Meta has discontinued its artificial intelligence tool Muse Image following backlash against the feature. This is the second security controversy to hit the company in a few weeks, following the release of WhatsApp’s usernames.
But Meta is not alone. Big Tech’s AI ambitions continue to run into the same walls: bold promises, rushed deployments, and the realities of the people the technology is supposed to serve. I was supposed to serve.
And there are several reasons why these organizations that are supposed to be at the forefront of technology, despite investing all their resources, people and money, still sometimes don’t get it. This problem has been around since long before AI became widespread.
Analysts at California-based AI firm IntutionLabs acknowledged in a report this week that “the rapid adoption of artificial intelligence in enterprise environments has led to some success stories” before Meta stopped using Muse.
“However, high-profile failures and widespread underperformance have exposed serious systemic problems,” they noted.
IntuitionLabs breaks down the reasons for failure into seven categories, and three stand out: excessive hype and unmet expectations, lack of clear strategy or alignment, and data quality and integration issues. Meta’s muse seems to be all three.
historical mistake
Some of the most notable companies have screwed up the “convenience” of AI. McDonald’s tried using bots for its drive-thru service in the US, but in 2024, only one customer was incorrectly charged for an order of 260 Chicken McNuggets.
Before AI exploded, IBM invested billions in Watson AI, which was launched to much fanfare in 2013 and aimed at making cancer treatment recommendations.
However, an independent review found that relying on simulated data with its own rationale rather than actual patient records often resulted in incorrect or unsafe treatments, forcing hospitals to abandon treatments. As a result, IBM suffered a loss of $4 billion.
Analysts at US research firm Forrester said in a recent report that many companies lack “understanding of AI.”
Even Google, one of the largest and most influential technology companies in history and developer of generative AI giant Gemini, couldn’t escape the inaccuracies of its AI overview.
In 2024, AI Overviews was accused of creating false answers after online users assumed that the information in the answers came from sites like Reddit and satirical news site The Onion.
Their inaccuracies persist to this day, sometimes comically. when The National We asked for an update on Tuesday’s World Cup quarter-final match between Argentina and Egypt – Pharaohs won 2-1 before the end of the match.albiceleste They equalized in the 83rd minute – AI Overovers had already declared Egypt the winners.
A study conducted in April by Oumi, a Seattle-based software development company, showed that AI summaries were still correct about nine out of 10 times. Meanwhile, Google has repeatedly acknowledged the shortcomings of AI Overviews and introduced several updates and safeguards. cc
Oumi analysts said, “This is a non-trivial number of inaccurate and unreliable summaries, given that Google users ‘Google’ their searches many times a year.”
Responsible development?
From a societal and market perspective, IntuitionLabs noted that while public sentiment toward corporate AI promises has cooled, media coverage has shifted from AI successes to failures, leveraging high-profile cases such as the lawsuit over Tesla’s Autopilot driving program.
“Continued failures in AI deployments have significant implications for how enterprises should approach AI going forward and the trajectory of the broader AI ecosystem,” IntuitionLabs said.
experience. The company has not commented on whether the tool will return in some form. Tristia Hennessy, a senior strategist at Portland-based Evolve Solutions Group, said she has “no idea” how Muse “surpassed a whole team of people who thought this was a good idea.”
“If you think you’re deploying AI responsibly, you need to think beyond the best possible use case. Meta/Instagram’s fiasco with the Muse tool had real-world consequences, not just someone being hurt by a doctored image,” she said.
“Hundreds of people suffered severe financial and health consequences, and it was all legal. There is no remedy. There is no protection for the victims. If you think you are safe, you are naive.”
