pilot, Office 365 Co-Pilot), Google (Duet), Slack (Slack AI) and others. GitHub’s co-pilot pre-dates all of these, announced in June 2021 and is probably the earliest example of a generative AI feature being released.
I expect the announcement of new generative AI features in apps and platforms to accelerate towards 2024 and I think in a couple of years’ time we’ll look back and wonder how we ‘survived’ without generative AI features in all our apps and platforms. In the future we’ll take generative AI features for granted in the same way we take search and social features for granted right now.
Generative AI as an… App… or Service
There has been an explosion of new generative AI apps and services in 2023, from only a handful in 2022. Now we have generative AI apps and services for chat, image creation/enhancement, voice synthesis, audio creation, video creation, 3D modelling, analytics and gaming to name a few! It’s fair to say that the chat, image, and voice apps have had the majority of the limelight but that will no doubt evolve in 2024.
I’ve purposefully not distinguished between generative AI apps and services here as I think in these early days of the technology it’s very difficult to draw the line between the two. The difference between the two for generative AI should become much clearer in 2024.
Most of the generative AI apps and services that we’ve seen so far, regardless of what type of content they output are ‘text-input’ apps. This means that many of them are essentially chat interfaces whether or not they’re actually chat apps. I think this is the area we’ll see evolve the most in 2024 (at least I hope!). I really like the idea of chat being a primary user interface, but it can’t be the ONLY part of a user interface. There’s a lot of great work ahead to work out what a chat-centric user interface looks like and I’m excited to see where it takes us.
Generative AI as a… Platform
I’d argue that we have nothing anywhere near a generative AI platform even on the horizon yet, but I’m convinced this is where the major players are heading. I’m not even sure ‘generative AI platform’ is the right phrase to use, even though the central technology will be generative AI in its broadest sense.
It’s such early days that it’s probably difficult to even envision what a generative AI platform could be. So, to help us, let’s go back to those features of a platform that I outlined earlier:
Multiple entry points
This should be an easy one - any generative AI platform needs to be accessible from a browser, an app (both mobile and desktop), have browser plugins/widgets and most important of all bring voice in as a major entry point. Voice is much easier said than done (#sorrynotsorry) and it hasn’t really taken off as a major input paradigm yet. At some point in the next few years, we’ll see that change and generative AI will be the catalyst for that change.
User profiles
Again, this should be an easy one, right? User profiles are user profiles. Some are good, some are bad, but they all serve a similar purpose…. until they don’t. For a generative AI platform, I think we need to see an evolution of the user profile. The reason for this is that users of a generative AI platform will need to set up and manage many more settings and features than on a typical digital platform today. For example:
Connections/integrations with a user’s other services that they want a generative AI platform to have access to. This will enable the platform to supplement its knowledge of users’ wider digital lives.
Managing the memory/knowledge of the generative AI platform has of a user. A successful generative AI platform should record every interaction a user has with it and that should be fully visible and editable for users.
Tool/plugin management. It’s likely that a generative AI platform has access to lots of different tools to help their users and these will need to be setup and managed in a user’s profile.
Customisation, inc. privacy controls
A generative AI platform should be highly customisable as it should ultimately be a very bespoke and tailored experience for every user. It will be truly personal.
Aesthetics are important. I’d love to see a platform where users can customise its look at feel, all the way from the structure of the interface to the sounds, fonts, and colour scheme. I see customisation of the user interface of generative AI platforms needing to be the digital equivalent of the personalisation options we currently see in wearable technologies.
AI preferences. Users should be able to customise the ‘personality’ of their generative AI account, being able to set gender, ethnicity, conversational style and level of creativity and original thinking amongst many other things. These setting might even be dynamic and evolve over time based on the interactions a user has with the platform.
More sophisticated privacy controls to allow users granular control of who has visibility of, interaction with and/or access to their generative AI platform account. This will be a core part of any social interactions available on future generative AI platforms.
Notifications
Another simple one, right? Wrong! Similar to privacy controls, users of a generative AI platform are going to want (at least to start with) much more granular control of their notifications. The reason? They’re going to be handing over control of what they get notified about and when to a generative AI platform for the very first time. I fully expect the need to configure notifications to go away over time as users become more comfortable with the idea of an AI handling their notifications and the platform learning a user’s preferences. But to start with, users are going to want to have full control of their notifications.
Notifications will also turn the current interaction model with generative AI technology on its head. At the moment, we have a very ‘one-way’ interaction with generative AI where we ask it something, which it then responds to. For a generative AI platform to be truly successful that interaction needs to become ‘two-way’ with the platform having the ability to get users’ attention and prompt them for input.
Search functionality
Nope. Don’t need this one. Why would you need search functionality when you have a generative AI platform that has knowledge/memory of every interaction you’ve had with it and is connected into your wider digital life? The answer: you don’t. If you’re looking for something, just ask the platform.
Activity feeds
This is both simple and complex and then simple again all at the same time. Simple in the sense that the activity feed could just be a ‘chat’ history/log of all the interactions you’ve had with the platform. But what if the generative AI platform has access to lots of connections/integrations with a user’s other services? What if a user allows others to view, interact or even access their account? This complicates things when it comes to an activity feed as I believe users should have a full log of every action a generative AI platform has taken. But that then brings me back to the search functionality - if you want to know what the generative AI platform has been up to just ask it - simple!
