Velox commentary
17 Jan 2023

Why OpenAI/ ChatGPT is not a code red for Developers

If your business revenues are quite dependent on “traditional” web search (sorry Google) then the maturation of OpenAI is most definitely a CODE RED, DEFCON1, THREAT LEVEL : CRITICAL or if you prefer the immortal words of Edmund Blackadder; “This is a crisis. A large crisis. In fact, if you got a moment, it's a twelve-story crisis with a magnificent entrance hall, carpeting throughout, 24-hour portage, and an enormous sign on the roof, saying 'This Is a Large Crisis'".

Even the most skilled searchers will spend quite a bit of time searching refining key words and searching again. ChatGPT has already read everything you could possibly read, has a better idea what you “meant” in your search terms and can organize its results into an actual answer to the question.

Being able to type "what is the capital of Ecuador" in google and get “Quito” as the answer has been around for a while, OpenAI/ChatGPT takes that to a whole new level and then some.

But it’s when the input / question is unambiguous that the OpenAI answer will be precise, which means it is not very well suited to dealing with situations that software engineers find themselves in all the time:

1. The problem to be described is complex (like build me an enterprise ready algo engine) rather than a smaller problem that can be well defined, like write me an efficient bubble sort;

2. The key success criteria of the project are in the head of a human or even worse multiple people;

3. There are high costs of failure or outages. AI can only give you the best possible answer based on what it knows, or the information surface that it has learnt from, if that’s inadequate you will get an answer anyway it just could be anything;

4. The project needs to be delivered within certain top-down constraints (like cost, time) that require judgment calls to be made and tradeoffs.

This is not to say that OpenAI won't make developers more productive because it totally will. There are many many tasks that engineers must undertake that don’t fall into the four above, tasks that can often involve web searches if you can narrow down the search terms. GPT-3 has already changed the game in this regard providing, faster, better results than traditional search and that’s with GPT-4 waiting in the wings (100 times more powerful than GPT-3 with a quantum leap in parameter size).

Link to the original article here.

Find out more about velox and how we #buildfaster here.

#technology #ai #veloxcore #vcore

Do you need to be faster?

Contact us to find out how we are helping our clients build|go|innovate faster and get a free evaluation
Contact us
New York