Keeping up with the latest developments in artificial intelligence (AI) is like drinking from the proverbial fire hose, as a recent 188-page overview by two tech investors Ian Hogarth and Nathan Benaich would attest.

Start-ups, major corporations, and state actors alike are rapidly expanding their work in this field, the report suggests, even if results of real-world AI applications remain uneven in many areas.

But I would argue, somewhat counter-intuitively, that the most exciting area for AI in 2021 is at the individual level, thanks to a growing ecosystem of open-source tools and affordable services that have redefined what non-experts can do with AI technology on their laptop or desktop.

Compared to a year or two ago, consumers can more easily use AI-powered products to handle a growing range of practical tasks, ranging from transcription of audio-visual recordings to “automated writing” services that can help you generate drafts for blogs or social media posts, or solve a coding problem.

 For those with basic coding skills, building and deploying your own AI prototypes can now be done with a fraction of the time and resources it used to take.

To be clear, many of these new tools are still somewhat rough around the edges. But taken as a whole, they hint at an interesting new phase to come in AI adoption.

ME, MYSELF AND A“I”

In popular culture, AI is often portrayed as a multi-purpose android with murderous intent or a sentient computer programme bent on global domination.

Modern AI services for consumers are far more modest in form and function, thankfully.

More likely than not, these AI-powered tools will come in the form of an app or desktop service that you’ll pay a few dollars a month to use. These services won’t have a wide-range of functions but they will be exceedingly good at one or two key tasks.

The best example of this new class of consumer AI products is Otter.ai — a service for turning audio or video recordings into text transcripts, a tedious and time-consuming task that most of us are familiar with at work.

The app can’t translate between languages, or accurately summarise a two-hour Zoom meeting for you. But it excels at the narrowly-defined task of producing an accurate text transcript from recorded or live conversations.

To try the service, just sign up for a free account, upload some audio clips, and see if the results meet your expectations. You don’t need any technical knowledge, or deal with a long-drawn customisation and installation process.

Other consumer-friendly AI products of note include Copy.ai and Quillbot, two services focused on so-called “automated writing”.

These services are based on recent breakthroughs in the text-generation capabilities of large language models, and can help users generate “human-like” summaries, generic website content, or social media posts. Newer versions of these models are even capable of generating code in different programming languages from simple commands in English.

The AI writing capabilities of these apps won’t produce literary gems anytime soon. But they could go a long way in easing the workload of, say, someone tasked with updating thousands of product descriptions for an e-commerce site.

The use of these AI services in schools and the workplace raises new questions about ethics and conduct, no doubt.

But on the plus side, they represent a dramatic lowering of the barriers to entry for the public adoption of AI, a trend that is set to accelerate in the coming years.

MOVE FAST AND BUILD THINGS

For those with coding skills or prepared to learn, the technical barriers are coming down even faster.

 In the past year, we’ve witnessed significant improvements in the speed and ease with which experienced users can build new AI apps for experimentation or development, thanks to the growing array of tools and models provided by companies like Hugging Face, Gradio and Streamlit.

For instance, it took me just one morning to deploy this auto-summarisation app online.

The app looks pretty barebones, but the building blocks required — two advanced language models from Google and Facebook, a template for the user interface, and the online resources to host and run the app — would have taken days, if not weeks, to pull together in the past.

This matters on many levels, particularly in helping to demystify what current AI technologies can or cannot do when pitted against real-world problems.

At the moment, there is a huge disconnect in the broader AI community between a small minority who have direct experience with the technology and hence are aware of its practical limits, and the vast majority who are inexperienced and have highly unrealistic expectations of AI.

This gulf in expectations can’t be bridged until the first group has an easier and faster way of putting working prototypes into the hands of the second group. This is precisely what the new tools and services from Hugging Face, Gradio and Streamlit have accomplished.

 Going into 2022, no company should accept a proposed AI solution at face value. Rapid prototyping and robust testing by users should be the industry standard, not the exception.

SMART NATION, PASSIVE USERS?  

In Singapore, the country’s AI and Smart Nation blueprints tend to focus on state and enterprise-level initiatives. That’s understandable, given the broader economic stakes involved.

But such an approach also tends to result in a passive public culture, where most users just sit back and wait to be told what new AI tools to use.

If we want to boost the overall level of AI-savviness in Singapore, we'd need to explore how individuals, particularly students, can be better empowered to seek out AI services and solutions on their own.

There’s still a widespread perception that you need a lot of technical knowledge or coding skills to get into AI, or that it’s the sole domain of computer scientists.

That’s not true, and developments in the past year have shown that the barriers of entry for AI, even for casual users, have come down significantly.

And this trend will only accelerate in the coming years, given the “low code/no code” movement in the broader tech industry that allows users to build apps and tools without programming in a conventional way.

 Sweeping overviews of the AI industry may give the impression that it’s too late for the uninitiated to catch up. But the opposite is true: There is no better time than now to get started with AI.

ABOUT THE AUTHOR:

Chua Chin Hon is Lead, Artificial Intelligence Strategy and Solutions for Mediacorp News Group. He was formerly a supervising editor at TODAY and before that bureau chief in China and the United States with The Straits Times.