Krutrim AI Gives Incorrect Prompts To Users | What's The Krutrim Controversy? | N18V | CNBC TV18

CNBC-TV18
1 Mar 202402:19

TLDRThe AI chatbot developed by Ola's founder Bhavish Agarwal, known as 'Crum', entered public beta testing on February 26th. Despite being promoted as a multilingual, advanced language model, it has faced user criticism for inaccuracies. Tests revealed that Crum provided incorrect answers on various topics, including historical facts and current affairs, unlike OpenAI's ChatGPT and Google's Gemini. The chatbot's struggles have raised questions about its training and the significant investment in the startup, which became India's first unicorn of 2024 after a $50 million funding round.

Takeaways

  • 🚀 Ola founder Bhavish Agarwal's AI startup has launched a chatbot for public beta testing.
  • 🌐 The chatbot is the first product of the company, powered by multilingual large language models.
  • 🔍 The AI chatbot is compared to OpenAI's GPT and Google's Gemini (formerly known as Bodo).
  • 🤔 Users have criticized the chatbot for not being properly trained and providing inaccurate results.
  • ❌ The chatbot incorrectly answered questions about the 1983 Cricket World Cup winner and Hillary Clinton's presidency.
  • 📅 It also struggled with basic questions like today's date and historical facts.
  • 🏆 When asked about India's first billion-dollar startup, the chatbot incorrectly identified its parent company, Ola.
  • 💡 In reality, InMobi was India's first billion-dollar startup.
  • 💰 The startup has attracted significant investment, with Ola becoming India's first unicorn of 2024 after raising $50 million.
  • 🔗 The investment came from Matrix Partners India, an early backer of Agarwal's ventures, including Ola Cabs and Ola Electric.
  • 📺 The report was presented by Ishwar Anand for CNBC TV8.

Q & A

  • What is the name of the AI chatbot launched by Ola's founder Bhavish Agarwal?

    -The AI chatbot is not explicitly named in the script, but it is a product of the startup launched by Bhavish Agarwal.

  • When was the AI chatbot released for public beta testing?

    -The AI chatbot was released for public beta testing on February 26th.

  • What language capabilities does the AI chatbot claim to have?

    -The AI chatbot is powered by multilingual large language models, suggesting it can handle multiple languages.

  • Which other AI chatbots does the script compare the new chatbot to?

    -The script compares the chatbot to OpenAI's ChatGPT and Google's Gemini (formerly known as Bodo).

  • What issues have users faced with the AI chatbot after its beta launch?

    -Users have criticized the chatbot for not being properly trained, producing inaccurate results in various areas such as general queries, translation, mathematical problems, and logical reasoning.

  • What incorrect answer did the chatbot give about the Cricket World Cup in 1983?

    -The chatbot incorrectly stated that the West Indies team won the Cricket World Cup in 1983, when it was actually India.

  • What other historical error did the chatbot make regarding US politics?

    -The chatbot erroneously claimed that Hillary Clinton became the US president by defeating Donald Trump in 2014.

  • What basic questions did the chatbot struggle to answer?

    -The chatbot struggled to answer questions about today's date, the government in India in 2014, who killed Mahatma Gandhi, and who was India's first woman prime minister.

  • What incorrect claim did the chatbot make about India's first billion-dollar startup?

    -The chatbot incorrectly claimed that its parent company, Ola, was the first unicorn of the Indian startup ecosystem, when it was actually InMobi.

  • How did the performance of the chatbot raise concerns about its model?

    -The series of inaccuracies by the chatbot raised questions about the effectiveness of its model, especially considering the significant investment in the startup.

  • What was the significance of Ola becoming a unicorn in 2024?

    -Ola became India's first unicorn of 2024 after raising $50 million from Matrix Partners India, an early backer of Agarwal's other ventures, Ola Cabs, and the IPO-bound Ola Electric.

Outlines

00:00

🤖 Launch and Criticism of Ola's AI Chatbot

Ola's founder Bhavish Agarwal's startup, Trm, released an AI chatbot for public beta testing on February 26th. The chatbot, the company's first product, is claimed to be powered by multilingual large language models and is compared to OpenAI's GPT and Google's Gemini. However, it has faced criticism for inaccurate responses across various topics, including general queries, translation, mathematical problems, and logical reasoning. The chatbot's errors include incorrect answers about the 1983 Cricket World Cup winner and the US presidential election in 2014. It also struggled with basic questions and provided false information about India's startup ecosystem.

