I’ve argued that “AI art” is doomed and I’ve put the “AI” hype into the perspective of past tech hype. Now I point out that AI itself is largely doomed.
Source: Food & Wine – Amazon Opens Its First Full-Size Grocery Store
It came out recently that Amazons just walk out service used a thousand Indian Workers to check out (pun intended) what people put in their carts because the AI that was supposed to do that and charge the customers wasn’t good enough. The “just walk out” service was supposed to get rid of it the check out by allowing people to pay when they left the Amazons physical grocery stores (yes Amazon has actual physical stores), it used cameras in the shopping carts to see what people put in the carts than automatically charged them for those items when they left the stores. 70% of the “just walk out” transactions had to be verified by human beings. In reality what “just walk out” did was outsource the check out to India. Amazon is not alone in this, Presto Automation claimed it developed an automated fast food drive through ordering system, instead they outsources the drive through ordering to the Philippians. Like Amazon’s just walk out service, 70% of the orders had to be put in by human beings. These are not alone, an MIT study fount that only 23% of wadge jobs could reliably be replaced by “AI” automation. On top of that for most jobs it cost more to replace people with “AI”. It turns out that is largely ineffectual.
Source: CTECH – The environmental pollution behind the boom in artificial intelligence
Then there is the money, specifically the money being lost. “AI” is loosing money. These systems require a lot of possessing power which requires a lot of electricity and a lot of water to keep things cool. It costs OpenAI 700 thousand dollars a day to run ChatGDP. That is 256 million dollar a years and that will only go up as more people use ChatGDP. Using lots of electricity and water has environmental costs. About 60% of electricity in the US comes from fossil fuels and fresh water is not an infinity available resources, some cities are already running out of water. These problem will not be solved in the foreseeable future. This is not sustainable, the environmental costs will only invite more scrutiny and regulation, and even the dumbest investor is only willing to loose so much money before they pull out.
What make this worse is how much human labor is involved in functioning “AIs”. “AI” requires humans to tag data (text and images) for the algorithm to be able to use that data. To prevent toxic data from being used by these algorithms and repeated by the algorithms humans have to sift through the information. Violent, toxic, and racist crap that have given the human filters PTSD, all for less than $2.00 an hour. Despite the psychological damage these people has suffered to keep racist, sexist, and toxic data from these “AIs”, these “AIs” still have racist and sexist biases.
Source: Tampa Bay Times – Pasco sheriff discontinues controversial intelligence program, court documents say
The Pasco County Sheriff’s Office in Florida tried to use an AI that was supposed to predict who was likely to commit crimes, then the police harassed and arrested people who hadn’t done anything wrong because the “AI” told them that these people had a high chance of break the law. The “AI” was wrong, meaning people were harassed for no reason other than a biased and broken “AI” and the people who trusted it. Harvey Murphy was arrested for a crime he didn’t commit because facial recognition software said he was a guy to robbed a Sunglass Hut. He was then sexually assaulted while in jail just hours before he was let go. Facial recognition technology still has trouble telling black people apart, it can’t tell white people apart either as it has about a 96%-98% failure rate. New York City Government used a chat bot on their website to help small businesses, it gave out false and harmful information telling people they can do thing that are in reality illegal. Multiple lawyers tried to use ChatGDP to write legal briefs only for ChatGDP to invent fake citations. “AI” has been hyped up far beyond its actual capabilities. This hype is harming people by leading people, including people in position of power and authority, to rely on “AIs” to do things it can not do.
These happen because there is no AI, what these are are algorithms that use statistical data to create statistically plausible results. These algorithms don’t look for factual or accurate results, just plausible results based on the statistical data it has access to. If there is any bias or flaws in that data or biases in the people who create the algorithm, the biases and flaws will be reflected in the results. This why a chat bot would create what amounts to lies, it does not look for factual results, only a plausible outcome based on the data it has. This is why an algorithm that uses police data would target minorities as there is a systematic bias against minorities in the policing and legal system, any data gathered from the legal and policing system is going to be biased and the out come from an algorithm that uses that data would inherit that bias. Because the data these algorithms use are created by, or gathered by people the biases in society and culture will be reflected in the results these algorithms create. Our societal, cultural, and human flaws will be reflected in any AI we create, real AIs or algorithms, as a result they would be as flawed as we are.
Of course there will be the response: the tech is new, it is still advancing. Self driving cars have been around the corner for twenty years and all we have gotten are accidents and dead people from them. Self driving cars have been ten years away since the 00s and they are still ten years away. As I have pointed out elsewhere there have been many technologies that have been “the next big thing” only to disappear or take far, far longer to get where people thought they would go. Hype can make people a lot of money, and “AI” has a lot of hype. Companies like OpenAI have made a lot of money in the form of investments from that hype, money that goes to salaries for people who work for OpenAI. Hype lies and hype masked flaws in technologies. These so called “AIs” are an extremely flawed technology and not nearly as capable as the hype presence them as. I’m not saying these are useless, they have uses but the reality of “AIs” is they are not nearly as capable as people think they are and are not profitable, so “AI” is, largely, doomed.
If you agree with me, just enjoined what I had to say, or hate what I had to say but enjoyed getting angry at it, please support my work on Kofi. Those who support my work at Kofi get access to high rez versions of my photography and art.
C.L. Who Writes 