The new products such as ChatGPT have caught the general public’s attention; however, what will the actual applications for making money be? Do they provide occasional business successes lost in the sea of noise, or are we beginning a paradigm change? What steps are needed to create AI systems that can be used?
It is possible to draw lessons from the prior technological breakthrough of the Big Data era to map the future of AI.
The Big Data Era
The rapid growth and commercialization of internet technology in the 1990s and early 2000s created and destroyed fortunes, laid the basis of corporations’ empires, and led to an exponential increase in website traffic. The traffic generated logs, which became extremely useful records of online activities. We soon realized that logs can help us understand why software fails and what combination of actions can lead to desired actions like buying an item.
The size of log files increased exponentially as they grew online, many of us believed that we had discovered something incredibly useful, and the excitement machine went up to 11. However, it remains to be determined if we can analyze the data to make it sustainable, especially when the data is spread across multiple ecosystems.
Google’s big data story is worth reliving as a reminder of how data helped it become an enormous company with a billion dollars that changed the market for good. Google’s search results were always good and helped build trust. Still, the company wouldn’t keep providing the ability to search on a massive size — or even all the other products we rely on Google for in the present before Adwords made it possible to monetize. Today, we want to locate exactly the information we need in just a few minutes, perfect directions for turning, collaborative documents, and cloud-based storage.
Numerous fortunes have been made on Google’s ability to transform data into products that are compelling, as well as a host of other titans, ranging from the re-launched IBM to the latest goliath named Snowflake and Snowflake, have created successful empires through aiding organizations to collect information, optimize and manage it.
What appeared to be a bit of nonsense initially eventually brought huge financial gains. This is exactly the path AI has to follow.
The AI Era
Internet users have written enormous amounts of text in natural languages such as English or Chinese, available as blogs, websites, PDFs, and much more. Due to the huge amount of data available for storage and analysis, storing and organizing the text is simple for researchers to create software that reads all this text and then teaches itself to write. Then, fast forward to ChatGPT coming in the latter half of 2022 and parents contacting their children to ask whether the machine had been able to come to life.
It’s a turning point in the area of AI in the development of technology and, perhaps, in the history of humanity.
Today’s AI excitement levels are exactly where we were in the past with huge data. The main question that the industry has to answer is: how can AI deliver the long-lasting business outcomes necessary to make this change benefit everyone?
Effective AI Let’s use AI to work
To discover viable, useful long-term applications, AI platforms should be able to incorporate three key aspects.
The AI models themselves are generative AI models themselves.
Business applications let users communicate with models that could be a stand-alone product or an AI-generated back office process.
A system that ensures confidence in the model, which includes the capability to continuously and effectively monitor the performance of a model and instruct the model to improve its capabilities.
Similar to how Google combined these elements to create workable big data, AI successes must achieve the same feat to produce what I call Workable AI.
Let’s take a look at the various elements and the current state of affairs:
Generative AI models
Generative AI is distinctive in its wildness, presents new challenges in unpredictable behavior, and requires constant training to improve. We can’t fix bugs the way we can with traditional procedural software. These models are programs that other programs have created, made up of hundreds of billions of equations that interact in ways we cannot comprehend. We’re unsure which neuronal weights need to be set to what values to stop chatbots from telling a reporter to break up with his spouse.
The only way the models can be improved is through feedback and additional opportunities to discover the characteristics of good behavior. A constant eye on the quality of data and algorithm performance is vital to avoid causing hallucinations which could deter prospective customers from using models in high-risk situations where the real money is being spent.
Building trust
Transparency, accountability, and governance, which are enforced by actual regulation, are vital for companies to have confidence that they can comprehend what AI will be doing when mistakes happen to limit the harm and enhance the AI. There’s plenty to appreciate in the initial efforts of the industry leaders to develop well-thought-out security measures backed by real teeth, and I encourage the rapid implementation and implementation of intelligent regulations.
I’d also like any media (text or audio, picture, and video) produced through AI to be identified with the words “Made with AI” when employed in a political or commercial context. Similar to the labels for nutrition or movie reviews, consumers need to be aware of what they’re buying -I’m convinced that many people will be amazed at the superiority of products created by AI.