Mindmap

Keywords

💡AI chatbot

An AI chatbot is an artificial intelligence application designed to simulate conversation with human users. It uses natural language processing and machine learning to understand and respond to user inputs. In the video, the AI chatbot developed by Ola's founder is being tested for its capabilities and accuracy, which is central to the narrative as it highlights the performance issues faced by the chatbot.

💡Public beta testing

Public beta testing refers to the phase where a software or application is released to the general public for testing before its final release. This process helps developers identify and fix issues. The video discusses the AI chatbot's public beta testing, which is significant because it reveals the chatbot's shortcomings in real-world usage.

💡Multilingual large language models

These are advanced AI systems trained on a vast amount of text data in multiple languages, enabling them to understand and generate text in those languages. The video mentions that the chatbot is powered by such models, which is crucial for its ability to communicate with users in various languages, although it seems to struggle with accuracy.

💡OpenAI GPT

OpenAI's GPT (Generative Pre-trained Transformer) is a state-of-the-art language prediction model that can generate human-like text. It is used as a benchmark in the video to compare the performance of the AI chatbot, showcasing the gap between the chatbot's capabilities and those of leading AI models.

💡Google's Gemini

Google's Gemini, formerly known as BERT, is another advanced language model developed by Google that excels at understanding the context of words in a sentence. It is mentioned in the video as a comparison to the AI chatbot's capabilities, emphasizing the chatbot's performance issues.

💡Inaccurate results

In the context of the video, inaccurate results refer to the incorrect information provided by the AI chatbot in response to user queries. This is a critical issue that the video highlights, as it affects the chatbot's reliability and usefulness.

💡Cricket World Cup 1983

The Cricket World Cup 1983 is a historical event in the sport of cricket, where India won the championship. The video uses this as an example to demonstrate the chatbot's failure to provide correct historical information, which is a significant flaw in its performance.

💡Hillary Clinton and Donald Trump

Hillary Clinton and Donald Trump are prominent political figures in the United States. The video mentions an incorrect statement made by the chatbot about their political history, specifically the year of a presidential election, which showcases the chatbot's inaccuracies in providing factual information.

💡India's first woman prime minister

India's first woman prime minister refers to Indira Gandhi, who held the position from 1966 to 1977 and again from 1980 until her assassination in 1984. The video uses this question to test the chatbot's knowledge of historical facts, further illustrating its inaccuracies.

💡India's first billion-dollar startup

The term refers to the first Indian startup to achieve a valuation of one billion dollars or more. In the video, the chatbot incorrectly claims that its parent company, Ola, was the first unicorn in the Indian startup ecosystem, which is factually incorrect, as InMobi holds that title. This highlights the chatbot's lack of accurate information about significant business milestones.

💡Investment and独角兽 (unicorn)

An investment is the act of committing money, time, or other resources to an endeavor with the expectation of a profit. A 'unicorn' is a term used in the business world to describe a privately held startup company valued at over $1 billion. The video discusses the significant investment in the AI startup, which became India's first unicorn of 2024, emphasizing the high expectations and the reality of the chatbot's performance.

Highlights

Ola founder Bhavish Agarwal's AI startup launches a chatbot for public beta testing.

The AI chatbot is the first product launched by the company.

The chatbot is powered by multilingual large language models.

It is compared to OpenAI's ChatGPT and Google's Gemini (formerly known as Bodo).

Users criticize the chatbot for not being properly trained.

The chatbot produces inaccurate results in various categories like general queries, translation, mathematical problems, and logical reasoning.

The chatbot incorrectly answers who won the Cricket World Cup in 1983.

Google's Gini and ChatGPT provide the correct answer to the Cricket World Cup question.

The chatbot also incorrectly states that Hillary Clinton became the US president in 2014.

The chatbot struggles with basic questions like today's date and historical facts.

The chatbot incorrectly claims that Ola is India's first billion-dollar startup.

InMobi was actually India's first billion-dollar startup.

The chatbot's inaccuracies spark questions about its model and the significant investment in the startup.

Kum became India's first unicorn of 2024 after raising $50 million from Matrix Partners India.

Matrix Partners India is also an early backer of Agarwal's other ventures, Ola Cabs and the IPO-bound Ola Electric.

The report is by Ishwar Anand for CNBC TV18.