Some recent AI misfires remind us of what the novel/movie/TV series “Westworld” and its writer/director Michael Crichton warned long ago: What could go wrong, wrong, wrong?
In recent weeks, artificial intelligence (AI) has endured some high-profile glitches from subways to transcription services that have thrown cold water on its prodigious potential and given pause to those not yet ready to embrace AI unconditionally.
The most-glaring occurred on some New York City subways, where a pilot program testing AI-powered weapons scanners this summer failed to pick up any passengers with firearms, according to police data released recently.
Compounding matters, the system made more than 100 times false alerts through nearly 3,000 searches. Last week, the New York Police Department said it performed 2,749 scans over 30 days at 20 stations. There were 118 false positives, a rate of 4.29%.
Embattled New York Mayor Eric Adams, a tech enthusiast, announced the portable-scanner pilot program with vendor Evolv to help deter what little violence is on the subway system.
But the program has drawn pushback from some riders and civil liberties groups, who believe it neither feasible nor constitutional to scan millions of riders daily.
Meanwhile, users of OpenAI’s Whisper audio transcription tool claim it often includes hallucinations in its results, usually in the form of inaccurate text passages. A University of Michigan researcher discovered garbled text in 80% of Whisper’s transcriptions.
In one recording, a speaker said, “Two other girls and one lady.” But the AI tool transcribed it as, “Two other girls and one lady, um, which were Black.” Another example mangled a medical expression into “hyperactivated antibiotics,” which do not exist.
Finally, AI wasn’t to blame for Delta Air Lines’ cancellation of 7,000 flights this summer, but the airline is suing security software maker CrowdStrike for breach of contract and negligence. According to the lawsuit filed last week, a software update snafu disrupted Microsoft Corp.’s Windows operating system, costing Delta $380 million in reduced revenue and $170 million in costs. Delta is seeking damages to cover its losses, as well as litigation costs and punitive damages.
“CrowdStrike caused a global catastrophe because it cut corners, took shortcuts, and circumvented the very testing and certification processes it advertised, for its own benefit and profit,” Delta said in the complaint. “If CrowdStrike had tested the Faulty Update on even one computer before deployment, the computer would have crashed.”
Some developers, however, aren’t blaming the technology altogether.
“Delta’s lawsuit against CrowdStrike feels like a ‘self-own’ once you dig into the details,” Richard Bird, chief security officer at Traceable, said in an email. “Every solution provider and customer agrees to liability terms when they agree to work together. Delta isn’t pursuing a contract remedy here — they are calling CrowdStrike negligent when it is clear in their own court filings that CrowdStrike was a catalyst and the root cause for the outage was antiquated technology and a completely non-functional business recovery plan. Delta failed when it came to resiliency, and would have at some point with or without CrowdStrike’s help.”
