This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy.
AI naysayers believe that this nascent technology is too temperamental at this stage to be truly revolutionary, relying on anecdotes of various AI-based Large Language Models (LLMs) hallucinating to bolster their case. However, one tech-savvy netizen from Russia just used OpenAI’s GPT-4 to find the perfect date on Tinder, illustrating the truly revolutionary characteristics of this emerging paradigm shift in AI’s use cases.
Сделал предложение девушке, с которой ChatGPT общался за меня год. Для этого нейросеть переобщалась с другими 5239 девушками, которых отсеила как ненужных и оставила только одну. Поделюсь, как сделал такую систему, какие были проблемы и что вышло с остальными девушками. Тред pic.twitter.com/fbVO7OmZhF
— Aleksandr Zhadan (@biblikz) January 30, 2024
After enduring a breakup in 2021, Aleksandr Zhadan, began actively dating women in Moscow and St. Petersburg only to see his efforts end abruptly each time. In 2022, Zhadan gained access to OpenAI’s GPT-3 API. He then developed a bot that would automatically swipe right on profiles with at least two pictures.
The tech-savvy Russian integrated GPT-3 API with his bot, allowing the LLM to chat with prospective dating candidates without his intervention. Zhadan used the following prompt for this task:
“You’re a guy talking to a girl for the first time. Your task, not immediately but to invite the girl on a date.”
This first attempt yielded less-than-stellar results, with the LLM often forgetting information conveyed in previous messages within the same conversation. This made Zhadan look like a “dory-pecked fish” to the offended dating candidates.
Zhadan’s dating fortunes improved markedly when OpenAI released ChatGPT, bringing a dedicated memory cache to the table. He trained his model so that it could write more naturally and installed additional filters to prevent the LLM from sending “cringe” stories. To better target his preferred dating vectors, Zhadan trained his bespoke bot on over 4,000 profiles that he had already swiped right on by integrating Torchvision with the web version of Tinder.
In fact, the Russian programmer was able to filter out most of the noise to specifically target his preferred vectors by using an exhaustive list of filters. Of the 353 matched Tinder profiles, the LLM was able to chat with 160 (45 percent), resulting in 12 dates.
Zhadan continued to refine his approach by asking ChatGPT to summarize his chats with each dating vector on both Tinder and Telegram. He also asked the LLM to build a profile of each girl that he actually dated.
Eventually, in what Zhadan defines as the V2 of his dating bot, he integrated Google Calendar and Telegram with GPT-4. The bot could now set up dates autonomously. So, what was the result? Well, our prolific Casanova ended up with 4,886 matched profiles on Tinder. He dated up to 4 times per day until he met with Karina – his current love interest. Yay for technology!