Dear Friends,
On March 7, 2023, I wrote an article on “The Changing Driving Forces of the Indian Marriage System!” (www.linkedin.com/pulse/changing-driving-forces-indian-marriage-system-saripalle/). Today, on average, any successful matchmaking passes through 15 parameters such as family history, horoscope, star, caste, nature of the job, package, location, appearance, social habits, the boy’s/girl’s previous social background, influences, the financial position of both families, educational background of both families, and many more. I am only taking basic parameters; in some cases, they go up to 32.
15–20 years ago, the conditions were heavily weighted toward the boy’s side, but today they are heavily weighted toward the girl’s side. Thanks to social media, both pains and perspectives are shared openly and publicly. Earlier, due to ego, they were either within the house or inside the heart. They were not publicly shared. Today, Gen Z kids are openly sharing those views and highlighting the flaws in the system.
A few days back, there was an article in the Hindustan Times. A woman lost 1 crore rupees or £100,000 in London, moved to Australia due to family constraints, worked as a product manager, and was fired due to market conditions. She now works as a cleaner and housekeeping manager at Airbnb houses. I appreciate this woman’s confidence and dignity of labor. In fact, many great founders did the same. Jan Koum worked as a grocery store cleaner in California during his youth to support himself and his mother after immigrating to the United States from Ukraine.
A married man or woman can publicly write, but can a prospective (yet to get married) boy or girl who is on the verge of the process confidently write in the current circumstances? I doubt it. Adding to this complexity, AI has now been added to these 15–30 parameters. To date, parents have not resorted much to AI for background verification, but thanks to free GPT tools, they are now asking AI for help as well.
Yesterday, there was an article in Moneycontrol. The summary of the article is: “When a father Googled his daughter's potential match, Google AI confidently declared the CEO already married to a woman named Joya. The problem? She does not exist.” He wrote, “I almost lost a marriage prospect to an AI hallucination. Her father did what every Indian father does before the match—he Googled me.” (https://www.moneycontrol.com/news/trends/google-ai-invented-a-wife-for-me-how-an-ai-hallucination-nearly-ruined-this-ceo-s-marriage-prospect-13912953.html)
This is the result of hallucination. What if this hallucination goes the other way, stating something positive about the prospect? That is also not correct. Where do LLMs get most of their data? Reddit. About 40.1% of references are now from Reddit. This has outpaced Wikipedia (26.3%) and YouTube (23.5%), among others. LLM output is shaped by fast-moving discussions, real user reviews, and the latest opinions. Trust is increasingly crowdsourced.
In this context, if a boy's or a girl's data is widely discussed on social media, their profile can be trained, and LLMs may provide more accurate information about that person. If the data is skewed, the answers will also be skewed.
So far, AI's impact has largely been on jobs. Now, slowly, it is entering personal lives as well. We need to do something for the age group of 1–8. According to the National Center on Birth Defects and Developmental Disabilities, the first 8 years can lay the foundation for future learning, health, and life success.
We need to initiate a special project for this age group. Next time, I will write about this project.
Ravi Saripalle
No comments:
Post a Comment