AI LLMs : Are we there yet? - PART 1

PART 1: Are we there yet with using Bard or ChatGPT for code reviews?

These comments are as of 05 April 2023 … I believe that continuous improvements on these new gen AI productivity tools are moving at “Ramming Speed” at Google and OpenAI.

I was so excited to hear that we could use the AI prodigies ChatGPT and Bard for reviewing our code or for generating new code.

As far as generating code is concerned, these tools could be a super saver on time for developers as long as we have the pseudo code correct for the problem statement ?? ,or a code snippet to be peer reviewed.

Nevertheless, I just wanted to checkout the current situation on coding help that these tools could help with, and had a chat with these tools themselves, with a question and an exercise ... that's how most of us get evaluated during job interviews!

a)?????Question 1: Which programming languages and scripting can <tool> generate code?

Bard>

No alt text provided for this image

…that’s a pretty precise statement; and is consistent when trying to regenerate the response by typing in the question again. However, when regenerating the response it was not that nice and precise.

ChatGPT>

No alt text provided for this image

…that’s a pretty precise statement too; and is consistent when trying to regenerate the response by typing in the question again. However, when regenerating the response was not that nice and precise at times and not consistent. The kids got lots to think about…is it too much learning or some other factor?

b)????Exercise 1: Generate a piece of code in Scala for data profiling a csv provided as input.

Bard>

No alt text provided for this image

Oops! Can't generate code in Scala! But Bard said it could when asked earlier!…anyways, Bard also says that it is in experimental phase only. I wonder why Bard would claim that it can generate code in Scala. Is Bard trying to get info on the demand from users and then prioritize their rollouts for capabilities on programming support? But, with less number of users compared to ChatGPT, does Bard have good amount of sampling data? Compared to ChatGPT, Bard having less # of users = less interaction = less learning = less intelligence and can expect not so helpful responses and not so great learning curve. Now, that’s too harsh on Bard; let’s leave the kid alone; still learning! Actually, I had also asked Bard to generate a piece of code on Scala and Bard generated the equivalent of the famous C program by Kernighan…the Hello World program, in Scala. Very basic, but at least that made me happy that the prodigy is hungry. ??.

ChatGPT>

No alt text provided for this image

Nice! Look at the output from this fellow! Not bad at all!

Ofcourse, we cannot judge the artificially induced intelligence based on a couple of questions! Lets keep checking these guys and see how soon they can grow? They could grow faster than a dog; till they are hit by another gamechanger.

Ponder: Are these tools really AI tools or can we only stick to calling them deep Machine Learning tools for now; either way these guys would need so much data to feed; and with so much data consumption would they become fat…can they manage not to become sluggish; do these tools have high metabolism? ??

"Stay Hungry, Stay Foolish : Steve Jobs". Will this apply to these two guys...Bard and ChatGPT (I wish OpenAI could have named this fellow; maybe something like...Jagger or Floyd)

So, for now, I think there is a long way to go when it comes to using these tools for code reviews and generating code...maybe another year. Lets see!


*All comments are my personal opinions and do not mean offence to either of the prodigies…Google’s Bard or OpenAI’s ChatGPT.?

Interesting references:

·??????Google’s Code Red

·??????clip from the classic 1950s movie: Ben Hur

要查看或添加评论,请登录

社区洞察

其他会员也浏览了