ChatGPT Fails UPSC Prelims
AIM Events
Hosting the World’s Most Impactful AI Conferences & Events. For Brand collaborations write to [email protected]
ChatGPT, a fulgent star in the classroom, has displayed sheer brilliance by passing all the major exams of the world (a parameter to judge intelligence in modern times). The outlier passed an MBA exam in the US, scored better than most students in the US medical exam and also cleared the?Google Coding Interview?for Level 3 Engineers. The newbie enjoyed a good run at outwitting humans in almost every field until, of course, it decided to take the dreaded Civil Services Examination – one of the world’s toughest exams conducted by UPSC.?
The only difference, here we became the examiners and tested ChatGPT on behalf of UPSC, based on the Prelims paper for 2022. Of the 100 questions from?Paper?1 (Set A), ChatGPT was able to answer only 54 correctly.
In?2021, the cut-off for this exam was around 87%, so going by that benchmark, ‘ChatGPT did not clear USPC’.
The paper tested the bot on varied subjects like Geography, Economy, History, General Science, Current Affairs, Social Development, and Indian Polity & Governance. The experiment sealed the hallucinatory nature of LLM-based chatbots that often get confused and give wrong answers.
Meanwhile, researcher Christian Terwiesch published the?white paper?titled, ‘Would ChatGPT Get a Wharton MBA?’ The paper noted that ChatGPT performed exceptionally well in answering fundamental queries within its area of expertise. However, it occasionally made unexpected errors in middle-school-level calculations.
There’s another important point which needs to be mentioned. In the UPSC exam experiment, we discovered that with a few changes in the prompt, ChatGPT does sometimes arrive at the correct response. This means that in certain cases, asking the same question differently could lead ChatGPT to giving the right answers. It’s all a matter of how you steer it and that brings the importance of a ‘prompt engineer’ to the fore. But that’s a topic for another day…
For now, read this one?here.
English, the default language for programming?
English is the only language you need to know in order to write computer programs. And now, with the advent of generative AI platforms like ChatGPT, English has captured the centre stage. A simple command in English can now give you an entire code base. However, there’s an attempt to break the language barrier and programmers are experimenting with other languages like Russian, Tamil, etc.
领英推荐
Read the full story?here.?
AI Threat is Real
The AI threat is looming large with generative platforms like DALL.E, Midjourney, ChatGPT and others posing threat to many a jobs now. However, Indian IT companies seem to be undeterred by the changing scenario and have assured employees that they shouldn’t be bothered by it. Experts, however, have raised concerns about programmers’ jobs being on the line. Meanwhile, a recent report by JP Morgan said that generative AI can be a “deflation driver” in the near term on legacy services as they compete on pricing, necessitate staff retraining and drive loss of competitiveness.
Read the full story?here.
Sam Altman is No More on Spotlight
Since the launch of ChatGPT, all discussion has been centred around Sam Altman, the CEO of OpenAI. However, he is not the only brain behind the generative AI platform. The latest name from the company to be enjoying the spotlight currently is Mira Murati. Murati is an American citizen of Albanian descent and has previously worked with Zodiac Aerospace, Tesla and Leapmotion before joining OpenAI in 2018. Her journey at OpenAI started as a senior vice president in the company and she was later promoted to the position of CTO, 10 months ago.
Read the full story?here.?