We asked the Initi8 LinkedIn community…

Survey says NO.

Thanks to the release of ChatGPT, which does a good job of imitating human conversation and sounding like it knows what it's talking about, the world of AI has been opened to a new crowd of admirers. Early users even used it to write Python code and reverse engineer shellcode and rewrite it in C. There's no denying the practical applications of AI like ChatGPT, but it begs the question; will it replace developers and programmers in the dystopic way that we've seen in so many Hollywood movies before?

Let's break it down...

Limitations of ChatGPT

Wrong answers

ChatGPT is a large-scale language model that is constantly trained to improve response accuracy. However, since this is a completely new technology, the model has not yet undergone sufficient training. Therefore, the AI ??chatbot may provide wrong answers.

As it’s based on data and not human perspective or intelligence, it can sound coherent but fail to provide critically informed responses. For example, in Dec 2022 when asked which if 1,000 or 1,062, was larger, ChatGPT confidently responded that 1,000 is larger.

Nor reliable for direct programming

OpenAI’s website shows an example of utilising ChatGPT to debug code. To generate responses, prior code is needed as the AI lacks the capability to imitate human based QA. Directly programming with ChatGPT is not without its risks of errors and bugs. OpenAI even acknowledged that ChatGPT can sometimes write “plausible-sounding but incorrect or nonsensical answers.”

Q&A troubleshooting site Stack Overflow, temporarily banned the use of ChatGPT, saying “the average rate of getting correct answers from ChatGPT is too low, the posting of answers created by ChatGPT is substantially harmful to the site and to users who are asking or looking for correct answers.”

Like all machine learning tools—ChatGPT is trained on existing data as it doesn’t possess the ability to do programming well. For software to be created well, engineers need to understand the intended purpose and goals, not just throw out half-baked code!

Training limitations and issues of bias

Like many AI models, ChatGPT has limitations in its training data. Both limitations in the training data and biases in the data can have a negative impact on model output. In fact, ChatGPT has shown a bias when it comes to training minority data groups. Therefore, it is important to improve the transparency of model data to reduce the bias of this technology.

It's not just ChatGPT!

With Google’s recent demonstration of their AI Bard answering a query incorrectly (and sinking parent company shares more than 7% as a result!), AI is still developing and building its capabilities.

To promote Bard, the bot was asked what to tell a child about discoveries from the James Webb Telescope.

It responded that the telescope was the first to take pictures of a planet outside the earth's solar system, when in truth, that honour was claimed by the European Very Large Telescope in 2004 - a mistake quickly noted by astronomers on Twitter!

Although impressive, ChatGPT still has limited effectiveness when it comes to creating more complex software that also relies on human creativity.

It goes without saying that things could change with future developments, but for the time being there is still a wide set of skills required for software development that AI cannot yet replicate. If you require a complex phone app or highly interactive game, developers are far more adept to tackle this.

HOW CAN ChatGPT BE USED BY PROGRAMMERS?

We need programmers, developers and engineers to create software that so many of us rely on. However, that’s not to say there’s no place for AI in software development! AI can potentially serve as a tool for lower-level and minor tasks to reduce time and effort.

This is already happening within big tech. GitHub CoPilot allows developers to use ChatGPT to improve their code and find bugs and Amazon’s CodeWhisperer is designed to help developer productivity with generated code recommendations.

Rather than dismissing or being frightened of ChatGPT, we can recognise it as a valuable tool:
- Outline App builds
- Quickly generate frameworks
- Code debugging
- Input questions around data structuring
- Input questions about interface features
- Accelerate routine tasks

As with all tools, learning how to use it efficiently comes with its own skillset. All your existing coding knowledge and experience can help you select your prompts to produce what you’re looking for.

What are your thoughts? Let us know! 

Are you an Employer or Recruiting?
Sign up to our newsletter for Employers & Recruiters

Find the latest tech tips, guides and hiring secrets within the tech industry, direct to your inbox each month.
Simply drop your email address below to subscribe: