Connect with us

International Circuit

AI washing could leave a big stain on the industry

You’ve heard of greenwashing and perhaps even cloud washing. Now, AI washing is the latest trend taking the tech and telecom industry by storm as companies rush to cash in on the artificial intelligence (AI) craze. But little white lies about what is AI — and more importantly, isn’t — could leave a stain that’s impossible to scrub out.

Sid Nag, Gartner VP of cloud services and technologies, told Silverlinings it’s “inevitable” there will be some degree of AI washing from vendors as they scramble to grab wallet share from buyers.

“Vendors and providers have to be extremely careful not to overplay the AI hand without concrete use cases to demonstrate the real value of AI and the pedestrian use of AI in their product offerings as opposed to referencing AI as some sort of an esoteric technology or it will go by the way of technologies such as Web3 or Metaverse I might dare say,” he said.

The term AI washing refers to the practice of calling something AI when it really isn’t. Or, using the term to hype up old tools that have been used in networks for years now when what people really mean by AI is technology that can take decisions out of human hands entirely.

In essence, AI becomes just another marketing term rather than a technology with concrete parameters and measurable — dare we say transformative — results.

Beyond being just off-putting and potentially souring buyers on a technology that has true potential, Nag said there are other pitfalls to AI washing too, including “issues related to copyright, ethics and legal.”

Bret Greenstein, data and AI leader at PwC, agreed. In addition to serious problems like hallucinations, bias and privacy issues, fake AI can lead companies to waste investments on solutions that fall short rather than “allocating resources to technologies that meet enterprise security, functional, and outcome expectations.” He added they may end up having to spend more than originally intended to get a solution that actually works.

Overall, the risk is that AI washing could “undermine customer confidence and cool interest if enough people are disappointed in fake AI solutions. When businesses invest in these solutions, they often have high expectations regarding the value and safety they will provide. However, when these expectations are not met, it can lead to a loss of trust and future investments being impacted,” he added.

Hanging out to dry
Danielle Royston, CEO of cloud consulting firm TelcoDR and acting CEO of telecom software company Totogi, said she expects AI washing to be rampant at Mobile World Congress Barcelona next week and urged companies heading to the show to bring a big shaker of salt with them. If our inbox is any indication, she’s bang on the money.

“I think because enterprise organizations and telcos are broadcasting that they plan to spend on AI, for vendors, they’re like ‘sh*t I need to put AI in my product ASAP,’” Royston told Silverlinings. “So, I think people are just issuing press releases without a lot of detail and they’re like ‘we’ll figure it out but we need to message to customers that we have it.’”

But that can lead to huge disappointments for customers, she added.

“Judgment will be swift and harsh,” Royston said.

Royston said that as CEO of Totogi, she has pushed the company to lean into an AI-first approach in part because it has spent the past several years building its own AI tools for the telco BSS environment. Its AI-first mission has seen Totogi experiment internally with a lot of AI tools, and not all of the them hit the mark.

For example, Royston said her staff tried to use the Einstein tool offered by Salesforce when it first came out but found it to be “lame.” They waited a few months and trued again but it was “still bad.” So, her team built their own AI tools that pull data from Salesforce to accomplish what they wanted to do.

But Royston said it shouldn’t have to be that way. It’s better, she said, if the vendors can build adequate AI into their own systems.

Spot the fake
“Savvy buyers will demand proof of business outcomes and KPIs and metrics as part of their buying decision to detect fake AI claims,” noted Nag.

But Royston offered more detail. She said buyers can specifically ask the vendor to distinguish what type of AI their solution uses (copilot vs. predictive analytics, for instance), question what large language model the AI was trained on if it’s a copilot, ask for metrics around how much the solution can increase productivity and revenue, and seek answers to questions about grounding, protection of proprietary data and the costs associated with tasks such as API calls. All of these should have answers, she said.

Additionally, Greenstein advised buyers to carefully scrutinize a product’s history.

“In most cases, products that existed before November 2022 have simply incorporated a GenAI tool as a superficial enhancement, giving the illusion of modernity without truly transforming the product,” he explained. “It is rare to find products that have undergone a complete re=engineering to fully embrace Generative AI.”

For those products that have truly incorporated generative AI, Greenstein said prospective customers should probe exactly how it is being used, and get detail about things like data privacy protections, whether the model it is based on is open source or commercial, safeguards to ensure compliance with local AI regulations, how the product will integrate into enterprise architecture, and how it aligns with other solutions from major cloud and enterprise providers.

More than anything else, though, Royston said “seeing is believing.” That means asking the vendor to show you the AI in action, preferably via a live (rather than canned) demo. More advanced solutions will also have customer proof of concept studies for prospective clients to review, she concluded.

Happy hunting! SilverliningsInfo

Click to comment

You must be logged in to post a comment Login

Leave a Reply

Copyright © 2024 Communications Today

error: Content is protected !!