Starke Frau .

From last couple of days i am looking at tweets on GPT-3 a OpenAI API in recruitment community. As much i explore to get understanding about it i have few queries maybe because i am not from tech background.

1.Is it hyped much in AI world?
2.Is it reliable with no mistake proof
3.What about copyright or cryptography permission before uploading any text?
4.Can we completly rely on Open AI API for our task?

Many recruitment tools like Cutshort who gives you complete ML model where you can enter your data as per requirements to get desired result .Some companies even for background verification running tools based on AI ML and blockchain.

How much authenticity we can expect from AI?

Originally Posted Here

Siddharth Verma

Here are my 2 cents

1)The hype is real. GPT-3 is a general AI which is a major improvement from its predecessor GPT-2(which was already awesome)

The major reason for hype behind GPT-3 is because it's easier to use

2)No AI is not fool proof that's why a lot of companies use it as an enabler rather than a solution. Like cutshort uses AI to give recommendations not automate the process of hiring.

3) Didn't get you on this one. Can you elaborate ?

4) It's pretty decent I wouldn't rely on it 100%. The motive of introducing an AI in a system is to make ones work easier not to do it for them

The models used by Cutshort and other companies is very different from GPT. It's a generative model not classification or recommendation one.

In recruitment I see GPT used for
-Generating Job Description
-Resume cover letter writing
-Automating outreach to candidates

Starke Frau

Thanks for giving me insight about GPT-3.

3) I was asking about what if upload any data so will it bypass the permission access and what if data is owned by another .

Yes as you said rightly about Cutshort Algo same wity Insytahyre.
They also automatically outreach to candidate as far as i used it.

Yes i also see use of GPT in above mentioned while we talk about recruitment.
Now if we talk about another step which is background verification,Here like Springrole based on AI blockchain module gives you platform to minimize this process by setting up critariea as per requisition. Now if we see GPT -3 in such process .Example
If candidate A go through all rounds and verificatiin comes in but problem occurs
a) No response from "A" previous employers after reminders or from any other colleagues who suppose to verify "A" .In this case it will bypass and may mark as unverified.
b) If "A" one of document which may be pending from previous employer but in system it is mandate to go ahead then you cannot bypass and process further.

Siddharth Verma

You will have full access to you data as far as my knowledge goes.

An outreach created by GPT will be far superior than the current Algos use by existing companies which are more or less rule based.

AI doesn't really works that way ideally it will keep the status as unverified. Unless it is trained to bypass the above use cases via the company

Starke Frau

points which i noted😃are giving me great idea about it.As a Recruiter it is wonderful to know more from you.

Apurva Raj

The hype is real. There should be more and more discussion around it.

It’s still in beta. No system will ever be mistake-proof but GPT will improve drastically in a couple of years.

Currently, they are saying, “You retain ownership of your data, and control whether we can use it for training models.” But no exact idea on how copyright will work. Awaiting to use GTP and connect with their team to understand this fully.

I am gonna rely on it for a lot of research-based tasks as of now. And probably coding too. Not fully though. It will take some time but I defiantly see it happening.

Starke Frau

Exactly .If i say it's still in trial and error mode so hope i m not putting wrong statement.

Training model cost i read somewhere is bit high.

Join India’s
most active community
and interact with 20k+
like-minded entrepreneurs