Content curation
Content curation on a ‘typical’ digital platform is usually down to user selection, i.e. what topic/person/company etc. they want to follow and see content from. A generative AI platform could be that simple, but that wouldn’t be fun, would it? Another option would be for the generative AI platform to learn the content you’re interested in and do the curation for you. This is pretty much how many social media feeds currently work, but it doesn’t lead to a very satisfactory experience (in my opinion).
The answer to content curation needs to be much more nuanced. It’s not just about what a user sees, it’s also about when, where, and how they see it. For a generative AI platform, I think the definition of content is much broader as well - it could be news, images, social posts, emails, text messages, chat messages, app notifications and all sorts of content coming from all the other connections the platform has with a users’ wider digital life. Bear with me on this one…
As a user I want to consume content in many different ways, for example:
There is content I absolutely want to be alerted to the moment they’re available/happen so that I’m fully up-to-date and able to respond immediately if necessary.
There are fleeting moments when I fancy seeing what’s going on in the wider world and want to be able to consume content casually.
There are moments when I have time to purposefully consume content that I’m really interested in.
There are moments when I want to see what my friends/family/colleagues have been up to and to communicate with them.
These moments all require different types of content curation, and my hope is that a generative AI platform would be able to help me with this, presenting me with the right content in the right moment and learn from my preferences over time.
Collaboration tools & sharing features
Ah, relief - another simple one I hear you say. Yep, of course it is… a user just shares what they want to share and has the ability to un-share whenever they please! Except… this is a generative AI platform and what if the generative AI platform wants to be able to share something? How would we manage the permissions for that and how granular would they need to be? What does sharing and collaboration even mean in the context of generative AI? Is it:
human ←→ human?
human ←→ machine?
machine ←→ human?
machine ←→ machine?
All or none of the above?
A user should also be able to control whether their content is visible, extendible, or editable, not just generically ‘shared’. Lots to think on with this one.
Integrations with other apps/services
I think this will be vast - it’s not going to be as simple as just ticking a box to say you want an update on one platform to be shared automatically on another platform. If you imagine some of the integrations you might want a generative AI platform to have in order to have a better understanding of your wider digital life it could mean literally anything - your music services, streaming services, news services, emails, messaging, social platforms, banking, subscription services, mobile app integrations, smart home integrations, internet of things integrations and the list will go on.
Some of these integrations will have to come with considerable safe-guards and require huge amounts of user trust, which won’t be easy to come by to start with. But I can absolutely see the benefits.
Just managing this broad an ecosystem of APIs and integrations will be a significant technical task all on its own for a generative AI platform, let alone all the smart stuff that could be enabled once the platform has access to all that data.
Could generative AI go further than a platform?
You may have noticed that a certain word or phrase is conspicuously missing from all of the content above. It’s something that’s often talked about in lofty terms and certainly something that all the major generative AI players have talked about creating. Have you guessed what it is yet?
Oh, you got it - you smart people!
The word is SAMANTHA ASSISTANT. Or personal assistant, virtual assistant, or life coach if you must. I don’t think the often used term co-pilot quite does this idea justice, so let’s ignore that one.
Yes, if you piece together all the generative AI platform features I’ve outlined above each user gets themselves their very own generative AI personal assistant. You’re welcome.
In my mind, a generative AI platform is an assistant and the natural evolution of the large digital platforms we currently have. If you go back to my original definition of a platform and put generative AI in the middle of it, you’ll see what I mean:
“A (generative AI) Platform is a comprehensive ecosystem that combines various features, apps, and services into a unified experience. (Generative AI) platforms offer a broad range of functionalities, often integrated in such a way that they augment each other, providing a cohesive user experience.”
But real life isn’t that simple - you’re not going to get to a generative AI assistant just by putting generative AI features into a digital platform. You’re also not going to build a generative AI assistant out from a generative AI app or service.
As you can see from the different platform features I outlined above, all of which need re-thinking and engineering for generative AI, there is a lot of work that needs to be done to get to anywhere near a generative AI assistant.
I don’t think the work is overly technical - I think it’s mostly in user experience design, spotting the opportunities we’ve previously missed and avoiding all of the mistakes we’ve made when it comes to giving users full transparency and control over their content and data. This will take time to get right, but it’s not out-of-reach for today’s technology.
For me, that’s a really exciting thought. I believe it’s technically possible today to deliver a technology that we’ve only dreamed of in science fiction for decades. Whilst there are challenges to work through, it’s absolutely within our reach. I’m also excited because I don’t think an assistant like this can come from any of the large digital platforms. And that means that at some point in the next 5 years (maybe less) we’re going to see a new platform emerge, built ground-up around generative AI, and seriously challenging and disrupting the large platforms we currently have in the digital ecosystem.
We’ve had the Internet (as we currently know it) for 30 years now and during that time we’ve seen a lot come and go. Of those large digital institutions that are still with us, we’ve had:
Amazon for 30 years.
Google for 25 years.
Facebook, YouTube, Twitter and LinkedIn for nearly 20 years.
iOS, Android, Instagram, SnapChat, WhatsApp, and WeChat for around 15 years.
Lastly, TikTok for 6 years.
We’re definitely overdue a change at the top and I for one can’t wait to see how generative AI shakes things up.
"The future is already here, it's just not evenly distributed."
William Gibson
This article was researched and written with help from ChatGPT, but was lovingly reviewed, edited and fine-tuned by a human.