Meta has pulled its Muse Image artificial intelligence tool after a backlash over the feature, the second security controversy to hit the company in as many weeks, following the WhatsApp username rollout.
But Meta is not alone. Big Tech's AI ambitions keep running into the same wall: bold promises, rushed rollouts, and real the people whom the technology was supposed to serve. was supposed to serve.
And there are several reasons why, despite all the resources, personnel and money being invested by these organisations supposedly at the forefront of the technology, they still sometimes don't get it – a problem that stretches back well before AI became a thing.
Analysts at IntutionLabs, a California-based AI firm, acknowledged in a report this week – before Meta pulled the plug on Muse – that the "rapid embrace of artificial intelligence in enterprise settings has delivered some success stories."
"But high-profile failures and widespread underperformance have revealed profound systemic issues," they pointed out.
IntuitionLabs tagged the reasons for failures into seven categories, with three standing out: overhype and unmet expectations, lack of clear strategy or alignment, and problems with data quality and integration. Meta's Muse appears to have fallen into all three.
Historic mistakes
Some of the most notable companies have botched their AI "conveniences". McDonald's tried to use bots for its drive-through service in the US, only for one customer to be erroneously charged for 260 orders of Chicken McNuggets in 2024.
Earlier, before AI exploded onto the scene, IBM invested billions in its Watson AI for cancer-therapy recommendations that was launched amid much hype in 2013.
But independent reviews revealed that it often resulted in incorrect or unsafe therapies, as it relied on simulated data using its own rationale rather than actual patient records, forcing hospitals to abandon it. That resulted in a $4 billion loss for IBM.
Many companies lack "AI understanding", analysts at US research firm Forrester said in a recent report.
Even Google, one of the biggest and most influential tech companies in history that is the creator of generative AI major Gemini, couldn't escape the inaccuracies of its AI Overviews.
In 2024, AI Overviews was blamed for producing wrong answers, with online users deducing that the information for the responses was pulled from sites like Reddit and the satirical news site The Onion.
Those inaccuracies – sometimes comical – persist to this day. When The National asked for an update on the Argentina-Egypt quarter-final match in the World Cup on Tuesday – while the Pharaohs were up 2-1 and before la Albiceleste tied it in the 83rd minute – AI Overviews already proclaimed Egypt as the winner.
An April study from Seattle-based software development company Oumi showed that AI Overviews was still correct roughly nine out of 10 times. Google, meanwhile, has repeatedly acknowledged AI Overviews' shortcomings and has deployed several updates and safeguards. cc
"Given Google users perform searches 'googols' of times a year, this is not an inconsequential number of inaccurate and untrustworthy overviews," analysts at Oumi wrote.
Responsible rollout?
From societal and market perspectives, IntuitionLabs noted that public sentiment has cooled regarding AI promises made by corporations, while media coverage has shifted more from AI successes to failures, exploiting notable incidents such as lawsuits against Tesla's Autopilot driving program.
"The litany of AI rollout misfires carries critical implications for how enterprises should approach AI moving forward, and for the broader AI ecosystem’s trajectory," IntuitionLabs said.
experience. TheThe company has not commented on whether the tool will return in any form. Tristia Hennessey, a senior strategist at Portland-based Evolve Solutions Group, said she had "no idea" how Muse "got past a whole team of people thinking this was a good idea".
"If you think you're rolling out AI responsibly, you better be thinking beyond the best possible use case. Meta/Instagram's fiasco with their Muse tool had real-life consequences - not just someone's hurt feelings over a manipulated image," she said.
"Serious financial and health consequences for hundreds of people. And it was all legal. There's no recourse. There's no protection for victims. If you think you're safe, you're being naive."
