Is OpenAI's ChatGPT Becoming Slower? Know Here

OpenAI's ChatGPT has transformed the AI landscape since its inception last year with human-like question-and-answer sessions with a chatbot. Some business experts believe ChatGPT is poised to threaten Google's search engine market dominance. However, as the generative AI explosion continues to acquire widespread acceptance, there is a rising fear among many users that ChatGPT is becoming incredibly slow and far less capable of giving accurate responses. This is leading to scepticism.

Performance of OpenAI

The performance of OpenAI's LLMs has drastically declined over time, according to Stanford and UC Berkeley researchers, making ChatGPT less precise. The research team examined the GPT-3.5 and GPT-4 versions of ChatGPT from March 2023 and June 2023 on four essential tasks. They created benchmarks to measure the model's aptitude for tackling mathematical difficulties, responding to delicate inquiries, producing code, and judging the models' capacity for visual reasoning.

The tests showed a considerable decline in performance over time between versions. In March, ChatGPT-4 had an accuracy rate of 97.6% for finding prime integers. However, ChatGPT's accuracy fell to 2.4% in June after only answering 12 questions correctly. ChatGPT once again performed poorly when producing a few lines of code. The investigation also discovered that ChatGPT's more recent version would not respond to some delicate queries.

According to Santiago Valderrama, an AI expert, "GPT-4 is getting worse over time, not better," on Twitter. He stated there are rumours that they are using several smaller, more specialised GPT-4 versions that function similarly to larger ones but are less expensive to operate. As a result, ChatGPT's performance has been worse during the past few months.

GPT-4's accuracy and performance: User's Verdict

Many users have recently complained on social media and online forums that GPT-4's accuracy and performance are worse than GPT 3.5 and the LLM used to power programmes like ChatGPT. According to one argument, businesses like OpenAI need to release the most advanced chatbots to the general public because the expense of operating the systems is so high. But despite pressure from several prominent AI figures who support slowing down AI research, the industry shows no signs of slowing down.

Microsoft invested enormous sums of money in Sam Altman's business after OpenAI debuted the ChatGPT AI chatbot and later the GPT-powered Bing browser. Google's Bard immediately after that. Both Google and Microsoft are currently integrating AI-powered tools into their main offerings.

Conclusion

Although the generative AI explosion is here to stay, more people aren't downloading the ChatGPT and Microsoft Bing apps lately. According to a Bank of America research that used Sensor Tower data, ChatGPT downloads on iPhones in the US fell by 38% in June. Downloads of the Bing app, which in the U.S. includes a ChatGPT-based chatbot, decreased by 38% in June. According to online analytics company Similarweb, the chatbot's website had a 9.7% decline in overall traffic and a 5.7% drop in unique visitors during the same month.