New products like ChatGPT have captured the public’s attention, but what are the genuine money-making applications? Will they provide infrequent corporate success stories lost in the shuffle, or are we witnessing the beginning of a major paradigm shift? What will it take to create AI systems that can truly be used?
To predict AI’s future, we can learn from the previous technological step-change advance: the Big Data era.
The Big Data Era, 2003-2020
The internet’s quick acceptance and commercialization in the late 1990s and early 2000s created and destroyed fortunes, provided the groundwork for corporate empires, and fuelled exponential rise in web traffic. This traffic generated logs, which proved to be a very helpful record of internet activities. We rapidly discovered that logs aid in understanding why software fails and which combinations of behaviours result in desirable actions, such as purchasing a product.
As log files swelled dramatically with the emergence of the internet, most of us realised we’d stumbled across something extremely significant, and the hype machine was turned up to 11. However, it remained to be seen whether we could properly analyse that data and transform it into long-term value, especially given that the data was spread across multiple ecosystems.
Google’s big data success story is worth revisiting as an example of how data converted it into a trillion-dollar firm that forever changed the market. Google’s search results were consistently outstanding, which established confidence, but the business couldn’t keep offering search at scale — or all of the other things we rely on today — until Adwords enabled revenue. We now expect to be able to find exactly what we need in seconds, as well as receive precise turn-by-turn directions, collaborative papers, and cloud-based storage.
Countless riches have been established on Google’s ability to turn data into appealing products, and many other titans, from a rejuvenated IBM to Snowflake’s new goliath, have built lucrative empires by assisting organisations in capturing, managing, and optimising data.
What began as incomprehensible gibberish eventually yielded enormous financial gains. This is the path that AI must take.
The Artificial Intelligence Era, 2017-2034
Internet users have created huge amounts of content written in natural language, such as English or Chinese, and made it available as webpages, PDFs, blogs, and other formats. Because of big data, storing and analysing this text is simple, allowing researchers to create software that can read all of it and educate itself to write. Fast forward to late 2022, when ChatGPT is released, and parents phone their children to see if the machines have finally come to life.
It is a watershed point in the history of artificial intelligence, technology, and possibly humanity.
Today’s AI hype levels are identical to those of big data. The crucial challenge for the industry is: How can AI produce the long-term business outcomes required to propel this paradigm shift forward?
AI that works: Let’s put AI to work.
AI platforms must incorporate three critical features in order to discover practical, valuable long-term applications.
- The models of generative AI themselves
- The interfaces and business applications that will allow users to interact with the models, whether as a standalone product or as part of a generative AI-augmented back office process.
- A system for ensuring model trust, including the capacity to continuously and cost-effectively monitor a model’s performance and train the model to improve its responses.
- Just as Google brought these aspects together to generate workable big data, AI success stories must do the same to develop what I refer to as Workable AI.
Let’s take a look at each of these components and where we are now:
AI models that generate data
In its wildness, generative AI is unique, posing issues of unexpected behaviour and requiring continuous education to progress. We cannot fix defects in the same way that we would in traditional, procedural software. These models are constructed by other software and are made up of hundreds of billions of equations that interact in ways we don’t understand. We just do not know which weights between which neurons must be set to which values in order to prevent a chatbot from telling a journalist to divorce his wife.
These models can only grow with feedback and more opportunities to learn what excellent behaviour looks like. Constant monitoring over data quality and algorithm performance is required to avoid damaging hallucinations that might turn off potential consumers from employing models in high-stakes scenarios where real money is invested.
Building trust
Governance, transparency, and explainability must be enforced by actual regulation to give businesses confidence that they can comprehend what AI is doing when mistakes arise, allowing them to reduce the harm and seek to improve the AI. There is much to praise in industry leaders’ initial steps to develop reasonable guardrails with genuine teeth, and I urge swift implementation of smart legislation.
Furthermore, any media (text, audio, image, or video) generated by AI must be explicitly labelled as “Made with AI” whether utilised in a commercial or political setting. Consumers ought to know what they’re getting into, just like they do with nutrition labels or movie ratings, and I believe many will be pleasantly surprised by the quality of AI-generated items.
Killer apps
Hundreds of businesses have sprung up in recent months to provide generative AI applications ranging from marketing collateral to new music to new pharmaceuticals. The simple prompt of ChatGPT might theoretically outperform the search engine of the Big Data Era — but many more applications in many verticals and applications could be just as powerful and profitable. We’re already witnessing significant gains in coding productivity thanks to ChatGPT. What else will happen? Experimenting to uncover AI applications that give a significant improvement in user experience and business performance will be critical to developing Workable AI.
Companies who will make their fortunes from this new class of technologies will be the first to break past these innovation hurdles. They’ll overcome the difficulty of consistently and cost-effectively increasing trust in AI while delivering killer apps with sound monetization based on powerful underlying models.
Big data went through the same cycle of noise and nonsense. Similarly, it will most likely take a few generations and blunders, but by focusing on the fundamentals of Workable AI, this new discipline will soon evolve to build a game-changing platform that will be just as disruptive as experts anticipate.