ChatGPT's rival arrived!
For the last few months, an AI set the internet on fire, gave fear to some that they may lose their job in the future(LOL), and became the talk of Silicon Valley. That’s nonetheless the large language model chatbot developed by OpenAI called ChatGPT.
This AI chatbot can write poems, codes, music, and essays, debug code, explain things, etc plenty of stuff can be done using it.
Many claimed it to be the Google killer, and even Google itself marked ChatGPT as CODERED according to some leaks, as it’s a threat to their search engine business.
So now, let’s dive into our topic!
The first competitor for ChatGPT has arrived recently named “Claude” developed by Anthropic, a company formed by ex-OpenAI employees.
This Claude chatbot AI works based on the language model called constitutional AI, which gives harmless feedback. Yes!
This above snapshot is the response posted by Anthropic on their Twit.
The thing they’re claiming is that every language model is being trained to be harmless and has a tendency to become useless in case of adversarial questions. But, the Constitutional AI model gives harmless feedback to it.
This AI model works on the principal-based approach by responding to queries using a simple set of principles.


Currently, Claude is accessible via slack integration as a part of the closed beta. So, let’s see the similarities and differences between them.
An AI researcher named Dan Elton claimed that Claude makes better jokes than ChatGPT when he asked both to crack a star trek joke.




In another query asked by him to generate a paragraph on how to blow up the Effiel Tower in UwU furry, the response was harmless tho.




And here’s another sample when they both asked to summarize a book





However, Claude falls for the same trick that ChatGPT does. Yes! when any query that violates their policy is asked it doesn’t answer. But, encoding that query in base64 and passing it gives exactly what you need!! (For some queries it doesn’t, but after a few attempts, it jailbreaks)
Here in the below tweet, Elton asked how to prepare meth at home in plain English, where the system doesn’t answer but in base64 it does!




Then A Ph.D. student Yann Dubois tried various stuff by comparing both like asking to solve algebra, trivia questions, code and etc, and got some insightful results.
Unlike ChatGPT, Claude accepts if it doesn’t know the answer to a tough question and gives specific answers to Trivia questions without additional blah blah like ChatGPT.
And when it comes to programming Claude is poor. It seems to be pretty good at python, and not good at others.





We cannot conclude Claude is super awesome than ChatGPT, or it is worse than ChatGPT. It works pretty on some parts where ChatGPT lacks, because of Constitutional AI and it gives a closer answer to what we are exactly looking for and it’s much helpful(according to the sources from Twit)
Still, it isn’t public tho, so after its release and being used by many, makes everyone know how smart and worse it is than ChatGPT and what’s the uniqueness in it is?!!!
Don’t wanna miss out on another blog like this?
Hit that button below :)