Transparent ‘Nutrition Labels’ Could Be Key to Scaling Generative AI

Generative artificial intelligence (AI) has been billed as an everyday technology.

And while consumers may already be reaping the technology’s benefits, there is a gap between consumers’ confidence in their knowledge of AI and the ongoing realities of its almost invisible integration into their daily lives.

“What is key for most consumers is knowing that [AI] goes well beyond just a large language model [LLM], it goes well beyond what you’re sort of seeing at the surface, and it’s been touching and permeating a lot of parts of your daily lives for years,” Shaunt Sarkissian, founder and CEO of AI-ID, told PYMNTS CEO Karen Webster.

Still, when you ask consumers the role AI’s capabilities play in powering their lives today, many struggle to explain where the technology sits and what it impacts.

Then again, who among us can speak to the intricate payment networks at play that spring into action for routine and simple on-the-surface tasks like buying a coffee or shopping online?

“The same way that the cloud is something new, how digital payment networks are interconnected beneath the surface, that’s how consumers’ understanding of AI relates to a lot of other technologies,” said Sarkissian. “But consumers are smart, they can handle the truth as long as you tell them what’s going on.”

That is why it is so important for AI firms to be transparent and tell them.

Knowing What AI Is Makes End-Users More Comfortable Trusting It 

While consumers generally think that AI can improve their daily lives, there exists an undercurrent of the unknown to their perception of the innovative technology — and that undercurrent is centered around doubts that AI will provide them with the right information 100% of the time. There’s also the uncertainty over whether the information AI platforms provide is safe and reliable, particularly when it comes to sensitive areas like banking and healthcare.

“The food industry was the first sector to really start adopting things like disclosure of ingredients and nutrition labels, providing consumers transparency and knowledge of what’s in their products — and with AI, it’s much of the same. Companies need to say, ‘Look, this was AI-generated, but this other piece was not,” Sarkissian said. “Get into the calorie count, if you will.

“As long as you can tell consumers what the content is made of, they can then choose to make decisions around that information based on what they see. But if you don’t give that to them, then it’s that shielding and blackboxing that the industry needs to be careful with, and where regulators can step in more aggressively if the industry fails to be proactive,” he said. “It is going to be critical for areas like intellectual property.”

Particularly as AI continues to evolve and integrate into various industries, it is crucial for consumers to have a clear understanding of its capabilities and potential risks. Transparency and disclosure will play a vital role in building trust and ensuring that consumers can make informed decisions in an increasingly AI-driven world, as well as help maintain a productive balance between public oversight and private innovation.

AI Shines in Data-Rich Environments 

Already, AI is showing its ability to outperform humans in certain tasks when measured on a scale and speed basis.

“If it is a data-intensive industry with a large data set, AI will perform very well. And you just need to be aware of, and wary of, that,” explained Sarkissian, noting that AI generated outputs may not perform as well when tasked with taste-based prompts, such as being asked to “make a painting that would sell in a gallery.”

Still, it is one thing to have a good idea — it is another thing entirely to go and execute it successfully, which is something that still requires a decidedly human touch.

That’s why, said Sarkissian, AI models are already being specialized to the degree that they can generate “good ideas” at scale for specific industries that are then able to be passed off to capable experts for execution.

“AI is all based on the training, so if you feed it a very specialized field of data, the chance of it becoming an expert or being the best in that particular category is very high,” he said. “And what might occur over time is that one large AI model might source information for multiple, specialized sub-models.”

Call it GPT Orchestration. But despite the promise, training an AI model to an expert-level degree of specialization is incredibly expensive, laborious and time-consuming.

“There really needs to be commercial utility, it has to be something like recognizing a tumor in an image,” said Sarkissian.

And he views healthcare as a promising opportunity area for AI applications to enhance care delivery by making doctors and provider systems more effective and efficient at scale.

“An AI-enabled diagnostic tool can be better at identifying things earlier in the prognosis around cancer than the human eye can, and those things are saving lives and saving money as well,” Sarkissian said.

It isn’t just medicine where specialized AI can have a game-changing impact.

“A lot of the commoditized part of business law, transactional law, even real estate law is going be automated in my mind, where it can process documents and give lawyers back hours and days of their time, making them hyper-efficient,” Sarkissian said.

Still, AI isn’t perfect — and today’s models still have a strong tendency to hallucinate and present end-users with misinformation or wrong results. That’s why, as mentioned above, transparency and accountability are crucial for the go-forward scalability of the still-nascent industry.