Team-BHP (https://www.team-bhp.com/forum/)

-   Shifting gears (https://www.team-bhp.com/forum/shifting-gears/)

-   -   The Generative AI Thread (chatbot, text to image, text to video etc) (https://www.team-bhp.com/forum/shifting-gears/259867-generative-ai-thread-chatbot-text-image-text-video-etc-4.html)

One still needs to ask right to get good answers
Also, It cant innovate. When I search on google I want to find what is already done and what people think about something. Its a fun play thing for sure. People should relax. Its not going to replace people !

Very interesting to know what ChatAI knows about Team-BHP

I am dreading the day when an overenthusiastic smart aleck uses Chatbot to gain entry into TeamBhp, posts very informative write-ups and gets applauded for his immense knowledge and rare insight!

Will the limited iterations of a human brain stand a chance against the masterful regurgitations of artificial intelligence?

A bot can be detected and deactivated; a human presence identified. But when an unethical man and an ethics-neutral machine are mated together, things get not just different but also extremely dangerous.

Check out this weird conversation I had:

I think it's the reverse. The limited (in creativity/quality, not quantity) regurgitation(s) of AI doesn't stand a chance against the masterful iterations of the human brain. (At least not yet.)

Those leaps of insight, even minor ones, that 'joining the not obvious dots', not just the brilliant insights, but the hundreds of small connections we make unconsciously everyday, are still the preserve of our organic brains.

AI is a tool. Even if someday someone uses it to generate passable posts on Tbhp, (some part of car review data, for example - publishing specs and spec comparos - that is already possible) it'll be more akin to someone using spell-check to correct their posts, rather than a truly non-organic sentient intelligence creating original content.

This thing is frustrating on many levels. I recently spent a day and a half writing code for a small college assignment, and when I discussed it with others I was asked "why didn't you just use ChatGPT?" :Frustrati Can't believe how quickly it has caught on.

ChatGPT is fully aware of the real Jeremy Clarkson. rl:

Context : Amazon cancels Grand Tour - Clarkson at it again

But it isnt a perfect tool yet. Anyone trying to use it for serious work needs to double check the content/quality/accuracy. The scary part it is , it will learn.

ChatGPT is lying when it says it cannot access the internet. Not only can it visit websites, it can instantly analyze the contents of webpage and and give cogent answers:



Context:
https://www.team-bhp.com/forum/shift...e-history.html

I don't think OpenAI claimed that ChatGPT cannot access the internet. They said that the training data set is last updated as of 2021. Here it is just trawling a website and succinctly summarising it.

It does not know of events that happened from 2022 onwards. Can you try asking "What was the impact on Adani group share price because of the Hindenburg research report?"?

Today had to run an experiment on writing an assignment in chatGPT vs checking whether the AI content detector tool finds it as AI content. Out of 5 runs the AI content detector found out it is AI written 4 times. Not bad as of now. Details in linkedin https://www.linkedin.com/feed/update...4981215277057/

ps - ChatGPT is going to be the biggest headache for educational institutions. Donno how are they going to figure out if an assignment is human written or AI written. And how are they going to grade a student if his assignment is AI written? will it be considered plagiarism? Plagiarism is outright copying of someones content say from wikipedia but if i give the right inputs to chatGPT and it generates the content which could or could not be from the same wikipedia page then am i guilty of plagiarism or not? Tough times ahead for teachers.

For me, ChatGPT is a tool, like a calculator. The rest of the system will adapt, recognizing its capabilities. Humans will have an advantage since they can figure out *what* to do & *why* to do it. Machines at best can tell *how* to do.

Would like to address a few concerns raised:
* Coding
We need good clean code. Doesn't matter whether humans write it or machines. On the question about faking during interviews, there are many ways to beat it and the easy one is asking mostly 'what' questions. Machines will give canned responses.

* School assignment
All most all exams are closed book, with pen-paper in an exam hall. If writing skill is what that will be tested then students will prepare for it.
On assignments, everyone uses AI, then all their answers look the same. The one with a different answer gets the marks! Also, the evaluators need to change their tactics. Just like a calculator didn't stop maths, AI won't affect language skills.

PS: I have never written a poem in my life. But here is it, courtesy of AI :).
A community of car enthusiasts,
united by a common bond,
Team-BHP, a forum of peers,
where knowledge is fond.

From the latest news, to DIYs,
reviews and comparisons,
there's always something new to see,
and discussions to have with passion.

From Maruti to Mercedes,
Honda to Hyundai,
all makes and models,
are welcomed with a sigh.

Team-BHP, a place to share,
your love for the open road,
where dreams of horsepower,
can be shared and bestowed.

So come and join the community,
of like-minded folk,
where the love of cars,
is the common stroke.

Today I asked ChatGPT the biggest confusion plaguing off-road buyers-

Mahindra Thar 2WD vs Thar 4WD vs Suzuki Jimny, which is better?

It gave a diplomatic answer at the moment. :)



But what baffled me is the fact even though it gave a diplomatic answer it was very close to what logically one might conclude. Although once indepth review of Jimny comes out will ask the same again and see if it has anything else to add.

That said, some of the folks are saying that ChatGPT has been trained on data set till 2021. But it has given answers on Russia Ukraine war which happened started in 2022. Also, Mahindra Thar 2WD came this month how on earth it knows about that? :D

w.r.t. accessing the Internet , I guess it is true to go with the assumption that whatever it answers is based on its acquired analytics which are as of 2021. Could be that their crawlers are polling for latest information and updating the Open AI servers. So you may get answers to recent happenings or you may not.

For example it was not able to answer these straight forward questions.

Amul special doodle on ChatGPT