Even though there is nothing (yes, literally nothing) about intelligence in AI technologies; even though I’m quite skeptical about pathway to the AGI; even though I think, that the AI development may pose several tangible threats to the world, I’d still like to share few words on near future of AI, or — of AI systems being integrated within software applications, to be more precise.
After going through several top-notch AI outlets (few small books, AI papers, lectures and podcasts) in the last few weeks, I can deduce the following (something like “integrated AI”) forecast for the AI development in 2024 and onwards.
Very soon, you won’t know and you won’t even care about what AI/GenAI system you’ll be using. Moreover, you won’t even naturally think of it as you’re using AI/GenAI, as these systems will be seamlessly integrated within otherwise traditional applications.
The way you’re using AI/GenAI today is mostly chatbots that surface the core AI systems and that are presented as “intelligence agents” you’re talking to. So, the way you’re using AI/GenAI systems today is using them as stand-alone systems, like ChatGPT, Dall E or Midjourney.
In the close future, the way you interact with AI/GenAI will change from using stand-alone systems to systems that are seamlessly integrated within otherwise traditional applications. Soon, you won’t care whether you’re using AI at all, or what AI you use, and like you’re using weather services, now, on your phone, absolutely not caring about how its internal implementation calls some remote [web-]services and aggregates their responses in a presentable to you way, you’ll also use AI/GenAI as a seamlessly integrated component of larger systems, as it will be present as integrated modules/blocks in traditional applications.
For instance, you may have a button in MS Word “Translate Text”, or a check-box in Photoshop “Improve Quality”, or a toggle button next to your Facebook wall “Detect false information posts” for performing corresponding actions, and you won’t even think (maybe not even realise) that back in the scenes, those functions will be enabled (provided) by AI/GenAI.
So, GenAI (and AI, in general) will become one of the building blocks of traditional applications, and AI-driven functionality will be available in such traditional applications out of the box (yet, still, using AI models in the backstage).
How this integration will happen?
AI/GenAI engines will be exposing APIs, which will enable other (traditional) applications to integrate those AI/GenAI engines in themselves. So, anytime some application will be integrating AI/GenAI system, it will be communicating to that system via internal APIs exposed by those AI systems.
AI/GenAI systems will soon be integrated, pretty much, in every (or — the most of) popular and otherwise traditional applications and systems.
One of the good illustrations of the difference between AI/GenAI stand-alone application and a AI/GenAI system that’s integrated as a building block in the larger application, is comparing ChatGPT and Microsoft 365 Copilot: with ChatGPT you get a chatbot that answers your prompts and with Copilot you get contextual help by an GenAI system which helps you based on what you are exactly working on, however, this Copilot is also based on the same GPT engines that ChatGPT is based on. The difference is that when you’re using MS 365 Copilot, it has an integrated GenAI component that helps you contextualising its responses on what you are exactly working on.. (cool, right?) So, it’s gonna be very much tailored on your work. But again, both ChatGPT and MS 365 Copilot are based on the same GPT models/engines, yet, when using MS 365 Copilot, you won’t think much of it like you’re using AI/GenAI, rather you’ll think being in the context of Copilot (which, in turn, will be using GPT), that is solely a new nice feature of your coding environment.
AI systems will become more and more seamlessly integrated components of otherwise traditional and ordinary [web] applications, and that’s all thanks to the GenAI/AI APIs.