How companies can beat the ChatGPT hype with ‘actionable AI’

New products like ChatGPT have dazzled people, but what will real money-making applications look like? Will they offer sporadic business success stories buried in a sea of ​​noise? , or are we at the beginning of a true paradigm shift? What does it take to develop a working AI system?

To chart the future of AI, we can draw valuable lessons from the big data era, the step-change advancement of the technology that precedes it.

2003-2020: Big Data Era

The rapid adoption and commercialization of the Internet in the late 1990s and early 2000s created and lost wealth, laid the foundations of corporate empires, and fueled exponential growth in web traffic. This traffic generated logs, which turned out to be a very useful record of online actions. We quickly learned that logs help us understand why software breaks and what combinations of behaviors lead to desired actions, such as purchasing a product.

With the rise of the Internet and the exponential growth of log files, many of us felt we had something of great value, and the hype machine increased to 11. However, it was still unclear whether that data could actually be analyzed and transformed into sustainable data. Its value increases, especially when the data is spread across many different ecosystems.

Google’s big data success story is worth revisiting as it epitomizes how data has transformed Google into a trillion-dollar company that changed markets forever. Google’s search results have been consistently good and we’ve built trust, but until Adwords allows us to monetize, Google’s search at scale or any additional services that currently rely on Google could not continue to provide Today, we all expect to find exactly what we need in seconds, complete turn-by-turn directions, collaborative documents, and cloud-based storage.

Innumerable wealth has been built on Google’s ability to turn data into compelling products. And many other giants, from the rebooted IBM to the new Goliath his Snowflake, have built successful empires by helping organizations acquire, manage, and optimize their data.

What started as confusing chatter ended up with huge financial benefits. It is precisely this path that AI must follow.

2017-2034: AI Era

Internet users can create large amounts of text written in natural languages ​​such as English and Chinese for use as websites, PDFs, blogs, and more. Thanks to big data, storing and analyzing this text is easy. Researchers will be able to develop software that can read all of that text and learn to write it. When his ChatGPT came out in late 2022, parents called their kids asking if the machine had finally come to life.

This is a turning point in the field of AI, in the history of technology, and perhaps in the history of mankind.

The level of hype for AI today is exactly the same as it was for big data. A key question that the industry must answer is how AI can deliver the sustainable business outcomes that are essential to propelling this massive change forever.

Workable AI: Let’s move AI

To find viable and valuable long-term applications, an AI platform must incorporate three key elements.

1. Generative AI model itself
2. Interfaces and business applications that allow users to interact with models. This could be a standalone product or a back office process enhanced with generative AI.
3. A system for ensuring the reliability of the model. This includes the ability to continuously and cost-effectively monitor model performance and teach the model to improve its response.

Just as Google combined these elements to create actionable big data, AI success stories must do the same to create what I call actionable AI.

Let’s take a look at each of these elements and the current situation.

Generative AI model

Generative AI is unique in its wildness, posing challenges of unpredictable behavior and requiring continuous education to improve. Bugs cannot be fixed like traditional procedural software. These models are software built by other software, made up of hundreds of billions of equations that interact in ways we can’t comprehend. I just don’t know what weights between which neurons should be set to what values ​​to prevent the chatbot from telling the journalist to divorce his wife.

The only way these models can be improved is through feedback and more opportunities to learn what good behavior looks like. To avoid the catastrophic hallucinations that can keep potential customers from using your model in a high-stakes environment where real money is spent, you should always be vigilant about data quality and algorithm performance. Essential.

building trust

Governance, transparency, and explainability enforced by actual regulation will give companies confidence that they can understand how AI behaves when mistakes are inevitable, so they can limit the damage and work to improve AI. essential for There is much to admire in the industry leader’s early move to create thoughtful guardrails with real teeth, and I urge the swift introduction of sensible regulation.

In addition, require AI-generated media (text, audio, images, video) to be clearly labeled as “made with AI” when used in commercial or political contexts . Just like nutrition labels and movie ratings, consumers have a right to know what interests them. And I believe many people will be pleasantly surprised by the quality of AI-generated products.

killer app

Hundreds of companies emerged within months offering applications for generative AI, from creating marketing collateral to producing new music to creating new drugs. ChatGPT’s simple prompts could potentially surpass search engines in the age of big data, but many more could be just as powerful and profitable across a variety of verticals and applications. Using ChatGPT has already been proven to significantly improve coding efficiency. What else will follow? To create Workable AI, experimentation to find AI applications that make a big difference in user experience and business performance is essential.

Companies that make their fortunes with this new kind of technology will break down these innovation barriers. They solve the challenge of continuously and cost-effectively building trust in AI while developing killer apps combined with healthy monetization built on a strong foundational model.

Big data has gone through the same cycle of noise and nonsense. Equally, it may take several generations and failures, but by focusing on the Workable AI philosophy, this new field will evolve rapidly and become as transformative a platform as experts expect. will create

Florian Douetteau is the CEO of Dataiku.