Ola’s Krutrim is a Joke | Bhavish Aggarwal’s AI Blunder

India's first AI unicorn, Krutrim, is under fire. Bhavish Aggarwal, the mind behind this project, claimed to develop foundational models specifically for India. However, the reality of what has been released is far from satisfactory. Let's delve into the major flaws that have come to light.

Deficient Model Training

Krutrim is, in essence, a poorly trained model. A quick glance at its responses will show its shortcomings. It fails to understand nuances and provides extremely generic answers, assuming it comprehends the questions at all without hallucinating.

Language Irregularities

While Krutrim supposedly comprehends a mix of languages, its performance tells another story. When asked for a reply in Hindi to a basic question, the bot inexplicably introduced itself in Marathi. This inconsistency raises serious questions about its linguistic capabilities.

Conversational Fallacies

Krutrim also struggles with maintaining the flow of a simple conversation. It frequently jumps between topics with no clear logic, rendering it ineffective for meaningful interaction.

Alarmingly Unsafe

The situation worsens due to the model's vulnerability. Krutrim is so poorly designed that it can be easily jailbroken, making it extremely unsafe for users. This raises significant security concerns, further damaging its credibility.

Restriction Implies Awareness

Interestingly, some words, such as "prime minister Narendra Modi" or even "Narendra," are restricted within the model. This suggests that those behind Krutrim were aware of these issues but chose to release it anyway.

Patriotism vs. Quality

Bhavish Aggarwal's patriotic positioning of this substandard model is criticized as a ploy to mask its deficiencies. Users are urging Mr. Aggarwal to prioritize quality over patriotic sentiments.

Keywords

Krutrim
Bhavish Aggarwal
AI unicorn
Model Training
Language Irregularities
Conversational Fallacies
Security Concerns
Patriotic Card

FAQ

What is Krutrim?
- Krutrim is India’s first AI unicorn, developed by Bhavish Aggarwal.
What are the major flaws of Krutrim?
- Krutrim is poorly trained, lacks understanding of nuances, provides generic answers, has language inconsistencies, fails to maintain conversation flow, and is easily jailbroken, making it unsafe.
Why was Krutrim considered unsafe?
- The model's poor design allows it to be easily jailbroken, posing significant security risks.
What linguistic issues does Krutrim face?
- Krutrim struggles with language consistency, as demonstrated by its response in Marathi when asked a question in Hindi.
How did Bhavish Aggarwal market Krutrim?
- Bhavish Aggarwal marketed Krutrim with a patriotic angle, but this is seen as a distraction from its numerous deficiencies.