Will you be replaced by ChatGPT-3?

 For this article in the EITF blog, we felt compelled to address the latest frenzy around chatGPT-3 and AI. Since January, everyone’s LinkedIn walls, including ours, have been spammed with mentions of ChatGPT-3 / AutoML. Most of them talk about how this shiny new development in AI could be the final say, in terms of replacing jobs and human resources.  There has also interestingly been an increased coverage in the media regarding ChatGPT-3 getting candidates through tough medical examinations.

So we at EITF decided to try out and use ChatGPT-3. Our initial assessment is that, while tools like ChatGPT-3 / AutoML have their own advantages in human intelligence augmentation, their adoption in an industry setting will still have to wait for more maturity. 





According to our evaluations, here are 5 reasons  why Generative AI tools like ChatGPT-3, BARD will take time to find adoption

1.        Bias:  A common pitfall in AI/ML is AI bias. This is a phenomenon when an algorithm is prejudiced due to erroneous assumptions.  The chatter about ChatGPT-3 is that it participates in anti-conservative bias. In short, bias is baked into the system.

2.        Outdated Data: From what we have seen and heard, these models have been trained with data from 2021. Therefore, new and trending events will not be covered by the model since it hasn’t been exposed to new events that followed the 2021 time stamp. This will continue to be a problem in the future unless there are changes made to address this in future releases for ChatGPT-3.

3.        Generic Output: While there have been talks about how ChatGPT-3 can help with customer support, in reality it is pretty generic and redundant to chatbots that companies already have. Organizations these days want personalized customer support to stay ahead of the game, and it is hard to remove the human interaction for this case and address this completely with ChatGPT-3.

4.        Contextuality: The data for most organizations these days is contextual and sometimes private. ChatGPT-3 is a generic AI service that has been built on primarily, open source data that isn't domain specific.

5.        Outdated answers: There are instances where ChatGPT-3 is erroneous in its outputs. Our opinion is that deploying this without a human in the loop will result in customers being fed incorrect information. Therefore, there will need to be an extraneous human verification check for the chatGPT-3 output.

In conclusion, while ChatGPT-3 is an interesting entry in the AI space, turning a lot of heads for the right reasons in terms of AI development, our opinion is, for AI adoption to be fruitful, tools like ChatGPT-3 will need to augment the human in the loop in order to reduce the time to take decisions. As fancy as it sounds, tools like AutoML and ChatGPT-3 still have a long way to go to make jobs obsolete. So, don’t panic just yet! 

Follow us on our social media platforms to get more updates:

Follow our LinkedIn page at : https://www.linkedin.com/company/eitf-project/

Our social media platforms: linktree/eitfproject

Comments

Popular posts from this blog

Confessions of an overthinker

Enter the FIRST EITF BLOG CONTEST!

Can you be ambitious at work and avoid negative feedback?